Beruflich Dokumente
Kultur Dokumente
ImprovingBEMbasedAerodynamicModels TonioSant
inWindTurbineDesignCodes
AerodynamicModels
inWindTurbineDesignCodes
Thewindenergyindustryisexperiencingremarkablegrowthsannually.
Despite the great progress made, further cost reductions in turbine
technology are necessary for wind energy to reach its full potential in
in
termsofthelargescalesupplyofelectricity.Improvingthereliabilityof
aerodynamicmodelsembeddedinthedesignsoftwarecurrentlyusedin WindTurbineDesignCodes
industry is indispensable to guarantee reductions in the cost of wind
energy.
Duetoitsrelativelyhighcomputationalefficiencycomparedtofreewake
vortex methods and CFD, the BladeElementMomentum theory still
formsthebasisformanyaerodynamicmodels.Yetvariousexperimental
campaigns have demonstrated that BEMbased design codes are not
always sufficiently reliable for predicting the aerodynamic load
distributionsonthewindturbineblades.
ISBN:9789993204831
DelftUniversityWindEnergyResearchInstitute
DUWIND
DepartmentofMechanicalEngineering
FacultyofEngineering
UniversityofMalta
January2007
TonioSant
Improving BEM-based
Aerodynamic Models
in
Wind Turbine Design Codes
Proefschrift
Tonio SANT
Bachelor of Engineering, University of Malta
geboren te Malta
Toegevoegd promotor:
Dr. ir. G.J.W. van Bussel
Samenstelling promotiecommissie:
This PhD thesis was supported by the University of Malta and Delft University of
Technology.
ISBN: 978-99932-0-483-1
All rights reserved. Any use or application of data, methods or results from this thesis will
be at the users own risk. The author accepts no liability for damage suffered from use or
application. No part of this book may be reproduced in any form, by print, copy or in any
other way without prior permission from the author.
TomywifeMarisandfamily
Summary
Duetoitsrelativelyhighcomputationalefficiencycomparedtofreewakevortexmethods
andCFD,theBladeElementMomentumtheorystillformsthebasisformanyaerodynamic
models integrated in design software for horizontalaxis wind turbines. Yet various
experimental campaigns have demonstrated that BEMbased design codes are not always
sufficientlyreliableforpredictingtheaerodynamicloaddistributionsonthewindturbine
blades.Thisisparticularlytrueforstalledandyawedrotorconditions.Thus,itispresently
necessary to try to improve BEM methods in order to provide more costeffective wind
turbinedesignsandhencereducethecostofwindenergy.
In this thesis, a detailed investigation of the aerodynamics of wind turbines in both axial
andyawedconditionsiscarriedoutbasedonwindtunnelmeasurementswiththeaimof
providingabetterunderstandingofthelimitationsoftheBEMtheory.Thisworkidentifies
the need to pursue turbine aerodynamics and modelling in an integrated approach,
emphasisingontheneedtounderstandthelocalbladeaerodynamics,inflowdistributionas
wellasthegeometryandvorticitydistributionofthewake.Toenablethisapproach,new
methodologies and analytical models are developed which compensate for the limitations
inexperimentaldata.Amongthesemodelsisafreewakevortexcode,whichisbasedona
prescribed bound circulation distribution over the rotor blades. This freewake vortex
modelHAWTFWCisdevelopedandvalidatedusingthehotfilmandthetipvortexsmoke
visualisationdatafromtheTUDelftrotorexperimentsandwillbeusedtogenerateinduced
velocitydistributionsforthemeasuredaerodynamicloaddistributionsattheNRELPhase
IVwindturbine.
In this research project, the aerodynamics of two different wind turbine rotors is
investigated:
(1) The TUDelft model rotor which is tested in the openjet wind tunnel facility at Delft
University of Technology. Detailed hotfilm measurements are performed in the near
wake of the model rotor when operating in attached flow conditions over the blades
(low angles of attack). The measurements are taken at different planes parallel to the
rotorplane,bothupstreamanddownstream.Smokevisualisationexperimentsarealso
carriedouttotracethetrajectoriesoftipvortexcoresintherotorwakeinattachedand
stalled flow conditions. These experiments are carried out in close collaboration with
WouterHaans,aPhdcolleagueatTUDelft.However,duetolimiteddimensionsofthe
bladesitisnotpossibletomeasuretheaerodynamicbladeloaddistributionsdirectly.
By applying bladeelement theory a methodology is developed to estimate the time
dependent aerodynamic load distributions at the rotor blades from the hotfilm
measurements:
(i) initially, inflow velocities at the blades are estimated from the hotfilm
measurementstakenatthedifferentplanesparalleltotherotorplane.
(ii) theresultsfromstep(i)areusedtoderivethesteady/unsteadyangleofattack
andtherelativevelocitydistributionsattheblades.
(iii) theresultsfromstep(ii)areusedinanadvancedunsteadyaerofoilmodelto
yieldthedistributionsforboundcirculationandaerodynamicloadingatthe
blades.Anewandefficientnumericalmethodforimplementingthisaerofoil
modelinrotoraerodynamicscodesisdeveloped.
Both the inflow measurements and the derived aerodynamic loads on the TUDelft
rotorareusedtocarryoutadetailedinvestigationoftheBEMtheorywhenmodelling
both axial and yawed conditions. Two different approaches are applied: the first
approach in which the inflow measurements and aerodynamic loads are used to
compute separately the momentum and bladeelement theory parts of the BEM
equation for axial thrust. The discrepancy between the two parts is a measure of the
incapability of the BEM theory to model axial or yawed conditions. In the second
approach,atypicalBEMcodeisemployedtomodeltheTUDelftrotorandtheresults
are compared with those obtained from the hotfilm measurements. Despite the fact
that only attached flowconditions are being studied andalso the fact that the results
derived from the inflow measurements have a rather high level of uncertainty in
general, this comparison results in a better understanding of the limitations of BEM
baseddesigncodesandfurtherinsightisobtainedofhowthesecanbeimproved.
(2) The NREL Phase VI wind turbine which was extensively tested in the NASA Ames
wind tunnel in 2000. The experimental data required for the study are obtained from
the NREL. This data consists of timeaccurate blade pressure measurements for the
rotoroperatinginbothaxialandyawedconditionstogetherwithmeasurementsofthe
localflowanglesmeasuredatdifferentradiallocationsinfrontofthebladesusingfive
holeprobes.Theexperimentaldataalsoconsistsofstraingaugemeasurementsforthe
output torque and the root flap/edge moments. However detailed inflow
measurementsattherotorarenotperformed.Inthisthesis,anovelandcomprehensive
methodology is presented for using the blade pressure measurements in conjunction
with the freewake vortex model HAWTFWC to estimate the angle of attack
distributions at the blades more accurately, together with the induced velocity
distributions at the rotorplane and wake geometry. This methodology consists of the
followingsequenceofsteps:Initially,aspanwisedistributionfortheangleforattackis
assumedattheblades.ThisisthenusedtogetherwiththevaluesofCnandCtobtained
from the blade pressure measurements to estimate the lift coefficients at the blades.
UsingtheKuttaJoukowskilaw,theboundcirculationdistributionatthebladesisthen
determined and prescribed to HAWTFWC to generate the free vortical wake. The
inducedvelocityatthebladesisestimatedandusedtocalculateanewangleofattack
distribution. The process is repeated until convergence in the angle of attack is
achieved.Oneadvantageforapplyingthismethodologyisthatthesolutionisinitself
unsteady and could be readily applied to study yawed conditions, under which
complexaerodynamicphenomenaareknowntooccur(e.g.dynamicstallandunsteady
induction).Asecondadvantageconcernsthefactthatthewakegeometryisinherently
partofthesolution.Thusitispossibletoderivethepitchandexpansionofthehelical
wake from the measured Cn and Ct, which otherwise can be obtained using time
consuming smoke visualisation experiments. The threedimensional vorticity
circulationdistributioninthewakecanalsobeinvestigatedunderdifferentoperating
conditions.
Usingtheabovemethodology,new3Dliftanddragaerofoildataarederivedfromthe
NREL rotor blade pressure measurements. This new data is considerably different
from the corresponding 2D wind tunnel aerofoil data due to the presence of blade
tip/root loss effects, stalldelay or else unsteady conditions resulting from rotor yaw
(mainly dynamic stall). Thenew 3D lift and drag aerofoil data is then used improve
BEM load predictions in axial and yawed conditions. It is found that with this new
data, the BEM predictions improved considerably even when dealing with highly
stalledandyawedconditions.Foryawedconditions,newinflowcorrectionstoaccount
forskewedwakeeffectsinBEMcodesarealsoderived.
From this research, it is possible to draw guidelines on how BEMbased models can be
improved.Theseguidelinescanbesummarisedintwo:
(1) Improvement of aerofoil data: It is clear from this study that BEM predictions improve
substantially when more accurate 3D aerofoil data is used. In this thesis, a new
engineering model for 3D lift and drag coefficients in axial conditions is developed
based on the measurements on the NREL rotor. A similar model for unsteady
conditionsisnotdevelopedsincetheamountofderivedunsteadyaerofoildatawasto
a certain extent limited. Yet this data is very useful for other researchers to develop
suchimprovedmodels.
(2) Improvement of engineering models for skewed wake effects in yaw: The BEM theory is
incapablewithmodellingtheeffectsofaskewedwakeontheinductionattheblades
thatresultinyawedrotors.Variousengineeringmodelstocorrectforthisincapability
aredevelopedinthepastyearsandareusedinstateoftheartdesigncodes.Yetthis
study has demonstrated that such models are limited for two reasons and better
modelsarerequired:
(a)firstofall,theunsteadyandperiodicinductiondistributionatthebladesresulting
fromrotor yawmay havea higherharmoniccontentthanthatcateredforbythese
presently available engineering models. Also the unsteady distributions are
dependentnotonlyontheyawanglebutalsoontheoperatingtipspeedratioand
rotor geometry. This study has shown that because the aerodynamics of yawed
rotors is complicated, it is vital to introduce more theoretically comprehensive
models. An approach is proposed for interfacing BEMcodes to prescribedwake
vortexmodelswhentreatingyawedconditions;
(b)secondly,thecurrentlyavailablemodelsonlycorrectthelocalaxialinductionatthe
bladestothecorrespondingannularaveragedvalue.Thisstudyshowsthat,dueto
the deficiency of the axial momentum equation in yaw, the annularaveraged axial
induction computed by BEM may also need to be corrected. It is found from the
analysisonboththeTUDelftandNRELrotorsthatthiscorrectionnotonlydepends
on the rotor yaw angle but also on the operating axial thrust coefficient. An
engineeringmodeltomodelthiscorrectionisrequiredifBEMpredictionsinyawed
rotorsaretobeimproved.
Samenvatting
(i)eerstzijndeinstroomsnelhedenterplekkevandebladengeschatuithittefilm
metingenindeverschillendevlakkenparallelaanhetrotorvlak.
(ii)deresultatenuitstap(i)zijngebruiktomdestationaire/instationaire
invalshoekendeverdelingvanderelatievesnelhedenterplekkevandebladen
afteleiden.
(iii)deresultatenuitstap(ii)zijngebruiktineengeavanceerdinstationairmodel
voorprofielaerodynamica,omdeverdelingenvandegebondenwervelsterkte
endeaerodynamischebelastingenopdebladentebepalen.Voorhet
implementerenvanditinstationalireprofielmodelinrotoraerodynamicacodes
iseennieuweefficintemethodeontwikkeld.
Zowel de instroommetingen als de berekende aerodynamische belastingen op de
TUDelftrotorzijngebruiktomeengedetailleerdonderzoekuittevoerennaardeBEM
theorievoorhetmodellerenvanenrechteenscheveaanstroming.Tweeverschillende
methoden zijn gevolgd: de eerste methode waarbij de instroommetingen en de
aerodynamische belasting gebruikt worden om afzonderlijk de impuls en de blad
elementtheoriewaardenvoordeaxiaalkrachtindeBEMvergelijkingteberekenen.De
discrepantie tussen deze twee waarden is een maat voor de toepasbaarheid van de
BEMtheorievooraxialeofscheveaanstromingscondities.Indetweedemethodeiseen
echte BEMcode gebruikt om de TUDelft rotor te modelleren en zijn de resultaten
vergelekenmetdieuitdehittefilmmetingen.Ondankshetfeitdatenkelaanliggende
stromingsconditiesbestudeerdzijnenhetfeitdatderesultaten,diebepaaldzijnuitde
instroommetingen,inhetalgemeentamelijkonnauwkeurigzijn,leiddedittoteenbeter
begrip van de beperkingen van op BEM gebaseerde ontwerpcodes en tot een verder
inzichtinhoedezetekunnenverbeteren.
(2) De NREL Fase VI windturbine die uitvoerig getest werd in the de NASA Ames
windtunnel in 2000. De experimentele data, die benodigd zijn voor het onderzoek,
werden verkregen van het NREL. Deze data bestaan uit tijdsafhankelijke
drukmetingen op het roterende blad in zowel rechte aanstroming als scheefstand,
gecombineerd met metingen van de locale stromingshoek, gemeten op verschillende
radialepositiesvoordebladenmetvijfgatsdrukmeters.Tijdensdeexperimentenzijn
ook rekstrookmetingen van het koppel en de klap en zwaaimomenten aan de
bladwortel verricht. Gedetailleerde instroommetingen aan de rotor zijn echter niet
uitgevoerd.Indezedissertatieiseennieuweenuitgebreidemethodetoegepastomde
invalshoekverdelingenopdebladen,deverdelingenvandegenduceerdesnelheidin
hetrotorvlakendezoggeometrienauwkeurigertebepalendoorgebruiktemakenvan
de drukmetingen op het blad in combinatie met het vrijewervel zogmodel HAWT
FWC,.Dezemethodebestaatuitdevolgendereeksstappen:eerstwordteenverdeling
inspanwijdterichtingvandeinvalshoekterplaatsevandebladenaangenomen.Deze
wordt dan gebruikt om, samen met de Cn en Ctwaarden verkregen uit de
drukmetingen op de bladen, de liftcofficinten op de bladen te schatten.
Uit dit onderzoek volgen richtlijnen voor de wijze waarop op BEM gebaseerde modellen
verbeterdzoudenkunnenworden.Dezerichtlijnenkunnenalsvolgtwordensamengevat:
(1) Verbetering van profieldata: het blijkt duidelijk uit deze studie dat BEM voorspellingen
aanzienlijk verbeteren wanneer nauwkeuriger driedimensionale profieldata gebruikt
worden.Indezedissertatieiseennieuwengineeringmodelvoordriedimensionalelift
en weerstandscofficinten in rechte aanstroming ontwikkeld gebaseerd op de
metingen aan de NREL rotor. Een vergelijkbaar model voor instationaire
omstandigheden is niet ontwikkeld omdat er onvoldoende instationaire profiel
gegevens afgeleid konden worden. Desalniettemin zijn deze data erg bruikbaar voor
andereonderzoekersomverbeterdemodellenoptestellen.
(2) Verbeteringvanengineeringmodellenvoordeeffectenvanhetzoginscheefstandcondities:de
BEM theorie is niet in staat de effecten van een scheef zog op de inductie ter plaatse
vandebladentemodelleren.Verscheideneengineeringmodellentercorrectiehiervan
(b)tentweedewordtindehuidigemodellenalleendelocaleaxialeinductiesnelheidter
plaatse van de bladen gecorrigeerd met de gemiddelde waarde over de annulus.
Dezestudielaatziendat,tengevolgevaneenfoutindeaxialeimpulsvergelijkingin
scheefstand,degemiddeldeaxialeinductieoverdeannulusberekenddoorBEMook
gecorrigeerdmoetworden.UitdeanalysevanzoweldeTUDelftalsdeNRELrotor
blijktdatdecorrectienietslechtsafhangtvandescheefstandhoekvanderotormaar
ook van de axiaalkrachtscofficint. Een gecoorigeerd engineering model is nodig
omBEMvoorspellingenbijscheefstaanderotorenteverbeteren.
TableofContents
Acknowledgements.........i
Abbreviations..........ii
Nomenclature..........iii
ListofDevelopedSoftwareCodes...............xi
Chapter1Introduction........1
1.1TheRoleofAerodynamicsinWindTurbineDesign......2
1.2PrinciplesofHAWTAerodynamics..........5
1.3 CurrentStatusofAerodynamicDesignModelsfor
HorizontalAxisWindTurbines........8
Chapter2AimofThesisandApproach..11
2.1ProblemStatement.......11
2.2AimofThesis...................14
2.3Approach.................15
2.3.1ResearchworkontheTUDelftwindtunnelturbine......15
2.3.2ResearchworkontheNRELPhaseVIwindtunnelturbine...15
2.3.3DevelopmentofFreewakeVortexModel....17
2.4OrganizationofWork.............18
2.5CoordinateSystemsAnalysis.......20
Chapter3TheBladeElementMomentumTheory..........23
3.1 TheSimpleLinearMomentumTheoryforaYawed
ActuatorDisc..........23
3.2TheMomentumEquations........25
3.3TheBladeElementTheory............26
3.3.1TheBladeElementVelocity........26
3.3.2TheInducedVelocityatEachBladeElement..........26
3.3.3FlowVelocityRelativetoaMovingBladeElement.........27
3.3.4AerodynamicLoads.............28
3.3.5TheBladeElementEquationsforThrustandTorque....................29
3.4TheBladeElementMomentumEquations.........30
3.5CorrectionstotheBladeElementMomentumTheory.....31
3.6DescriptionofProgramHAWT_BEM...........40
3.6.1TimebasedNumericalSolutionfortheBEMEquation.........40
3.6.2ProgramStructure............43
Chapter4AerodynamicAnalysisoftheDelftModelTurbine..................45
4.1Introduction.................45
4.2WindtunnelMeasurements.............48
4.2.1WindTunnel.................48
4.2.2PartI:InflowMeasurements.......49
4.2.3PartII:SmokeVisualizationMeasurements...........76
4.3EstimatingtheAerodynamicLoadsattheBlades
fromtheWindTunnelInflowMeasurements.........87
4.3.1Methodology................87
4.3.2ATheoreticalMethodforFindingtheUnsteadyLift
CoefficientinAttachedFlowforaRotatingBladeinYaw....90
4.3.3DevelopedSoftwareTools..........98
4.3.4AssessmentofWindTunnelBlockageEffects.............110
4.3.5QuantificationofErrorsinDerivingBladeLoadingdueto
ErrorsinInflowMeasurements........121
4.3.6ResultsandDiscussion.................128
4.4InvestigatingtheLimitationsoftheBEMTheoryfor
Axial/YawedWindTurbines...........................182
4.4.1ResultsfromApproachA..............183
4.4.2ResultsfromApproachB...........188
4.5Conclusions............207
Chapter5DevelopmentofaFreeWakeVortexModel..........211
5.1Introduction.......211
5.2FreewakeVortexNumericalModel......................212
5.2.1BladeModel...............212
5.2.2NearWakeModel..........213
5.2.3FarWakeModel........221
5.2.4NumericalSolution...........222
5.2.5ProgramStructure........226
5.3VerificationandValidationofFreeWakeVortexModel........228
5.3.1VerificationandValidationMethodology.........228
5.3.2ResultsandDiscussion..............234
5.4Conclusions.............274
Chapter6AerodynamicsAnalysisoftheNRELPhaseVIWindTurbine......277
6.1Introduction.......277
6.2NASA/AmesUAEWindTunnelDataUsed.........279
6.3EstimatingtheAngleofAttackfromBladePressure
MeasurementsusingtheFreeWakeVortexModel.....283
6.3.1Methodology..............283
6.3.2AxialConditions.......285
6.3.3YawedConditions.............320
6.4BEMPredictionsfortheNRELPhaseVIRotorwithNew
AerofoilDataandInflowCorrections...........362
6.4.1AxialConditions.............363
6.4.2YawedConditions.............373
6.5Conclusions..........393
Chapter7GuidelinesforImprovingtheReliabilityofBEMbasedDesign
Codes..........397
7.1ModellingofAerofoilData..........397
7.2Correctingforskewedwakeeffectsinyaw...403
7.2.1Developmentofengineeringmodelusingaheuristic
approachbasedonFourierseries..........404
7.2.2Alternativeapproachusingprescribedwakevortex
model........405
7.2.3Correctiontotheaxialmomentumequationfor
yawedrotors.............411
Chapter8ProjectOutcome,ConclusionsandRecommendations................413
8.1ProjectOutcome........413
8.2Conclusions........416
8.3Recommendations....417
References..419
Appendices
AppendixAMaximumPowerCoefficientforaYawedActuatorDisc..........429
AppendixBCalculationofAerodynamicLoadsInducedatthe
YawBearing...........432
AppendixCCalculationoftheInducedVelocityfromaVortexFilament
usingtheBiotSavartLaw............436
AppendixDLinearandSplineNumericalInterpolation...........441
AppendixEVortexFilamentStretching.................446
CurriculumVitae...........449
AuthorPublications...........450
Acknowledgements
ThisPhdthesiswouldhavenotbeenpossiblewithoutthesupportandhelpofmanypeople.
FirstofallIamthankfultoGodforgrantingmethehealthandenergy,especiallyinthemost
difficultmoments.Iwouldliketoexpressmyheartfeltthankstomysupervisorprof.dr.ir.
GijsvanKuikandcosupervisordr.GerardvanBusselfortheirguidanceandtheinvaluable
advices they gave me during my research. I am greatly indebted to prof. ing. Peter Paul
Farrugiaforgivingmetheopportunitytocarryoutthisworkandforrelievingmefrommy
dutiesattheUniversityofMaltatobeabletofinishmystudiesontime.Manythanksalsogo
totheUniversityofMaltaandDelftUniversityofTechnologyforthefinancingoftheproject.
I am very grateful to Wouter Haans for his cooperation in the experimental work on the
TUDelftwindtunnelturbineandforthemanydiscussionswehadthroughoutthecourseof
this work. Many thanks also go Simon Toet for his technical assistance in using the wind
tunnel equipment. I would like to extend my gratitude to my other colleagues at the wind
energy research group at Delft University of Technology for their hospitality, sharing their
knowledge and for offering fruitful suggestions: Ruud van Rooij, Nando Timmer, Carlos
Ferriera,ToniSubroto,WimBierbooms,MichielZaaijer,DickVeldkamp,EricvanderPoland
former colleagues Nord Jan Vermeer, Sander Mertens and Albert Bruining. Our secretary,
Sylvia Willems is thanked for her practical assistance, which saved me lots of time and
allowed me to concentrate on my work during my short stays in Delft.My colleague at the
UniversityofMalta,dr.ing.MartinMuscat,isacknowledgedforhissupportinfacilitatingmy
accesstotheMechanicalEngineeringComputerLabrequiredfortheextensivecomputations.
Dr. Scott Schreck and his colleagues at the National Renewable Energy Laboratory are
acknowledgedforprovidingthedataoftheNASAAmeswindtunnelexperiments.Thisdata
wasverystimulatingfortheintendedresearchwork.HermanSnelandGerardSchepersfrom
the Energy Research Centre of the Netherlands as well as Jeppe Johansen from the Ris
LaboratoryofDenmarkarealsothankedfortheirinterestandvaluablediscussions.
ThemembersoftheDoctoralExaminationCommitteearealsothankedfortheirattentionto
thisthesis.
Mostofall,IwouldliketoexpresstogratitudetomywifeMarisandfamilyfortheirsupport
andprayers.Withouttheirpatienceandencouragement,completionwouldhaveneverbeen
reached.
TonioSant
Delft,September2006
Abbreviations
AEPAnnualEnergyYield
ATCAnnualTotalCost
BEMBladeElementMomentumTheory
BETBladeElementTheory
COECostofEnergy
CFDComputationalFluidDynamics
DTUDenmarkTechnicalUniversity
ECNEnergyResearchCentreoftheNetherlands
EAWEEuropeanAcademyofWindEnergy
EWEAEuropeanWindEnergyAssociation
HAWTHorizontalaxiswindturbine
IEAInternationalEnergyAgency
NASANationalAeronauticsandSpaceAdministration,USA
NLRNationalAerospaceLaboratory,theNetherlands
NRELNationalRenewableEnergyLaboratory,USA
PIVParticleImagingVelocimetry
TUDelftDelftUniversityofTechnology,TheNetherlands
UAEUnsteadyAerodynamicsExperiment
WECSWindenergyconversionsystem
ii
Nomenclature
a indextorepresentvortexageoftrailingorshedvortexfilamentorparameter
usedinengineeringmodelforstalldelayorwakeskewangleorparameter
equaltoinunsteadyaerofoiltheoryforattachedflow
a1 diskaveragedorazimuthally(annular)averagedaxialinductionfactor
(=ua/U)
a1,m optimumdiskaveragedaxialinductionfactorforagivenyawangle
a1,c axialinductionfactoratbladeliftingline(=ua,c/U)
a2 azimuthally(annular)averagedtangentialinductionfactor(=ut/R)
a2,c tangentialinductionfactoratbladeliftingline(=ut,c/R)
a3 azimuthally(annular)averagedradialinductionfactor(=ur/U)
a3,c radialinductionfactoratbladeliftingline(=ur,c/U)
b indextorepresentbladenumberorparameterusedinengineeringmodelfor
stalldelayorparameterequaltohalfchordlengthinunsteadyaerofoil
theoryforattachedflow(m)
b1 constantinexponentialdecayfunctionapproximatingWagnersfunction
b2 constantinexponentialdecayfunctionapproximatingWagnersfunction
bs midspanofellipticalwing(m)
c localbladechord(m)
co maximumchordlengthofellipticalwing(m)
d rotordiameter(m)orparameterusedinengineeringmodelforstalldelayor
distancemeasuredfromsmokevisualisationphoto,correctedforparallax
effects(m)
da distancebetweenrotorhubcentreandyawbearingcentre(m)
dA tangential(chordwise)componentofaerodynamicloadingatabladeelement
(N/m)
dA normalcomponentofaerodynamicloadingatabladeelement(N/m)
dT axialthrustloadingatabladeelement(N/m)
dT2D axialthrustloadingatabladeelementcomputedfrom2Dliftanddrag
coefficients(N/m)
dT3D axialthrustloadingatabladeelementcomputedfrom3Dliftanddrag
coefficients(N/m)
dQ torqueloadingatabladeelement(Nm/m)
dQ2D torqueloadingatabladeelementcomputedfrom2Dliftanddrag
coefficients(Nm/m)
dQ3D torqueloadingatabladeelementcomputedfrom3Dliftanddrag
coefficients(Nm/m)
f Prandtltipandrootlossfactor
fd parameterusedtocorrect2Ddragcoefficientfor3Deffects
iii
fl parameterusedtocorrect2Dliftcoefficientfor3Deffects
fr Prandtlrootlossfactor
ft Prandtltiplossfactor
fL aerodynamicbladeliftloading(N/m)
fQ aerodynamicbladetorqueloading(Nm/m)
fT aerodynamicbladethrustloading(N/m)
h perpendiculardistanceofvortexfilamentfromagivenpoint(m)or
parameterusedforengineeringmodelforstalldelayorangularcalibration
constantforhotfilmprobeorverticaldistancebetweensmokevisualisation
planeandmeasuringgrid(m)
i bladestationnumberortrailingvortexnumberorindextodenoteradial
locationofhotfilmprobeinmeasuringplane
ip indextodenoteradiallocationofpointatwhichtheinducedvelocityis
computedbyprescribedwakeorfreewakevortexmodel
j indextodenoteangularpositionofhotfilmprobeinmeasuringplane
jp indextodenoteazimuthallocationofpointatwhichtheinducedvelocityis
computedbyprescribedwakeorfreewakevortexmodel
k angularcalibrationconstantforhotfilmprobeorindextonumberofvortex
filamentalongagivenhelixinprescribedwakevortexmodelorreduced
frequency
ka parametertocorrectaxialmomentumequationforyawedconditions
kc ratiooftheexitjetvelocitytothetruefreestreamvelocity
ke ratiooftheaxialinducedvelocityatthetunnelexitcomputedbythe
prescribedwakevortexcodetothetunnelexitjetvelocity
l lengthofvortexfilament(m)
m indextorepresentrotortimestep
n totalnumberofbladestationsandtrailingvorticesperbladeorconstantfor
speedcalibrationofhotfilm
nRev numberofrotorrevolutionstogeneratefreewake
nfwRev numberofhelicalrevolutionsinfarwakemodeloffreewakecode
nwRev numberofhelicalrevolutionsinwakeofprescribedwakevortexmodel
p tipvortexpitchmeasuredalongadirectionparalleltothefreewindspeed
(m)
pw tipvortexpitchtakenalongadirectionparalleltotherotoraxis(m)
pfw tipvortexpitchinfarwakemodeloffreewakecode(m)
r positionvectororradiallocationalongblade(m)
rc viscouscoreradiusofvortexfilament(m)
rceff viscouscoreradiusofvortexfilament,correctedforfilamentstraineffects(m)
rw radiallocationofvortexnodeonagivenhelixinprescribedwakevortex
model(m)
s reducedtime
t time(sec)
iv
ua diskaveragedorazimuthally(annular)averagedaxialinducedvelocity(m/s)
ua,c axialinducedvelocityatbladeliftingline.Canbedirectlyonliftinglineorat
agivenaxialdistancefromit(m/s)
ua,exit axialinducedvelocityattunnelexitjetascomputedbyprescribedwake
vortexmodel(m/s)
ut azimuthally(annular)averagedtangentialinducedvelocity(m/s)
ut,c tangentialinducedvelocityatbladeliftingline(m/s)
ur azimuthally(annular)averagedradialinducedvelocity(m/s)
ur,c radialinducedvelocityatbladeliftingline(m/s)
ux tangentialinducedvelocity(m/s)
uy axialinducedvelocity(m/s)
uz radialinducedvelocity(m/s)
uX inducedvelocityatnearwakenodealongtheXaxis(m/s)
uY inducedvelocityatnearwakenodealongtheYaxis(m/s)
uZ inducedvelocityatnearwakenodealongtheZaxis(m/s)
3Dinducedvelocityvectoratbladeelement(m/s)
c 3DinducedvelocityvectoratbladeelementinBEM,correctedforskewed
wakeeffects(m/s)
v localbladedeflection(m)
wa axialflowvelocityinrotorwake(alongtheyorYaaxis)(m/s)
wa,aver azimuthally(annular)averagedaxialflowvelocityinrotorwake(alongthey
axis)(m/s)
wa,c axialflowvelocityatbladeliftingline(alongtheyaxis).Maybedirectlyon
liftinglineoratagivenaxialdistancefromit(m/s)
wb velocityoffluidbypassingrotorwake(m/s)
wh horizontalflowvelocityinrotorwake(alongtheXaaxis)(m/s)
wr radialflowvelocityinrotorwake(alongthezaxis)(m/s)
wt tangentialflowvelocityinrotorwake(alongthexaxis)(m/s)
wv verticalflowvelocityinrotorwake(alongtheZaaxis)(m/s)
z parameterforviscousmodellingofvortexcore
A rotorcrosssectionalarea(m2)orconstantforspeedcalibrationofhotfilm
(V2/0C)
A0,A1,..A3 amplitudesinFourierseriesbasedengineeringmodelforskewedwake
effects
A1 constantinexponentialdecayfunctionapproximatingWagnersfunction
A2 constantinexponentialdecayfunctionapproximatingWagnersfunction
A tangential(chordwise)aerodynamicloadatbladeelementactingalongthe
axis(N)
A normalaerodynamicloadatbladeelementactingalongthe axis(N)
B totalnumberofbladesinrotororconstantforspeedcalibrationofhotfilm
BF blockagefactorforrotorinwindtunnel
Cd dragcoefficient
Cd,2D dragcoefficientfor2Dflow
Cd,2DMIN minimumdragcoefficientfor2Dflow
Cd,3D dragcoefficientcorrectedfor3Deffects(tip/rootlossand/orstalldelay)
Cd,s dragcoefficientatstall
Cd,max maximumdragcoefficient
Cdp pressuredragcoefficient
Cn normalcoefficient
Cm momentcoefficient
Cl liftcoefficient
Cl,2D liftcoefficientfor2Dflow
Cl,3D liftcoefficientcorrectedfor3Deffects(tip/rootlossand/orstalldelay)
Cl,lin liftcoefficientthatwouldbeobtainedifthe2Dliftslopeisextendedlinearly
beyondstall
Cl,s liftcoefficientatstall
Clc circulatoryliftcoefficient
Clnc noncirculatoryliftcoefficient
Ct tangentialcoefficient
CT rotoraxialthrustcoefficient
CQ rotortorquecoefficient
CP rotorpowercoefficient
CP,Max optimumrotorpowercoefficientforagivenyawangle
Dj diameterofopenjetwindtunneltube(m)
Dj diameteroftunneljetatagivendownstreamdistancefromrotorplane(m)
D diameterofrotorwakeatagivendownstreamdistancefromrotorplane(m)
E hotfilmvoltage(V)
F locationofrotorhubcentre
Fsa correctionfactorusedinBEMmodeltocorrectforskewedwakeeffectsin
yaw
FA1 righthandsideofBEMequationforaxialflow(bladeelementtheorypart)
(m2/s2)
FA2 lefthandsideofBEMequationforaxialflow(momentumpart)(m2/s2)
GXB geometricinfluencecoefficientforXcomponentofinducedvelocityfrom
boundvortexinwakeofprescribedwakevortexmodel(m1)
GYB geometricinfluencecoefficientforYcomponentofinducedvelocityfrom
boundvortexinwakeofprescribedwakevortexmodel(m1)
GZB geometricinfluencecoefficientforZcomponentofinducedvelocityfrom
boundvortexinwakeofprescribedwakevortexmodel(m1)
GXT geometricinfluencecoefficientforXcomponentofinducedvelocityfrom
trailingvortexinwakeofprescribedwakevortexmodel(m1)
GYT geometricinfluencecoefficientforYcomponentofinducedvelocityfrom
trailingvortexinwakeofprescribedwakevortexmodel(m1)
GZT geometricinfluencecoefficientforZcomponentofinducedvelocityfrom
vi
trailingvortexinwakeofprescribedwakevortexmodel(m1)
H towerheightorverticaldistancebetweensmokevisualisationcameraand
measuringgrid(m)
IBX geometricinfluencecoefficientforXcomponentofinducedvelocityfrom
boundvortexinnearwakeoffreewakevortexmodel(m1)
IBY geometricinfluencecoefficientforYcomponentofinducedvelocityfrom
boundvortexinnearwakeoffreewakevortexmodel(m1)
IBZ geometricinfluencecoefficientforZcomponentofinducedvelocityfrom
boundvortexinnearwakeoffreewakevortexmodel(m1)
ISX geometricinfluencecoefficientforXcomponentofinducedvelocityfrom
shedvortexinnearwakeoffreewakevortexmodel(m1)
ISY geometricinfluencecoefficientforYcomponentofinducedvelocityfrom
shedvortexinnearwakeoffreewakevortexmodel(m1)
ISZ geometricinfluencecoefficientforZcomponentofinducedvelocityfrom
shedvortexinnearwakeoffreewakevortexmodel(m1)
ITX geometricinfluencecoefficientforXcomponentofinducedvelocityfrom
trailingvortexinnearwakeoffreewakevortexmodel(m1)
ITY geometricinfluencecoefficientforYcomponentofinducedvelocityfrom
trailingvortexinnearwakeoffreewakevortexmodel(m1)
ITZ geometricinfluencecoefficientforZcomponentofinducedvelocityfrom
trailingvortexinnearwakeoffreewakevortexmodel(m1)
K parameterusedinBEMmodeltocorrectforskewedwakeeffectsusingthe
Glauertmodelorfunctionusedtosmoothenexperimentaldatausingthe
Gaussiankernel.
Kv correctionfactorappliedtoBiotSavartequationtocorrectforviscouscore
effects
L distancetakenfromsmokevisualisationphoto,uncorrectedforparallax
errors(m)
LFA localflowangle(deg)
LSSTQ lowspeedshafttorque(Nm)
Mtot totalnumberoftimemarchingstepsinfreewakecode
O locationofrotoryawbearingcentre
P rotorpowerdeveloped(W)
Q rotoroutputtorque(Nm)
QNORM dynamicpressure(N/m2)
R rotortipradius(sameasRt)(m)
Rr rotorhubradius(m)
Rt rotortipradius(sameasR)(m)
Rr,w innerwakeboundaryradiusforprescribedwakevortexmodel(m)
Rt,w outerwakeboundaryradiusforprescribedwakevortexmodel(m)
Rt,w1 outerwakeboundaryradiusforprescribedwakevortexmodelatthefirsttip
vortexcorelocation(m)
vii
Rt,w2 outerwakeboundaryradiusforprescribedwakevortexmodelatthesecond
tipvortexcorelocation(m)
Re Reynoldsnumberatbladesection
ReleaseRoot radiallocationatwhichinboardedgeofvortexsheetisshedfrombladein
prescribedwakevortexmodel(expressedasfractionofR)
ReleaseTip radiallocationatwhichoutboardedgeofvortexsheetisshedfrombladein
prescribedwakevortexmodel(expressedasfractionofR)
RCTF relativecomputationaltimefactorforfreewakesolution
REM bladerootedgebendingmoment(Nm)
RFM bladerootflapbendingmoment(Nm)
Sc viscouscoregrowthconstant(sec)
T towerbaselocationorrotoraxialthrust(N)
Ta measuredflowtemperature(0C)
Tf presetflowtemperature(0C)
U freewindspeedorwindtunnelspeed(m/s)
Ujet openjettunnelexitvelocity(m/s)
Up flowvelocitycomponentmeasuredalongxpaxisofhotfilm(m/s)
Ux freewindspeedcomponentparalleltorotordisk(m/s)
Uy freewindspeedcomponentnormaltorotordisk(m/s)
U resultantflowvelocityatyawedactuatordiscinaccordancewithsimple
momentumtheory(m/s)
V flowvelocityrelativetoaerofoil(m/s)
Veff measuredeffectiveflowvelocitybyhotfilm(m/s)
Veff,Xa measuredeffectiveflowvelocitywithhotfilmalignedalongtheXaaxis(m/s)
Veff,Ya measuredeffectiveflowvelocitywithhotfilmalignedalongtheYaaxis(m/s)
Veff,Za measuredeffectiveflowvelocitywithhotfilmalignedalongtheZaaxis(m/s)
Vp flowvelocitycomponentmeasuredalongypaxisofhotfilm(m/s)
Vn normalcomponentofflowvelocityrelativetobladesection(m/s)
Vr 2Dresultantflowvelocityrelativetobladesectionactinginplane(m/s)
Vrel 3Dresultantflowvelocityrelativetobladesection(m/s)
Vt tangentialcomponentofflowvelocityrelativetobladesection(m/s)
V tangentialcomponentofflowvelocityrelativetobladesection(sameasVt)
(m/s)
V normalcomponentofflowvelocityrelativetobladesection(sameasVn)(m/s)
V radialcomponentofflowvelocityrelativetobladesection(m/s)
VA, absolutevelocitycomponentofbladeelementalongtheaxis(m/s)
VA, absolutevelocitycomponentofbladeelementalongtheaxis(m/s)
VA, absolutevelocitycomponentofbladeelementalongtheaxis(m/s)
Wp flowvelocitycomponentmeasuredalongzpaxisofhotfilm(m/s)
WX velocityofnearwakenodealongtheXaxisinfreewakevortexcode(m/s)
WY velocityofnearwakenodealongtheYaxisinfreewakevortexcode(m/s)
WZ velocityofnearwakenodealongtheZaxisinfreewakevortexcode(m/s)
viii
Yap axialdistanceofplaneparalleltorotorplaneatwhichinducedvelocity
distributioniscomputedusingprescribedwakeorfreewakevortexmodel
(m)
GreekNomenclature
angleofattack(deg)orviscouscoregrowthconstant
zeroliftangleofattack(deg)
e equivalentangleofattack,accountingforunsteadyeffects(deg)
sweep sweepangleofattack(deg)
s stallingangleofattack(deg)
rateofchangeofangleofattackwithtime(deg/s)
bladeconingangle(deg)
rotoraxistiltangle(deg)
s wakeskewangle(deg)
cutoffdistance(m)
v viscouscorediffusivitycoefficient
vortexfilamentstrain
a relativeerrorinua,cbetweenthatpredictedbyprescribedwakevortexcode
andthatobtainedfromhotfilmmeasurementsbyassumingthatthefree
streamvelocityisequaltothetunnelexitjetvelocity(%)
kc errorinthecalculatedaxialinducedvelocityduetodiscrepancybetween
tunnelexitvelocityandtruefreewindspeed(%)
Q errorinthederivedbladetorqueloadingduetoerrorsintheinflow
measurements(%)
T errorinthederivedbladeaxialthrustloadingduetoerrorsintheinflow
measurements(%)
wa,c errorintheflowvelocityattheliftinglineobtainedfromtheinflow
measurements(%)
rotororbladeazimuthangle(deg)orindicialresponsefunctionderivedby
Wagnerforunsteadyaerofoils
p angularpositionofhotfilmprobe(deg)
w angularpositionofwakevortexfilamentnodeinprescribedwakevortex
model(deg)
parameterusedinthecosinesegmentationofradialsegmentsofliftinglines
(deg)
rotoroperatingtipspeedratio
bladeaspectratio
a airdynamicviscosity(Ns/m2)
localinflowangle(deg)
0,1,..3 phaseanglesinFourierseriesbasedengineeringmodelforskewedwake
ix
effects(deg)
bandwidthusedinsmoothingmethodusingtheGaussiankernel
localbladepitchangle(deg)
tip pitchangleatbladetip(deg)
densityofair(kg/m3)
indextodenotetimestep
tot totalnumberofequallyspacedtimestepsinonewholerotorrevolution
p indextodenoteazimuthallocationofpointatwhichtheinducedvelocityis
computedbyprescribedwakeorfreewakevortexmodel
kinematicviscosityofair(m2/s)
n relativeerrorwhenvaryingn(%)
fw relativeerrorduetofarwake(%)
wp relativeerrorforwakeperiodicity(%)
relativeerrorwhenvarying(%)
vorticity(s1)
phaseshiftangleusedintheBoeingVertolmodelfordynamicstall(deg)
azimuthalstepforonerotorrevolution(deg)
incrementaltimestep(sec)
rotorangularspeed(rad/s)
collectivepitchangleofblade(deg)
yawangle(deg)
circulation(m2/s)
B bound circulation(m2/s)
B,MAX maximumboundcirculationalongellipticalwing(m2/s)
B,2D bound circulationbasedon2Dliftcoefficient(m2/s)
B,3D bound circulationbasedon3Dliftcoefficient(m2/s)
T trailingcirculation(m2/s)
S shed circulation(m2/s)
ListofDevelopedSoftwareCodes
HAWT_BEMBladeelementmomentummodel
HAWT_LFIMModeltoderiveaerodynamicloadsfromwakeinflowmeasurements
HAWT_PVCPrescribedwakevortexmodel
HAWT_FWCFreewakevortexmodel
xi
Chapter1Introduction
1.Introduction
Energy is fundamental to economic and social development. On the dawn of the 21st
century we are being faced with one of the toughest challenges ever that of securing
energy supply. We are still heavily dependent on oil resources which will eventually
becomedepletedwithinafewdecades.Energyconsumptionisalsoexpectedtoincreaseup
to about 40% by the year 2010. An increasing world population, an enlarged global
economyandanimprovedstandardoflivingallcontributetogreaterdemandsforenergy.
At the same time, we are facing the greatest threat to oursurvival on planet earth: global
climate change. Climate change is not just an environmental threat but also an economic
threat. Rising sea levels, more severe droughts and health issues will increase insurance
costsradicallyinthefuture.
Wind energy is one of the most effective power technologies that is ready today to be
deployed globally on a scale that can aid in tackling this problem. Wind energy is a
significant and powerful resource and is safe, clean and abundant. It is being very
successfulinpenetratingtheenergymarket,especiallyinEuropewhichhasashareof70%
oftheglobalwindenergyindustry.Backedbyeffectivepolicies,thewindenergyindustry
isexperiencingaremarkablegrowthof2025%perannum(EWEA,[23]).Thisisconsidered
tobeveryhighforanindustrymanufacturingheavyequipment.Inthepast20years,the
sizeofwindturbinesincreasedrapidlyfromabout15mdiameterhavingacapacityof50kW
to about 120m having a capacity of 5MW. The initial capital cost of the turbines also
decreasedfrom3500Euros/KWtoabout1000Euros/KW.Anumberofassessmentsconfirm
that the worlds wind resources are enormous and well distributed. The total available
resourcethatistechnicallyrecoverableisestimatedtobe53,000Terawatthours(TWh)/year,
which is over twice as large as the projection for the worlds entire electricity demand in
2020 (EWEA, Greenpeace, [24]). Lack of resource is therefore unlikely to be a barrier to a
penetrationofwindenergyintheenergymarket.Amajorbarrieriscostsincewindenergy
hastofacefiercecompetitionfromconventionalsourcesofenergybasedonfossilfuelsand
nuclear energy. Despite the great progress made, wind energy still has a long way to go
before it reaches its full potential in terms of the largescale supply of electricity. While it
can already be cost competitive with newly built conventional plants at sites with good
wind speeds, significant further cost reductions are necessary. In the IEA report Long
TermResearchandDevelopmentNeedsforWindEnergyfortheTimeFrame2000to2020
[94], it has been estimated that if wind energy is going to supply 10% of the worlds
electricityneedsby2020,costreductionsinthetechnologyof30to50%arestillnecessary.
Research and development work could contribute up to 40% of those reductions. The
1
Chapter1Introduction
challengefacedbythewindenergycommunityistoproducemorecostcompetitivewind
turbinesthroughhighlyoptimizeddesigns.
1.1TheRoleofAerodynamicsinWindTurbineDesign
Awindturbineisacomplexsystemworkinginacomplexenvironment.Itiscomposed
of subsystemsworking together in a tightly coupled manner. Furthermore, wind turbines
must be adapted to specific meteorological and topographical characteristics of each
particular site. This makes the design of a HorizontalAxis Wind Turbine (HAWT) a
complex process that is characterized by several tradeoff decisions aimed at finding the
optimumoverallperformanceandeconomy.Thedesignapproachismultidisciplinaryand
integrates several branches of engineering including aerodynamics, mechanical, materials,
electrical, control and manufacturing engineering. The design objectives are to maximize
energy yield, reduce overall costs and maximize the lifetime of the system (see Fig. 1.1).
Theseobjectiveswilldeterminetheminimumcostofenergy(COE).
Electricalpower
Mechanical generation
Control
transmission
Siting
Aerodynamics
DesignObjectives Operation&
Structure 1.MaximizeEnergyYield maintenance
2.ReduceCosts
3.MaximizeLifetime
Gridconnection
Materials
Fatigue Windfarmlayout
Manufacturing Foundation
Figure1.1DesignconsiderationsforaWindEnergyConversionSystem.
A typical design process starts off with the identification of subsystems and components
makingupthewholewindenergyconversionsystem(WECS)(seeFig.1.2).Basicallythree
types of models are integrated in the design process: (1) an aerodynamics model that
estimates the aerodynamic performance, loads and annual energy yield (AEP) for a given
rotorgeometryandoperatingsite;(2)astructuresmodelthatwillcalculatethetotalloads
andinducedstressesontheloadbearingcomponentsresultingfromtheaerodynamicloads
(computedbytheaerodynamicsmodel)andduetogravityanddynamics.Thismodelwill
2
Chapter1Introduction
alsoestimatethefatiguelifetimeofeachcomponent;and(3)acostmodelthatcomputesthe
expenses required to manufacture the WECS components, together with other costs
requiredtoinstallandoperatethesystemattheinstallationsite.Thecostmodelcalculates
the equivalent annual total cost (ATC) taking into account all costs incurred over the
expected lifetime of the system. Finally, the annual cost of energy (COE) is determined
(equal to the ratio AEP/ATC) (see Fig. 1.2). Throughout the design optimization, the
differentvariablesoftheaerodynamic,structureandcostmodelsarealteredsystematically
toyieldtheminimumCOE.Foroffshorewindturbines,thedesignprocessbecomesmore
complicated since it should cater for more costly foundations and for a tougher
environment. For description of integrated design approaches for offshore wind turbines
refertoworkofKuhn[43]),Hendriksetal.[38]andvanderTempel[92].
As illustrated in Fig. 1.2, aerodynamics plays a vital role in the design process as it will
determine the AEP and the aerodynamic loads which in turn influence the costs of the
differentWECScomponents.TheaerodynamicdesignofaHAWTrotorhastheobjectiveof
providingtheoptimizedgeometry(diameter,numberofblades,bladeshape)thatwillyield
the lowest COE possible. It is subject to three constraints that may be conflicting: (1)
Maximizationofpowercoefficient;(2)Maximizationofenergyyield;and(3)Reductionof
bladeloads(seeFig.1.3).
Mechanical& Windspeed
electricallosses distribution(Weibull)
AnnualEnergy
PowerCoefficient Power Powerof
producedatsite
ofrotor ofrotor windturbine
(AEP)
Rotoraerodynamicperformance
Aerodynamicsmodel
Objective:
Aerodynamicloads CostofEnergy
COE=AEP/ATC
Gearbox Yawdrive
Blade
Lowspeedshaft Nacellestructure
Hub
Loads&materialstress
Brakes Controls
Bladeflange
Generator Otherequipment
Costmodel AnnualTotalCost
(ATC)
Maintenance
Foundation
Remotemonitoring Transport
Gridconnection Installation
OtherCosts
Figure1.2Typicalschemeofmodelsfordesignoptimization.
3
Chapter1Introduction
Maximizingthepowercoefficientwillincreasetheenergyyield.Howevertomaximizethe
energyyield,thepowercoefficientshouldbemaximizedoverawiderangeofwindspeeds.
Intheearlierdays,aerodynamicdesignfocusedonlyonmaximizingthepowercoefficient
CP.ButitwasdiscoveredthatthemaximumCPwasonlyachievedatasmallrangeofwind
speeds, especially in stallregulated turbines with the detriment of reducing the annual
energyyield(Snel,[81]).Increasingthechordandtwistofthebladeswillhelpinincreasing
the energy yield at low windspeeds. This is helpful for sites where the mean annual
windspeed tends to be on the low side. Yet doing so may increase the blade weight and
thusthegravitationalloadsofthesupportingstructure.
MaximizePower MaximizeEnergy
Coefficient Yield
ReduceLoads
Figure1.3Constraintsforaerodynamicdesign.
Windturbineoperationislimitedbyacutoutwindspeedbeyondwhichtherotorhastobe
broughttoastandstillduetohighwindspeeds.Ahighercutoutwindspeedwillcontribute
toalargerenergyyieldsincebenefitwillbetakenfromthehighwindspeeds(whichhave
largerenergyintensities).Ontheotherhand,thiswillpushtogreaterstructuraldemands
resultingfromhigherloads.
Itshouldbeclearthatalthoughprofoundknowledgeofaerodynamicsisanindispensable
requirement, it alone is not sufficient to determine the loads and stresses on the WECS
components. These depend on the deformation (which may be unsteady) of the system
componentsinresponsetotheexternalloads.Theaerodynamicloadsonthebladesactas
externalloads.Theinteractionbetweentheseaerodynamicloadsandthedynamicbehavior
ofthesystemcomponentsisknownasaeroelasticity.Aprequisteistoensurethatthewind
turbineisaeroelasticallystableduringtheoperations.Inacomplexoperatingenvironment,
the unsteady aerodynamic loads may cause the WECS to become aeroelastically unstable
whichcausesthelargevibrationsthatreducethefatiguelifetimeofthesystem.Aeroelastic
analysis demands that aerodynamic models inherently form part of structural dynamics
models.Thisaddstothecomputationalpowerrequiredbydesignsoftwaretools.
4
Chapter1Introduction
1.2PrinciplesofHAWTAerodynamics
Insimplewords,aHorizontalAxisWindTurbine(HAWT)isapropellertyperotorthat
extracts energy from the wind. Since the turbine extracts kinetic energy from the fluid
stream, the air flowing through the rotor experiences a decrease in the velocity. The
momentumchangeoftheairwillexertanaxialthrustontherotor.Therotoralsoimpartsa
swirlvelocitycomponenttotheairinadirectionoppositetothatoftheshaft.
Eachrotorblademaybeconsideredasarotatingwing.Theairflowingaroundtheblades
causesthelattertoexperiencelift(resultingfromboundcirculationaroundtheblades)and
dragforces.Thecombinedactionoftheseforcesyieldsanoutputtorqueattherotorshaft,
therebyproducingpower.Duetothefactthatthebladesrotateinsteadofmovinglinearly
asinanormalwing,thelocalaerofoilliftanddragcoefficientsmaybedifferentfromthose
obtainedin2Dwindtunneldata.Thisisespeciallynotedathighanglesofattackwhenthe
phenomenonofstalldelayisknowntotakeplace,resultinginsignificantlylargeraerofoil
coefficients.StalldelaywillbedescribedinChapter3.
Thewakefromtherotatingbladecomprisesavorticalshearlayerorvortexsheet,asshown
in Fig. 1.4, which is a schematic reconstruction of the wake formed by a rotating blade as
observedinflowvisualizationandfieldmeasurements.
Boundcirculation r/R
B Bladetip
Rotorhub
Trailingvortices
Bladeroot
Shedvortices
Rotoraxis
Vortexcore
Root
wind Tip
Vortexsheet
Figure1.4Schematicdiagramshowingwakedevelopedbyarotatingbladeofawindturbine.
5
Chapter1Introduction
The circulation distribution in the vortex sheet originates from the bound circulation (B)
developed at the blades. The former circulation is composed of two vector components:
trailing circulation () that is released from the blades in a direction perpendicular to the
blades trailing edge and is related to the spanwise variation of the bound circulation
( B r );shedcirculation(S)thatisreleasedfromthebladesinadirectionparalleltothe
blades trailing edge and is related to variation of bound circulation with time ( B t ).
Thegeometryofthevortexsheetemergingfromthebladeswillchangesuchthattheedges
willrollup(similartothatobservedonawinginlinearflight)toformatipandrootvortex
asshowninFig.1.4.
AHAWTrotorisnormallyorientedwiththewindsuchthattheaxisofrotationisparallel
to the incoming wind velocity vector (often referred to as the axial condition). The wake
formedbehindaHAWTconsistsofvortexsheets,oneperblade(asdescribedinFig.1.4),
thattraceahelicalpathasaresultofrotorrotation.Thevortexsheetswillrolluptoforma
tip and root vortex, as illustrated in Fig. 1.5. The root vortex is usually distorted by the
presence of the turbine nacelle and consequently it is very difficult to observe it in
experiments. The wake (slipstream) boundary which is usually defined by the radial
locationofthetipvortices,expandsdownstreamasaconsequenceoftheretardationofthe
flow.Thewakevorticityisresponsibleforslowingdownoftheairasitflowsthroughthe
rotor. It will also alter the local angle of attack at the blades, thereby influencing the
aerodynamicforces.Whenthewindspeedandrotorspeedareconstantwithtime,thereis
noshedcirculationinthewake.
Trailingvortices Tip
Root
U
Root
Tip
Figure 1.5 Schematic diagram showing helical wake developed by a wind turbine in axial
conditions.
6
1.3CurrentStatusofAerodynamicDesignModelsforHorizontalAxisWindTurbines
Ayawedrotorisonewhichisoperatingwithitsaxisofrotationnotparalleltotheincoming
wind velocity vector. In the real operating environment of a HAWT, the direction of the
windchangesfrequentlywithrespecttotherotoraxis.Asaresult,theturbinemayoperate
inyawforconsiderableamountsoftime.Thisyawedconditionintroducesacyclicangleof
attack at the blades and causes the helical wake to become skewed as shown in Fig. 1.6,
yieldinganunsteadyandcomplexinductiondistributionattherotorplane.Infactthewake
isquitesimilartothatofahelicopterrotorinforwardflightwiththemaindifferencebeing
thatitexpandsinsteadofitcontracts.Thetimedependentaerodynamicloadsattheblades
willcauseshedcirculationinthewake.Whentheangleofattackatabladesectionexceeds
the aerofoilsstalling angle, dynamic stall takes place causing the maximum aerodynamic
loads to be much higher than those predicted by 2D static aerofoil data. The hysterisis
effectsintroducedbydynamicstallmayhaveanegativeeffectontheaeroelasticdamping
behaviour of wind turbine blades. They reduce the fatigue lifetime leaving an adverse
impactontheeconomicsofthesystem.Thephenomenonofdynamicstallwillbedescribed
inChapter3.
The flow field across a wind turbine may be separated intuitively into two regions: the
global flow field which extends far upstream of the turbine to far downstream and a local
(rotor/blade)flowfieldwhichistheflowaroundtheindividualblades.Physicallythesetwo
partsareinherentlytiedtogether,butforboththeunderstandingofthephysicsandalsofor
Trailingvortices
Shedvortices Tip
U
Root
Root
Tip
Figure 1.6 Schematic diagram showing helical wake developed by a wind turbine in yawed
conditions.
7
1.3CurrentStatusofAerodynamicDesignModelsforHorizontalAxisWindTurbines
modelling,itmaybeconvenienttotreatthemseparately.Theinteractionbetweenthetwo
regionsisstrong:theflowintheglobalregiondeterminestheinflowconditionattherotor
blade and the forces on the blades (which can be seen as a localized pressure change)
influencestheflowintheglobalregion.
1.3CurrentStatusofAerodynamicDesignModelsforHorizontal
AxisWindTurbines
Since aerodynamic modelling should ultimately serve as a design tool, fast and robust
codes are required. Throughout the design process, different wind turbines should be
modelledoverwiderangeofoperatingconditions(includingyaw,extremegusts,start/stop
sequencesandstandstillconditions).
Various mathematical models exist to model the aerodynamic loads on rotors: Blade
ElementMomentum methods, Prescribed or Freewake Vortex methods, Acceleration
PotentialmethodsandCFDtechniques.Anoverviewofthesemethodsmaybefoundinthe
followingreferences:(Snel,[80];Leishman,[50];Conlisk,[18];vanBussel[15]andVermeer
etal.,[99]).Thesewereinitiallydevelopedtotreatpropellerandhelicopteraerodynamics,
butcouldthenbeeasilyadaptedtomodelHAWTs.Duetoitsrelativelyhighcomputational
efficiency,manyaeroelasticdesigncodesstillrelyontheBladeElementMomentum(BEM)
theoryforpredictingtheaerodynamicloads.However,thistheoryissimpleandlacksthe
physicstomodelthecomplexflowfieldsaroundarotorandconsequentlyitsaccuracymay
be unsatisfactory. Its limitations are mostly observed when treating stalled flows and
unsteadyconditionssuchasinrotor yaw.Othermethods aremuchmorecomprehensive,
especially freewake vortex methods and CFD. Yet unfortunately, with present computer
capacity, they are still too computationally expensive to be fully integrated into wind
turbinedesigncodes.Thecomplexityofwindturbinedesignisprohibitingtheuseofthese
more elaborate methods that are systematically used today in other aerodynamic
applications.
Inthepastyears,severalcorrectionswereaddedtoBEMcodestoimprovetheiraccuracy.
These mainly took the form of engineering models that mainly fall under two
classifications: (1) aerofoil data models that correct 2D static aerofoil for 3D effects (blade
tip/rootloss,stalldelayanddynamicstall)and(2)inflowmodelsthatcorrectfortheuneven
inducedvelocitydistributionattherotorplaneduetoskewedwakeeffectsinyawaswellas
forconditionsofheavyand/orunsteadyloadingontherotor.Abriefoverviewofsomeof
thesemodelswillbepresentedinChapter3.Theengineeringmodelsweredevelopedusing
experimentaldataorusingthemoreadvancedmodels,includingCFD.
It is often thought that in the future, the increased computer power that will become
availabletothewindturbinedesignerwillmakeitpossibletointegratethemoreadvanced
8
Chapter1Introduction
modelsinaeroelasticdesigncodesandthusreplacingBEMbasedmodelscompletely.This
would make research in the field of BEM improvements futile. However, considering the
present situation, there are still various reasons why effort should still be devoted to
improvingBEMcodes:
First of all, one should keep in mind the fact that the increased computer power that
will be available in the future for more sophisticated aerodynamic modelling will be
partly limited by the structural dynamicists request to employ more accurate (thus
more computational demanding) structural analysis codes. A typical example would
betheinclusionofcomplexwaveandfoundationdesignmodelsforoffshoresystems.
Secondly, the BEM method is considerably accurate when treating attached flow
conditions (low angles of attack) in axial flow. In the recent years, it was shown that
this method is also sufficiently accurate for stalled conditions provided that reliable
aerofoildataisused.Accordingly,moreeffortshouldbemadetoobtainmorereliable
aerofoil data from wind tunnel rotor experiments and CFD. Also, it is still unclear to
what extent is BEM accurate in yaw when reliable aerofoil data is used. Further
researchisrequiredtoinvestigatethis.
Thirdly, even though the more elaborate methods are comprehensive, they do not
necessarily always yield better results than BEM. This has been realized in a recent
Europeanproject(Schepersetal.,[70]).
9
Chapter1Introduction
10
Chapter2AimofThesisandApproach
2.AimofThesisandApproach
2.1ProblemStatement
Inthepastyears,BEMcodeswereextensivelytestedagainstexperimentalmeasurements.
Inmanysituations,thereliabilityofsuchcodeswasfoundtobeunacceptable,inparticular
whentheangleofattackatthebladeswaslargeandinyawedconditions.Forinstancethis
wasobservedafewyearsago,inyear2000,inablindcomparisonstudyorganizedbythe
NREL.Inthisstudy,atwobladedwindturbinewasextensivelytestedintheNASAAmes
windtunnelforawiderangeofoperatingconditions(Schreck,[73]).Whencomparingthe
predicted results by different aerodynamics/aeroelastic codes from various
universities/institutions with the measured data considerable inconsistencies were found
(Simms et al., [78]). In some cases, deviations of the BEM predictions from the
measurements exceeded 200%, even though the simplest operating conditions of a wind
turbine were being considered (i.e. uniform windspeed and constant rotor speed, blade
pitchandyawangle).Thishasshownthattheaerodynamicinteractionbetweentherotor
blades and the wake is nonlinear and more threedimensional in nature than for fixed
wings in linear flight. Certain aerodynamic phenomena associated with wind turbine
bladesarestillpoorlyknownandarethereforechallengingtopredictaccurately,asinthe
caseofstalledbladesandtheunsteadyeffectsexperiencedinyawedconditions.Foragiven
aerofoilgeometry,thesteadyandunsteadyaerofoildataofawindturbineblademaydiffer
considerably from that normally obtained in 2D static wind tunnel experiments. Todate,
fundamental limits exist in the validity of models used for wind turbine design and
certification. As explained by Leishman [50], a major challenge for researchers is to better
understand the aerodynamic issues associated with wind turbines to develop more
rigorous models suitable for a wider range of applications and to better integrate and
validate these models with reference to good quality experimental measurements. These
modelsshouldalsobecomputationallyefficientiftheyaretobeusedindesigncodes.
Wind tunnel tests on model turbines are indispensable to have a better understanding of
theunderlyingphysicsandtoimproveengineeringmodelsfordesigncodes.Thecontrolled
environmentofferedbyawindtunnelprovidesasetofmeasurementsthatisfreefromthe
uncertainties caused by the different atmospheric effects that are always present in open
fieldtestsofturbines.ToimprovethepredictionsofBEMbaseddesigncodes,morereliable
aerofoil data models and inflow correction models are required. However, using the
experimental data to improve these models is not an easy task. Two major problems are
encountered:
11
Chapter2AimofThesisandApproach
ProblemI:Windtunnelexperimentaldataisusuallyrareandlimited.Thisisbecauseturbine
testing is very expensive. Also certain parameters may be very difficult to measure
accurately. To be able carry out a detailed experimental investigation of a turbines
aerodynamic behaviour, the following set of measurement data would be ideally
required:
(3) Measurements of the wake geometry to establish the expansion of the wake,
location of the tip vortices and the wake skew angle in the case of yawed
conditions. This is usually accomplished using smoke visualization techniques
(Vermeeretal.,[99]).
Despite the fact that over the past years various databases of wind tunnel data have
been produced, a complete set of datacomprising the above three measurement data
sets for a wind turbine operating over a wide range of operating states in both axial
andyawedconditionsisstillpresentlyunavailableinthewindenergycommunity.
Problem II: There is a difficulty in determining accurately the angle of attack. To be able to
derivethelocalaerofoilliftanddragcoefficientsClandCdfromthemeasuredCnandCt
obtained from blade pressure measurements knowledge of the angle of attack is
requiredasillustratedinFigure2.1.
Theangleofattackmaybeestimateddirectlyfromdetailedinflowmeasurementsbutthese
are not always available. Alternatively flow direction probes may be installed at different
radial locations of one blade, just in front of the leading edge to measure the local inflow
angle(LFA)asshowninFig.2.2.Though,duetotheinfluenceoftheboundcirculationat
thebladesandthewakevorticity,theinflowanglemaydiffersignificantlyfromtheangle
ofattack.Acorrectionhastobethenappliedtoestimatetheangleofattackfromtheinflow
angle,asdiscussedbyRooijet.al.[66].Becauseoftheflowfieldacrosstherotoriscomplex,
thecorrectionthatisusuallyobtainedfromsimple2Dwindtunnelcalibrationproceduresis
unreliable.Also,theprobesmaydistorttheflowoverthebladesandthismaycauseerrors
inbladesurfacemeasurements.Whendealingwithyawedconditions,theinfluencesofthe
unsteady shed vorticity and the effects resulting from the skewed wake will make the
requiredcorrectionverydifficulttoestablish.
12
Chapter2AimofThesisandApproach
Cthrust
Cn
Cd
Cl
line
chord
Ctorque
planeofrotation
Ct
Vn
Vr Cl = Cn cos + Ct sin
Cd = Cn sin Ct cos
Vt
Figure2.1Bladesectionaerodynamicloadcoefficientsandrelativevelocityflowcomponents.
line
chord
flowdirectionprobe planeofrotation
LFA
Vr
Figure2.2Measuringthelocalflowangleusingaflowdirectionprobe.
Another method to determine the angle of attack is the socalled inverse BEM method
which makes use of the BladeElementMomentum equations to estimate the axial and
rotation induction factors from the known blade loading, thereby finding the angle of
attack.ThismethodhasbeenappliedbySneletal.[82],Bruiningetal.[13]andlateronby
Lainoetal.[44].TheaccuracyofthismethodislimitedbythecapabilityoftheBEMtheory
in predicting accurately the induction factors at the rotorplane. This method would not
alwaysbereliable,especiallyinhighloadingandyawedconditions.
Research showed that when the new aerofoil data, derived from blade pressure
measurementsinconjunctionwithanyoftheabovemethodsforfindingtheangleofattack,
was used in BEM codes, correlation with the experimental load measurements generally
improved.Yettheproblemofaccuratelyderivingtheangleofattackforthemeasurements
remained a major source of uncertainty. Different researchers used different methods for
estimating the angle of attack and consequently discrepancies resulted in the derived lift
anddragdata,eventhoughthesamebladepressuremeasurementswerebeingused.Asa
13
Chapter2AimofThesisandApproach
result, it may be argued that the uncertainty in deriving the angle of attack is a major
stumblingblocktocarryoutaclearquantitativeassessmentofthetrustworthinessofBEM
basedcodes.
2.2AimofThesis
Inthisthesis,adetailedinvestigationoftheaerodynamicsofwindturbinesinbothaxial
andyawedconditionswascarriedoutbasedonwindtunnelmeasurementswiththeaimof
providingabetterunderstandingofthelimitationsoftheBEMtheory.
Asalreadymentionedinsection2.1above,tobeabletoaccomplishadetailedexperimental
investigation of wind turbine aerodynamics, it is very helpful that the experimental data
consists of the following data sets: blade pressure measurements (to derive the
aerodynamicloading),inflowmeasurementsinthenearwakeandtherotorplaneaswellas
measurements concerning the wake geometry. It goes without saying that, when any of
these three data sets is unavailable, it will impose restrictions to which detail the
aerodynamic analysis can be performed. This study focused on developing new
methodologies that make use of limited experimental data in conjunction with advanced
aerodynamicmodelstoderivetheadditionallyrequiredaerodynamicperformancedatafor
bothaxialandyawedrotors.Inthisway,adeeperaerodynamicstudycouldbeperformed.
Themethodologiesweredevelopedfortwoparticularcases:
(i) Case A: The wind turbine experimental data only consists of detailed inflow
measurementsinthenearwakeandwakegeometrydata
(ii) Case B: The wind turbine experimental data only consists of blade pressure
measurements.
Eachmethodologyisdescribedindetailanditslimitationsexamined.
The experimental data and the new aerodynamic data derived using the respective
methodology were used to carry out a thorough assessment of a BEM code. The major
scopeofthisassessmentwastoprovideguidelinesthatwouldbeusefulindevelopingnew
engineeringmodels.
14
Chapter2AimofThesisandApproach
2.3Approach
In the research work, the experimental data of two different wind turbines were
considered: (1) The Delft University of Technology (TUDelft) wind tunnel model turbine
and(2)theNRELPhaseVIwindturbine.
2.3.1ResearchworkontheTUDelftwindtunnelturbine
ThefirstpartoftheprojectdealtwiththeTUDelftmodelturbine.Aseriesofexperiments
were conducted on this rotor for both axial and yawed conditions in the openjet wind
tunnel facility of Delft University of Technology with the close collaboration of another
Ph.DresearcherWouterHaans.Theexperimentsconsistedofthefollowing:
a. Detailed hotfilm measurements in the near wake along planes parallel to the
rotorplane(bothupstreamanddownstreamoftherotorplane)
b. Smokevisualizationexperimentstotracethetipvortexpathsoftheturbinewakeand
thusobtaindetailedregardingthewakegeometry.
Unfortunately, the apparatus was incapable of measuring the pressure distributions over
theblades.ThesituationwasthereforeidenticaltoCaseAdescribedinsection2.2above.A
methodologywasdevelopedtoderivethetimedependentaerodynamicloaddistributions
at the rotor blades from the hotfilm measurements in conjunction with an advanced
unsteadyaerofoilmodel.Theapplicationofthismethodologywaslimitedtoattachedflow
conditions(lowanglesofattack)onlyforwhichunsteadyaerofoilmodelsareknowntobe
reliable. Very briefly, the sequence of steps in applying this method are as follows: the
angleofattackandflowrelativevelocitiesatthebladesarefirstestimateddirectlyfromthe
hotfilminflowmeasurements.Theadvancedunsteadyaerofoilmodelisusedtoderivethe
liftcoefficientdistributionsattheblades.Thedragcoefficientsareestimatedfrom2Dwind
tunnel static aerofoil data. This is acceptable since small angles of attack were being
considered (attached flow conditions). Finally the aerodynamic loads at the blades are
computed using the bladeelement theory equations. The results from this method were
thencomparedwiththosepredictedbyaBEMcode.
2.3.2ResearchworkontheNRELPhaseVIwindtunnelturbine
ThesecondpartoftheprojectdealtwiththeNRELPhaseVIwindturbine.Thisturbinewas
tested in the NASA Ames 80ft X 120ft wind tunnel way back in the year 2000. The data
collected from these experiments [73], usually referred to as the NASA Ames Unsteady
Aerodynamics Experiments (UAE), is very extensive and is currently being analyzed by
severalinstitutionsthroughtheIEAAnnexXX.Itisbeingusedasabenchmarkbythewind
turbine aerodynamics community in assessing the validity of improved aerodynamics
codes based on BEM, CFD or Vortex Methods [19, 44, 45, 72, 87, 90, 95]. In these
15
Chapter2AimofThesisandApproach
experiments, timeaccurate blade pressure measurements were taken with the rotor
operatinginbothaxialandyawedconditionstogetherwithstraingaugemeasurementsfor
theoutputtorqueandtherootflap/leadmoments.However,detailedinflowmeasurements
at the rotorplane were not carried out. The situation is therefore identical to Case B
described in section 2.2 above. In this project a novel and comprehensive methodology is
beingproposedforusingthebladepressuremeasurementsinconjunctionwithafreewake
vortex model to estimate the angle of attack distributions at the blades more accurately,
togetherwiththeinflowdistributionsattherotorplaneandwakegeometry.Thenew3Dlift
anddragdatatogetherwiththederivedinflowdistributionsattherotorplanearethenused
to assess the improvement in BEM load predictions in axial and yawed conditions.
EmphasishereismadeindeterminingtheaccuracytowhichtheBEMtheoryiscapableto
modelaerodynamicloadsinhighlystalledandyawedconditionsifreliable3Daerofoildata
areused.
The proposed methodology for coupling the blade pressure measurements with a free
wake vortex model is based on the principle that, in a wind turbine wake, it may be
assumedthatvorticityisconserved.Thusthecirculationinthewakecorrespondstothat
around the blades. From the blade pressure measurements, it is possible to estimate the
bound circulation at the blades which may then be used to generate the freewake. The
sequenceofstepsinapplyingthismethodareasfollows:Initially,aspanwisedistribution
fortheangleforattackisassumedattheblades.Thisisthenusedtogetherwiththevalues
ofCnandCtobtainedfromthebladepressuremeasurementstoestimatetheliftcoefficients
at the blades. Using the KuttaJoukowski law, the bound circulation distribution at the
bladesisthendeterminedandprescribedtothefreewakevortexmodeltogeneratethefree
vorticalwake.Theinducedvelocityatthebladesisestimatedandusedtocalculateanew
angleofattack.Theprocessisrepeateduntilconvergenceintheangleofattackisachieved.
Originally,thismethodwasappliedbyTangleretal.[90,91]butusingaprescribedvortex
model and treating axial conditions only. However a freewake vortex model is a more
realisticrepresentationbecausethewakegeometryisallowedtodevelopfreelydepending
onthecirculationthatisshedfromthebladesintotheglobalflowfield.Thisisevenmore
importantforyawedconditionssincetheresultingcomplexskewedwakegeometryismore
difficulttoprescribe.Anotheradvantageofusingafreewakevortexmethodconcernsthe
factthatthewakegeometryisinherentlypartofthesolution.Thusitispossibletoderive
thepitchandexpansionofthehelicalwake,whichotherwisecouldbeobtainedusingtime
consumingsmokevisualizationexperiments.
Fig. 2.3 summarises the main problems and possible solution methodologies proposed in
thisprojectforusinglimitedexperimentaldatatoinvestigateindetailtheaerodynamicsof
windturbinesandperformathoroughassessmentofBEMbaseddesigncodes.
16
Chapter2AimofThesisandApproach
ResearchObjective:TouseexperimentaldatatocarryoutadetailedinvestigationofHAWTaerodynamics
andtoprovidefurtherinsightfordevelopingimprovedengineeringmodelsforBEMbaseddesigncodes
Requirements: Experimentaldatashouldideallyconsistof:
a.bladepressuremeasurements
b.detailedinflowmeasurementsinthenearwake
c.wakegeometrymeasurements
ProblemStatement: Thearetwomainproblems:
(1)Experimentaldatameetingtheabovethreerequirementsisusuallyunavailable
(2)Methodsusedforderivingtheangleofattackfrommeasurementsarestill
unreliable,especiallyforyawedrotorconditions.
TwoProblemCasesconsideredinProject
CaseA: CaseB:
Experimentaldataconsistsofdetailedinflowmeasurements Experimentaldataconsistsofbladepressuremeasurements
andwakegeometrydatabutbladepressuremeasurements butdetailedinflowmeasurementsandwakegeometrydata
arenotavailable arenotavailable
ThismethodologywasappliedontheNRELrotor
Figure 2.3 The main problems and possible solution methodologies proposed in this project for
using limited experimental data to investigate in detail the aerodynamics of wind turbines and
performathoroughassessmentofBEMbaseddesigncodes.
2.3.3DevelopmentofFreewakeVortexModel
The freewake vortex model used to analyse the NREL rotor (see section 2.3.2 above) was
developed during this project. This model is somewhat different than other freewake
vortexmethodsthatrelyonaerofoildatatoiterativelydeterminethebladeloading.Itwas
specificallydesignedtobeusedintheproposedmethodforfindingtheangleofattack.The
input to this code is a prescribed spanwise distribution of bound circulation that may be
timedependent.Fromthisprescription,thecodewillgenerateawakeandthencalculates
the3Dinducedvelocitiesatdifferentpointsintheflowfieldoftherotor.
The project also focused on the verification and validation of this new freewake vortex
model.ThehotfilmnearwakeinflowmeasurementscarriedoutontheDelftwindturbine
(refertosection2.3.1)wereusedasabasesforthevalidation.Theseinflowmeasurements
wereusedtogetherwiththeunsteadyaerofoiltheoryandtheKuttaJoukowskitheormto
determinetheboundcirculationdistributionsattheblades.Thesedistributionswerethen
17
Chapter2AimofThesisandApproach
prescribedtothevortexmodel.Thelatterthencomputedthewakeinducedvelocitiesand
thesewerecomparedwiththeinducedvelocitiesobtainedfromthehotfilmmeasurements.
TheprocedureforvalidatingthevortexmodelisillustratedinFig.2.4.Alimitationofthis
approach is the uncertainty due to the employed unsteady aerofoil theory. However the
measurementsarelimitedtoattachedflowconditionsonly,forwhichtheunsteadyaerofoil
model is considerably accurate. Apart from inflow measurements, smoke visualization
experiments were also carried out on the Delft rotor to measure the location of the tip
vortex paths. These measurements were also used to validate the freewake vortex model
(seeFig.2.4).
Deriveboundcirculationfromhotfilmmeasurements,
unsteadyaerofoilmodelandKuttaJoukowski theorm
Prescribeboundcirculationtofreewake
vortexmodeltogeneratehelicalwakeand
calculateinducedvelocitiesatrotorplane
Validationusinghotfilm Validationusingsmokevisualization
nearwakeinflowmeasurements measurementsoftipvortexpaths
Procedure: Procedure:
Figure2.4Validationprocedureofdevelopedfreewakevortexmodelusingmeasurementdatafrom
theTUDelftwindtunnelrotor.
2.4OrganizationofWork
Theresearchworkwasorganizedintodifferentprojectphases.Fig.2.5liststhesephases
in a chronological order. This dissertation documents the work carried out as follows: In
Chapter3,areviewoftheBEMtheoryforayawedHAWTispresented.Abriefliterature
survey of various engineering models developed in the past years for BEM is also
presented.Chapter4describesindetailthewindtunnelexperimentsandtheaerodynamic
analysiscarriedoutontheTUDelftturbineduringPhasesI,IIandIIIoftheproject.Chapter
5 presents the details of the freewake model developed in Phase IV together with its
verification and validation undertaken in Phase V. Chapter 6 describes the analysis
18
Chapter2AimofThesisandApproach
accomplished using the NREL experimental data in Phases VI and VII. From this study
furtherinsightonthelimitationsofBEMcodeswasobtainedandanumberofguidelineson
how the reliability of such codes can be improved are presented and discussed. These
guidelinesaregiveninChapter7.
aerodynamicloadsonblades
PhaseIII: AssessmentofBEMtheoryforattachedflowconditionsusingthe
TUDelft windturbineasacasestudyandresultsfromPhaseI &PhaseII
PhaseIV: Developmentoffreewakevortexmodeltobeusedintheproposed
methodforfindingangleofattackfromblade pressuremeasurements
PhaseV: VerificationandvalidationoffreewakevortexmodelusingtheTUDelft
experimentaldataobtainedusingPhaseI
PhaseVI:Applicationofproposedmethodforfindingangleofattackfromblade
pressuremeasurements&freewakevortexmodeltoNRELwind
turbine
PhaseVII: AssessmentofBEMtheoryforattached&stalledconditions
usingtheNRELwindturbineasacasestudyandresultsfrom
PhaseVI
PhaseVIII: GuidelinesforimprovingthereliabilityofBEMbaseddesign
codes
Figure2.5ProjectPhases.
19
Chapter2AimofThesisandApproach
2.5CoOrdinateSystemAnalysis
InFig.2.6,theXYZaxesarethefixedreferencesystemwhoseoriginisatthepivotcentre
atO.TheZaxisisverticalandalignedwiththetower.Itcoincideswiththeyawaxis.The
rotoraxismaybetiltedintheverticalplanebyafixedangle. istheyawangle.Thehub
centreisatFandisatadistancedafromOalongtherotoraxis.Eachblademayalsohavea
coning angle as shown in the diagram. is the azimuth angle of the first blade and is
equaltozerowhenthebladeisverticalandpointingupwards.Thetowerbaseislocatedat
T. T lies on the ground and vertically below O such that distance H is equal to the tower
height.TheXtYtZtaxesareidenticaltotheXYZaxeswiththeonlydifferencethattheir
originisatT.
ThecoordinatesystemswhoseoriginisatFincludethe:
xryrzraxesthesearerotatingaxeswiththeyraxisalignedwiththerotoraxis.The
anglebetweenthezraxisandthezaxisisequalto.
XaYaZaaxesthesearenonrotatingaxeswiththeYaaxisalignedwiththerotor
axis.TheZaaxiscoincideswiththezraxisatequaltozero.
The coordinate systems that are located locally at all the elements of each rotor blade are
the:
xyzaxesthesearelocatedinthesurfaceofrevolutionthatarigidbladewould
traceinspace,withtheaxisnormaltothissurface.When isequaltozero,they
axisbecomesparalleltotheYaaxis.
xpypzp axes these are the principal bending coordinates, where the zp axis
coincideswiththebladeselasticaxis.
axesthesearetheprincipalcoordinatesofthedeformedbladealongeach
pointontheelasticaxis.Forarigidrotortheseaxescoincidewiththexpypzpaxes.
The blade is considered to deflect in the flapwise direction, i.e. about the yp axis. For the
sake of simplicity, edgewise (leadlag) deflections are neglected. is equal to the angle
betweenthexand axisatagivenbladeelementandisequaltothelocalpitchangle.The
XnYnZncoordinatesystemshowninFig.2.6issimilartotheXaYaZabuthasitsoriginat
O.In theXYZ system, theZ axis coincides with the Z axis and the angle between the Y
andYaxesisequalto.
20
Chapter2AimofThesisandApproach
Za
Win
dve Zn
locit
y Z,Z
F
da
Rotor axis
O
Y
Xa Ya ,yr,Yn
Deformed
blade
X
Rigid Y
blade
Xn,X
zr y y
Zt
x
z,zp T
(Elasticaxis)
x Yt
Xt
Za zr
Rotorplane
Rotor axis
Ya
Viewstandinginfrontof
rotorlookingdownstream
Windvelocity
Xa
Figure2.6Coordinatesystemsusedformodellingthewindturbines.
21
Chapter2AimofThesisandApproach
TransformationMatrices
G G
A vector G in the XYZ reference may be transformed into an equivalent vector E in the
moving referenceframebymeansoftransformationmatrixSwhere
G G
E = S * G X Y Z
where
S = A6 * A5 * A4 * A3 * A2 * A1
A1A6 are orthogonal matrices that transform from one coordinate system to another
where
1 0 0 Cos Sin 0 1 0 0
A4 = 0 1
A5 = Sin Cos 0 A6 = 0 1 v'
0 1 0 0 1 0 v' 1
22
Chapter3TheBladeElementMomentumTheory
3.TheBladeElementMomentumTheory
ThischapterpresentsareviewoftheBladeElementMomentum(BEM)theoryforyawed
HAWTs.Itismainlyareformulationofthetheorytobefoundinmanytextbooks[53,77,
88]. The limitations of this theory are discussed and a brief overview of various BEM
engineeringcorrectionmodelsdevelopedoverthepastyearsisalsopresented.
3.1TheSimpleLinearMomentumTheoryforaYawedActuator
Disc
U
Skewed wake boundary
UC
n
o
i
s
US U
Way upstream
UC
n
os
Si
U
+u a
Actuator disc
UC
n
Si
o
s
U
+u a
'
Way downstream
Fig.3.1Yawedactuatordiscinskewedflow.
23
Chapter3TheBladeElementMomentumTheory
Theseyieldaresultantflowvelocityatthediscequalto
FardownstreamthevelocityperpendiculartotherotorplaneisequaltoUy +ua.Usingthe
Bernoulli energy equation together with the linear momentum equation, it can also be
provedthatuaistwiceua.Sinceawindturbineinitsnormaloperatingconditionextracts
energyfromthefluidstream,thentheflowvelocitydecreasesacrosstherotorandtherefore
ua is negative. For a yawed rotor, Glauert [30] expresses the momentum equation for the
axialthrustTas
T (3.3)
CT =
1 AU 2
2
Aisthecrosssectionalareaoftherotordisc.BysubstitutingEqts.3.1and3.2inEqt.3.3and
puttingua=a1UthefollowingexpressionforCTresultsin
Eqt. 3.4a is invalid for high loading conditions in which for CT approaches and exceeds
unity.Andersonetal.[3]haveobtainedthefollowingempiricalequationforvaluesof(a1)
largerthanabout0.38andatzeroyawangle:
TotheauthorsknowledgethereisyetnoempiricalequationavailablesimilartoEqt.3.4bthat
accountsforyawedflowinHAWTs.
Usingtheabovesimpletheory,itcanbeshownthatthemaximumpowercoefficientthatcan
beachievedbyayawedturbinediskisgivenby
4 ( a1, m ) [Cos ( ) + a ]
3 2
2
(
CP , Max a1, m , = ) 1, m
(3.5)
[Cos ( ) + 2a ]
1
2
1, m
24
Chapter3TheBladeElementMomentumTheory
=0= deg
0 ==15
15deg == 30
30 deg ==45
45deg . (5.4)
Eqt. 3.4b
3.5
2.5
CT (a1 ,)
1.5
0.5
0
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 1.1
- a1
Figure3.2Variationoftheaxialthrustcoefficientwiththeaxialinductionfactorfordifferentyaw
anglesaspredictedbyEqts.3.4aandb.
3.2TheMomentumEquations
Inderivingthelinearandangularmomentumequations,thefluidflowstreamatthediscis
divided into independent annuli or streamtubes. The elemental axial thrust T resulting
fromachangeinlinearvelocityineachstreamtubeisgivenbysubstitutingEqt.3.1inEqt.
3.2andreplacingAbythecrosssectionalareaofanannularelement.Thisresultsin
ua is the azimuthal averaged axial induced velocity for the given annulus. For (ua)>0.38U
Eqt.3.6aisinvalidandthefollowingequationisusedinstead
T = CT rU 2 r (3.6b)
whereCTisanempiricalequationsimilartoEqt.3.4b.Tofindtheelementaltorque Qata
givenannulus,itisassumedthattheswirlvelocitiesatagivenannulusfarupstreamand
fardownstreamoftherotoractinimaginaryplanesparalleltotherotorplaneofrotation.
The elemental torque is given by the rate of change of moment of momentum due to the
swirlinthestreamtube.UsingthisprincipleinconjunctionwithEqt.3.1resultsin
25
Chapter3TheBladeElementMomentumTheory
3.3TheBladeElementTheory
Thebladeelementtheory(BET)isusedtocalculatetheaerodynamicforces(andmoments)
on the blade due to its motion through the air (combination of wind velocity, induced
velocityandrotationalvelocity)todeterminetheperformanceandloadsontheentirerotor.
It is assumed that each blade behaves like a twodimensional aerofoil to produce
aerodynamicforces(liftanddrag)andmoments(pitchingmoments),withtheinfluenceof
the wake and the rest of the rotor contained entirely in an induced velocity at the blade
element.
3.3.1TheBladeElementVelocity
ThevelocityofapointAatagivenpointatradiusalongthe axisofarigidbladeinthe
reference frame is expressed by the following three equations of motion given in
Spera,[88]:
VA, = rCos ( da + r ) Cos + v Cos rSin Sin
(
VA, = rSin + v da + r SinCos + rCos Sin )
VA, = ( v ' r v ) Sin da + ( v v ' r ) cos Sin
(3.8)
3.3.2TheInducedVelocityatEachBladeElement
Inyawedrotors,thebladetobladeaerodynamicinterferenceandtheskewedwakeinduce
athreedimensionalinducedvelocityateachbladeelementwhichmay berepresentedby
thefollowingvectorinthexryrzrreferenceframe
ut
f
(3.9)
u = ua f
u
r
ua and ut are the average axial and tangential induced velocities respectively at a given
annulus. ur is the radial component of induced velocity. The BEM theory is incapable of
26
Chapter3TheBladeElementMomentumTheory
calculating the radial component of induced velocity and this is taken as zero. f is the
Prandtltip/rootlossfactorthataccountsforthefactthattherotorhasanoninfinitenumber
ofbladesandforthereducedloadingatthetip/rootoftheblades.Atagivenbladeradial
position,thisfactorisgivenby
f = f t * f r (3.10a)
where
B
2 1 r R (3.10b)
2
f t = Cos 1 exp t
r Sin
Rt
B r R
2 R r R (3.10c)
2
f r = Cos exp t
1 t
Rr Sin
Rt
is the inflow angle which is described in section 3.3.4. For the derivation of the Prandtl
tip/rootlossfactorreferto[77].
3.3.3FlowVelocityRelativetoaMovingBladeElement
Theflowvelocityrelativetoamovingbladeelementcanbecomputedbytransformingthe
windvelocityvectorfromtheXYZreferenceframetothe referenceframe,adding
theinducedvelocityvectorandsubtractingthebladeelementvelocity.ThematrixS(refer
toChapter2,page22)isusedtotransformUfromtheXYZframetothe frame.The
matrix A6A5A4 transforms the induced velocity vector u from the xryrzr frame into the
frame.Theflowrelativevelocityateachbladeelementbecomes
V( ) = SU ( X Y Z ) + A6 A5 A4u( xr yr zr ) VA( )
(3.11)
Themagnitudeoftheresultantflowrelativevelocityinthe planeisgivenby
Vr = V 2 + V 2
(3.12)
Themagnitudeoftheresultantflowrelativevelocityisthengivenby
Vrel = V 2 + V 2 + V 2
(3.13)
27
Chapter3TheBladeElementMomentumTheory
3.3.4AerodynamicLoads
The aerodynamic loads on each blade element are assumed to act on the axis which is
locatedatc/4awayfromthebladesleadingedge.Fig.3.3showsthevelocitytrianglesona
bladeelementatagivenradius.Theaerodynamicloadsarealsoshown.
In Fig. 3.3, the reference chordline and the axis are considered to coincide with one
another. Since the flow relative velocity has three components, there are two velocity
diagrams: one in the plane (Fig. a) and the other in the plane (Fig. b). The two
correspondinganglesofattackare:
1. normalangleofattack:
= tan 1 V (3.14)
V
2. lateralangleofattack(usuallydefinedasthesweepangle):
1 V (3.15)
sweep = tan V
sweep indicatesthepresenceofspanwiseflowanditsdirection.Itcontributestodelaythe
onsetofstall.Theinflowangle isequaltothesumofthelocalvaluesofand.
Bladesection
TrailingEdge
V
q,a y
V , z
Vr
A
V
LeadingEdge
A sweep
x Fig.(a) Fig.(b)
Figure3.3Velocitytrianglesandaerodynamicloadsatagivenbladeelement.
28
Chapter3TheBladeElementMomentumTheory
Theexpressionsfortheaerodynamicloadsontheelementofchordcandfinitewidth r
are
A = 1 cVr Cl ( , Re)V Cd ( , Re)V r
2
A = 1 2 cVr Cl ( , Re)V + Cd ( , Re)V r (3.16)
Theaerodynamicforceinthespanwisedirectionhasbeenneglected.TheReynoldsnumber
attheelementistakenas
V c
Re = r
a (3.17)
Thebladeelementalsoexperiencesapitchingmomentgivenby
q , a = 1 c 2Vr 2Cm r ee , a c A (3.18)
2
xa,c xe,a
ee,a = (3.19)
c
where
q,aincrementofaerodynamicpitchmomentloading(Nm)
Cmaerodynamicpitchingmomentcoefficientabouttheaerodynamiccentreof
theaerofoil
ee,arelativeeccentricityoftheelasticcentreofaerofoil
xa,cdistancefromleadingedgetoaerodynamiccentre(usuallyc/4)(m)
xe,adistancefromleadingedgetoelasticcentre(m)
3.3.5TheBladeElementEquationsforThrustandTorque
Referring to Fig. 3.3 and neglecting the effect of small blade deflections, the rotor axial
thrustandtorqueduetoabladeelementaregivenby
T = A Cos A Sin
(3.20a)
(
Q = rCos A Cos + A Sin )
(3.20b)
29
Chapter3TheBladeElementMomentumTheory
3.4TheBladeElementMomentumEquations
The BladeElementMomentum (BEM) Theory combines the momentum theory with the
blade element theory (BET) to determine the axial and tangential induced velocities. The
basicassumptionisthattheforceofabladeelementissolelyresponsibleforthechangeof
momentumoftheairwhichpassthroughtheannulussweptbytheelement.Itistherefore
assumed that there is no radial interaction between the elements. The BEM equation for
axialthrustisobtainedbyequatingEqts.3.6toEqt.3.20aandsimplifying.Thisyieldsthe
followingequationthatisusedtodetermineua.
2 r b =0
b =0
For(ua)>0.38U,thelefthandsideoftheaboveequationisreplacedbyCTU2,whereCTisan
empiricalexpressionsuchasEqt.3.4b.
The BEM equation for angular torque is obtained by equating Eqt. 3.7 toEqt. 3.20b. After
simplifying,thisyieldsanequationthatisusedtofindut.
b =0 b =0
IntheBEMtheory,Eqts.3.21and3.22aresolvediterativelytofindtheaxialandtangential
induced velocities. These are then used in the BET theory to find the required spanwise
aerodynamic load distributions. These loads are then integrated along each blade span to
yieldtheglobalrotorloads.Thesolutionprocedureisdescribedinsection3.6.
30
Chapter3TheBladeElementMomentumTheory
3.5CorrectionstotheBladeElementMomentumTheory
Previous validation efforts have revealed that the BEM theory may be considerably
accuratewhenmodellingaxialconditions(norotoryaw)wherethelocalanglesofattackat
the blades are small. However for conditions of high angles of attack and/or yawed flow,
the theory fails to predict accurately the blade load distributions that are required for
aeroelastic tailoring of the blades. Basically, the main sources for inaccuracies in the BEM
theoryaretwo:
(1) thelimitationsofmomentumequations:Eqts.3.6and3.7arebasedontheassumptionthat
each individual streamtube (or strip) can be analyzed independently of the rest of the
flow.Suchanassumptionworkswellfornonyawedconditionsandwhenthecirculation
atthebladesisrelativelyuniformsothatmostofthecirculationisshedatthebladeroot
and tip. However when the turbine is yawed, a timedependent circulation that varies
radially along the blades is formed. This creates a radial interaction and exchange
between flows through adjacent stream tubes and thus invalidates this assumption. As
alreadydescribedinsection1.2,asthewindflowsthroughtheyawedturbine,avortical
wake is created downstream of the rotor similar to that created by a helicopter rotor in
forwardflight,themajordifferencebeingthatthewakeexpandsratherthanitcontracts.
TheBEMtheorylacksthephysicstomathematicallymodelhowthewakecharacteristics
affect the distribution of induced velocity at the rotor disk. When a rotor is yawed, the
wake becomes skewed. In a yawed rotor, the trailing and shed vorticity shed from the
blades into the wake is on average closer to the downwind side of the rotor plane
resulting in higher induced velocities in this region. The upwind side will experience a
lowerinducedvelocity.Consequently,thelocalinducedvelocitiesatthebladeswillvary
considerably from the azimuthally (annular) averaged values. However the BEM
equations 3.21 and 3.22 are only capable of calculating an axial and tangential induced
velocityateachstreamtubethatareazimuthallyaveraged( ua and ut ).
(2) theinaccuraciesintheaerofoildata:Intheearlydays,2Dstaticwindtunnelaerofoildata
(Cl and Cd) were used to compute the aerodynamic loads on wind turbine blades with
BEMtheory.Duetothecomplex3Dnatureoftheflowoverrotatingwindturbineblades,
the aerofoil characteristics will vary considerably from the 2D static aerofoil
characteristics,especiallyattheinboardsectionsandatthetip/rootregionsoftheblades.
As a result, the use of 2D static aerofoil data did not yield a good correlation of the
calculatedaerodynamicloadswiththoseobservedinexperiments.Inawindturbinetwo
aerodynamic phenomena take place: (a) Stalldelay phenomena and (b) Unsteady flow
phenomena.
(a) Stalldelay: Since the 1940s, Himmelskamp [39] investigated the aerodynamic
behaviourofpropellersandnotedthattheliftforcesonarotatingbladearelargerthan
those on a nonrotating one. The same phenomenon was observed on wind turbine
31
Chapter3TheBladeElementMomentumTheory
bladesinthepastyears[13,62,74,75,82,87,103].Evidenceshowsthatforattachedflow
conditions, below what would otherwise be the stall angle of attack of a nonrotating
aerofoil, there is little difference between the 2D flow conditions and the rotating
conditions.However,whenstalloccurs,theairintheseparatedregion,whichismoving
veryslowlywithrespecttothebladesurface,isrotatingwiththebladeandthereforeit
experiencesacentrifugalforcecausingittoflowradiallyoutwards.Theflowtowardsthe
tiponthesuctionsideexperiencesaCoriolisforceinthemainflowdirection,actingasa
favourable pressure gradient. This reduces the displacement thickness of the boundary
layer,delayingtheonsetofstallandresultinginhigherliftcoefficients.Thisphenomenon
isoftenknownasstalldelayandismostpredominantintheinboardsections.Theeffect
ofstalldelayontheliftcharacteristicsofarotatingwingisillustratedinFig.3.4.
Cl
Cl,3D
Cl,2D
Figure 3.4 Comparison of the lift characteristics of a rotating wing (Cl,3D) with those of a static
nonrotatingwingin2Dflow(Cl,2D).
(b) Unsteady flow: In certain operating conditions of a wind turbine such as rotor yaw,
eachbladeelementaerofoilissubjectedtoanunsteadyangleofattackandflowvelocity.
Thiscausestheliftanddragcoefficientstobedifferentfromthecorresponding2Dstatic
values at the same angle of attack. When the timedependent variation of the angle of
attackisbelowthestallangle(s)forstaticconditions,theflowoverthebladesremains
attachedandthevariationofliftwillbesimilartothatshowninFig.3.5.Whentheangle
of attack at a blade section exceeds the aerofoils static stalling angle, dynamic stall
occurs. In a 2D nonrotating environment, the phenomenon of dynamic stall is
characterizedbyadelayintheonsetofflowseparationtoahigherangleofattackthan
would occur statically. This results in higher lift. When flow separation does occur, a
sheddingofaconcentratedvorticaldisturbanceisformedattheleadingedge.Aslongas
thisvortexremainsontheupperaerofoilsurface,itproducesenhancedlift.Howeverthe
flow causes the vortex to be swept over the chord towards the trailing edge. This
produces a state of full separation resulting in a rapid aft movement of the centre of
pressureandanincreaseinthepitchingmoment,togetherwithadropinlift.Iftheangle
32
Chapter3TheBladeElementMomentumTheory
Cl
Cl,Unsteady
Cl,2D
s
Figure3.5Typicalvariationoftheunsteadyliftcoefficientforsmallanglesofattack(<s).
ofattackisreducedwellbelowthestaticstallangle,flowreattachementmaytakeplace.
Yetflowreattachmentcanonlytakeplaceiftheangleofattackbecomessmallenough
again. There is generally a significant lag in this process until the fully separated flow
reorganizesitselfuntilitisreadyforreattachment.Fig.3.6showsatypicalvariationof
the lift coefficient with angle of attack together with a schematic explaining the flow
topologiesobservedindynamicstall.
The delay in flow separation and the lag in the flow reattachment process results in a
hysteresis variation. These hysterisis effects introduced by dynamic stall may have a
negativeeffectontheaeroelasticdampingbehaviourofwindturbineblades.Thisinturn
reducesthefatiguelifetimeleavinganadverseimpactontheeconomicsofthesystem.
Thephenomenonofdynamicstallisnotfullyunderstoodandisstillundergoingmuch
research.Muchofwhatisknownaboutdynamicstallhasbeenobtainedfrom2Dwind
tunnel experiments on nonrotating wings. Dynamic stall also occurs in a rotor
environment where it has a much more threedimensional character and depends on
boththeradialandazimuthpositionsontheblades.
33
Chapter3TheBladeElementMomentumTheory
(3)
Cl
(2) Cl,Unsteady
(1) (4)
Cl,Steady
(5)
Stage 1: Aerofoil exceeds and flow reversal Stage 3-4: Lift stall occurs. After vortex reaches
s
occurs in upper boundary layer trailing edge, full separation occurs.
Figure 3.6 Schematic showing the unsteady lift coefficient and the basic flow topologies during
dynamicstall.AdaptedfromLeishman[49].
34
Chapter3TheBladeElementMomentumTheory
To reduce the uncertainties in the BEM theory due to the limitations of the momentum
equationsandduetoinaccuraciesintheaerofoildata,variouscorrectionswereincludedin
thepastyears.ThishasresultedinthesocalledextendedBEMtheory.Thesecorrections
took the form of basic engineering models that were derived from experimental data or
datafrommoreadvancedcodes(basedonvortextheoryorCFD).Ashortreviewofsome
ofthesemodelsisnowpresented.Thesemodelsfallundertwodifferentclassifications:
Type I models: those that correct the axial induction factor computed by the BEM
theoryforthenonuniforminductiondistributionattherotorplaneresultingfromthe
skewed wake of a yawed rotor. An early model for skewed wake effects has been
proposedbyGlauert[30].Itestimatestheaxialinducedvelocityatthebladesusingthe
equation
r
u
a ,c = ua 1 + K * * Sin
R
(3.23)
whereKdependsontheyawangle.Eqt.3.23wasderivedfromthesmokevisualization
of the fully rollup strong tip vortices formed on helicopter rotors in forward flight.
VariousformulasforKwereproposed[49]whicharegiveninTable3.1. sisthewake
skewanglewhichiscalculatedfrom
tan USin (3.24)
=
s
UCos + ua
Since in the equation above ua will vary radially, then s will also vary radially.
Howeversisusuallytakentobeequaltothatbetween7080%R.
Table3.1VariousmodelsforparameterK(Eqt.3.23)
Author(s) K
Othermodelsweredevelopedinthepastyears,someofwhichwereexaminedagainst
measurementsintheJOULEDynamicInflowprojects[83,84].Thesemodelshavebeen
implemented in various BEMbased aeroelastic models. One such model is that
developed by the DTU [60] that is similar to Glauerts model but introduces a radial
variationfortheinduction.Themodelwasderivedwithacurvefittingprocedurefrom
35
Chapter3TheBladeElementMomentumTheory
anactuatordiskvortexringmodel.Thelocalinducedvelocitywasfoundtodependon
theradiallocation,azimuthangleandthewakeskewangleaccordingto:
ua , c
= ua 1 + f 2,tudk r
( R ) tan 2 sin ( ) (3.25a)
s
where
( R)
3 5
r r r
f 2,tudk r = + 0.4 + 0.4 (3.25b)
R R R
Other engineering models for yaw were developed in the JOULE I and II projects.
Furtherdetailsmaybefoundinreferences[83]and[84].
AmajorshortcomingofEqt.3.23isthatitconsidersonlytheinducedvelocitydueto
thetipvorticityalone.Itisshownthatvorticityoriginatingfromthebladerootaswell
asshedvorticitywillcausetheinductiondistributionattherotordisktohaveahigher
harmonic content than that modelled by Eqt. 3.23. This was revealed in past inflow
measurementstakenontheDelftwindtunnelmodel[69,96].ECN[69]hasdeveloped
anewengineeringmodelthataccountsforsucheffects.Thisconsistsofasecondorder
Fourierserieshavingtheform:
ua , c = ua [1 A1 cos ( 1 ) A2 cos ( 2 2 )] (3.26)
whereamplitudesA1andA2andphases 1and 2havebeenmodelledasafunctionof
radialpositionandyawangle.
IntheBEMmodeldescribedinsections3.3and3.4,skewedwakeeffectsareaccounted
forbymodifyingEqt.3.9andintroducingacorrectionfactorFsaasfollows:
ut
f
ut ,c
F = u
u
uc =
f S a a ,c (3.27)
a
ur ,c
0
where Fsa determines the ratio of the local axial induced velocity at the blades to the
azimuthally averaged value as modelled by anyone of the engineering models
describedabove.Forinstanceforf=1,ifGlauertsengineeringmodelistobeused,then
Fsawouldbeequalto
r
Fsa = 1 + K * * Sin (3.28)
R
36
Chapter3TheBladeElementMomentumTheory
AsitcanbenotedfromEqt.3.27,thecorrectionforskewedwakeeffectsisonlybeing
applied to the axial component of the induction. No correction is done for the
tangentialandradialcomponentssincetheseareusuallyverysmallinmagnitudeand
thustheirinfluenceonaerodynamicloadingisinsignificant.
Type II models: those that correct static 2D aerofoil data for 3D rotating effects (stall
delay)andunsteadyaerodynamiceffects(unsteadyaerofoilmodelsforbothattached
flowanddynamicstall).
EngineeringModelsforStallDelay
An early empirical model for modifying 2D static aerofoil data to represented more
accurately thepower augmentation at highangles of attack resulting from stalldelay
has been developed by Viterna and Corrigan [100] in 1981. This model was used
extensivelyinthepastyearsforwindturbinemodellinginstalledflowconditions.The
equationsforthismodelareasfollows:
For s :
Cd ,max cos 2
Cl = sin 2 + K l (3.29a)
2 sin
Cd = Cd ,max sin 2 + K d cos (3.29b)
sin s
K l = ( Cl , s Cd ,max sin s cos s ) (3.29c)
cos 2 s
Cd , s Cd ,max sin 2 s
K d = (3.29d)
cos s
50 : Cd ,max = 1.11 + 0.018
(3.29e)
> 50 : Cd ,max = 2.01
where Cd ,max isthemaximumdragcoefficientandisthebladeaspectratio.
Inyear1993,Sneletal.[82]presentedamethodtoevaluatethefirstordereffectsofthe
bladerotationonstallcharacteristicsthroughasimplifiedsolutionofthe3Dboundary
layers equations. An order of magnitude analysis of the different boundary layer
equations was carried out to enable the identification of the most important
37
Chapter3TheBladeElementMomentumTheory
parameters. It was shown that the local blade solidity (c/r) is the most influential
parameteraffectingstalldelay.Inthiswork,asimpleempiricalmodeltocorrect2Dlift
coefficient data for stalldelay was developed with observation of wind tunnel data
(Ronsten, [62]) and results from a CFD model ULTRAN_V developed at NLR. The
modelisgivenby
b
c
Cl ,3 D = Cl ,2 D + a Cl (3.30a)
r
where
Cl = Cl ,lin Cl , 2 D (3.30b)
aandbareempiricalconstants.Cl,linistheliftcoefficientthatwouldbeobtainedifthe
linear part of the static 2D Cl curve is extended beyond stall. Although this model
improvedagreementforpowerpredictionwhencomparedwithexperimentaldata,it
is limited due to the fact that the drag coefficient remains uncorrected. It is a well
knownfactthat3Drotatingeffectsmayalterthe2DCd valuessignificantly,especially
attheinboardregions.
AthirdmodelwasdevelopedbyDuandSelig[22].Thistracesitsrootsintheworkof
Snel et al. [82] as it also originates from the 3D incompressible boundary layer
equations for a rotating system. A rigorous analysis of the integral boundary layer
equationsisapplied.Themodelcorrectsboththeliftanddragcoefficientsasfollows:
Cl ,3 D = Cl ,2 D + fl ( Cl ,lin Cl ,2 D ) (3.31a)
Cd ,3 D = Cd ,2 D f d ( Cd ,2 D Cd ,0 ) (3.31b)
Cl ,lin = 2 ( 0 ) , Cd ,0 = Cd ,2 D for = 0 (3.31c)
fl and fd are factors that depend on the separation point of the flow on the aerofoils
upper chamber as predicted by the boundary layer theoretical analysis. These factors
are related to the local solidity, c/r, (which accounts for rotor geometry) and the
modifiedtipspeedratio,(whichaccountsfortheeffectsofrotation). isgivenby
R U + ( R ) .flandfdaregivenby
2 2
dR
1 1.6 ( c r ) a ( c r ) r
f l = 1 (3.32a)
2 0.1267 b + c r r
dR
( )
d R
1 1.6 ( c r ) a ( c r ) 2 r
f d = 1 (3.32b)
2 0.1267 b + c r 2 r
d R
( )
38
Chapter3TheBladeElementMomentumTheory
In the above equations, terms a, b and d are empirical correction factors. Another
engineeringmodelwasdevelopedbyChaviaropoulosetal.[17]usingCFDandcorrects
all aerofoilcoefficients (Cl, Cd and Cm) for 3D rotating effects. The model was derived
basedonresultsobtainedfroma3DincompressibleflowNavierStokessolver.Italso
accountsfortheeffectsofbladetwistsincethiswasfoundtoplayanimportantrolein
massivelyseparatedflow.Themodelequationsare
= C X ,2 D + a ( c r ) cos n ( twist ) C X where X = l , d , m
h
C X ,3 D
(3.33a)
where
Cl = Cl ,lin Cl , 2 D (3.33b)
Cd = Cd ,2 D Cd ,2 D MIN (3.33c)
Cm = Cm ,2 D Cm ,lin (3.33d)
where a, h and n are empirical constants. However validations studies [45, 65, 73] on
differentrotorssuggestthesemodelsmaynotalwaysbesufficientlyaccurate.
EngineeringModelsforUnsteadyFlowEffects
Examples for unsteady aerofoil models used in attached flow conditions are
Theordorsens model [93] and Leishmans indicial response method using Duhamels
superimposition [49, pp 336340]. The latter model is described in detail in chapter 4,
section 4.3.2. Examples of dynamic stall models include the BoeingVertol model, the
ONERAmodel[11],andtheBeddoesLeishmanmodel[47].Themoststraightforward
model is the BoeingVertol model which is based on correcting the static 2D lift
coefficientinaccordancewiththefollowingequations
0
Cl ( , ) = Cl ,2 D ( ) (3.34a)
0
istheshiftintheangleofattackgivenby
1
c 2
= sign ( ) (3.34b)
2Vr
where is an empirical constant. A brief description of the different dynamic stall
models is presented by Leishman [49, 50]. Snel [79] developed a heuristic model for
dynamicstallbasedontheobservationofexperimentaldata.
39
Chapter3TheBladeElementMomentumTheory
Most unsteady aerofoil models for attachedflow conditions have been derived based
on2Dnonrotatingwingsandthereforetheremaybeinaccuratewhenappliedfor3D
conditions on a rotating wind turbine blade. Several dynamic stall models are semi
empirical and were also derived from nonrotating 2D wing experiments.
Consequently they are also inaccurate when treating 3D dynamic stall on a rotating
blade. Due to the blade advancingandretreating effect resulting from yaw, each
rotating blade is subjected to unsteady radial flow components that may be much
larger in magnitude than in a nonyawed rotor. Such radial flows influence the
dynamicstallbehaviorsignificantly,asdescribedinreference[48].Althoughitisawell
known fact that radial flow over the blades helps in preventing flow separation over
thebladesathighanglesofattackandthuscontributestostalldelay,itisstillunclear
how3Deffectsinfluencestallinanunsteadyenvironment.
Although the inclusion of both Type I and Type II engineering models improved BEM
aerodynamic load predictions, better models are still required. This wasnoted during the
blind comparison investigation organized by National Renewable Energy Laboratory
(NREL) way back in the year 2000 [73, 78].Considerablediscrepanciesbetween predicted
and experimental results were observed even at low windspeeds at which the angle of
attackissmall.
3.6DescriptionofProgramHAWT_BEM
HAWT_BEMisaBEMcodedevelopedinthisprojectusingMathCadversion11andis
applicabletobothaxialandyawedrotors.Thiscodewasusedforallcomputationsrequired
withtheBEMtheorythroughoutthisproject.
3.6.1TimebasedNumericalSolutionfortheBEMEquations
ThissectiondescribesthenumericalsolutionoftheBEMtheoryequationsdescribedabove
asimplementedinHAWT_BEM.Sinceinayawedturbinetheflowatthebladesbecomes
unsteady,thenthesolutionhastobesolvedasafunctionoftime(orrotorazimuthangle ).
Consider the situation in which the rotor is rotating at constant angular speed in a
uniformwindfrontequaltoU.Onewholerotorrevolutionisdividedintoafixednumber
ofazimuthsteps(tot),asshowninFig.3.7.Anindex isusedtodenotetheazimuthangle
of the first blade. =0 denotes when the first blade is at an azimuth angle equal to zero
(vertical pointing upwards). The azimuthal step, , is equal 2/tot while the incremental
timestep,,isequalto/.
Eachturbinebladeisdiscretizedintoafixednumberofequallyspacedsectionsasshownin
Fig.3.8.Thetotalnumberofbladesectionsisequalton.Thecrosssectionalareaofthe
40
Chapter3TheBladeElementMomentumTheory
=0
= tot-1
=1
= 10
D
ro irec =2
to tio
rr n
ot o
ati f
on
=9 =3
=8 =4
=7 =5
=6
Figure3.7Divisionofonewholerotorrevolutionintoafixedamountofazimuthalsteps.
rotorplane swept by the blades (from r=Rr to r=Rt) is divided into n annular elements (as
illustrated by the shaded area in Fig. 3.9). An index notation is used to represent each
parameterateachrotortimestep(), blade(b),andradiallocation(i).Theparameterslocal
to the blade (equal angle of attack, Reynolds number, flow relative velocity and
aerodynamicloading)aredenotedbythethreeletterindexnotation(,b,i).Forinstancethe
angleofattackatparticularbladeelementisdenotedby ( ) ,b ,i .Theparametersthatare
onlyafunctionoftheradiallocationandrotorazimuthangleareonlydenotedbyatwo
letter suffix notation (, i). Thus the azimuthally averaged axial and tangential induced
velocitiesateachannulusaredenotedby ( u a ) ,i and ( ut ) ,i respectively.Thetotalnumber
ofbladesisequaltoBandb=0denotesthefirstblade.Giventhat ( ) istherotorazimuth
angleattimestep,thentheazimuthangleofeachbladeisgivenby
2 b
( ) , b = ( ) +
B
rn-1 = Rt
ri+1
ri
r
ri-1
r0 = R r
Rotoraxis
Figure3.8Discretizationofbladeintoafixednumberofequallyspacedbladesections.
41
Chapter3TheBladeElementMomentumTheory
ri
Rt
Rr
Figure3.9Divisionofsweptareabybladesintoafixednumberofannuli.
The solution starts by assuming initial values for the azimuthally averaged axial and
tangential induced velocities for each time step and radial location ( ( u a ) ,i and ( ut ) ,i ).
The solution is started with an impulsive start of the rotor. Initially, the rotor is at an
azimuth angle of zero (=0) and the following sequence of steps is applied to each blade
elementforallthetimesteps(tot)inonewholerotorrevolution:
Step1:Theabsolutevelocities, VA , ( ) ,b ,i
(
, VA , ) ,b ,i
(
and VA , ) ,b ,i
arefoundusingEqts.3.8.
Step 3: The Prandtl tip/root loss factor, ( f ) ,b ,i is then calculated in accordancewith Eqt.
3.10.
Step4:Thelocalinducedvelocitiesateachbladeelement, ( uc ) ,b ,i arefoundfromEqt.3.27
Step 6: The resultant flow relative velocities, (Vr ) ,b ,i and (Vrel ) ,b ,i are found using Eqts.
3.12and3.13.
42
Chapter3TheBladeElementMomentumTheory
Step 7: The angle of attack, sweep angle and Reynolds number, ( ) ,b ,i , sweep ( ) ,b ,i
and
( Re ) ,b ,i arefoundfromEqts.3.14,3.15and3.17.
Step 8: The lift and drag coefficients, ( Cl ) ,b ,i and ( Cd ) ,b ,i together with the pitching
momentcoefficient ( Cm ) ,b ,i arefoundfrom2Dwindtunneldatawhichmaybecorrected
forstalldelay,orunsteadyflow(e.g.dynamicstall)usingaTypeIIengineeringmodel(refer
tosection3.5.)
Step9:Eqts.3.21and3.22arethensolvedtoyieldnewvaluesfor ( ua ) ,i and ( ut ) ,i .
( A ) ,b,i and ( q ) ,b,i are evaluated using Eqts. 3.16 and 3.18. These are then
,a
integrated numerically to find the resulting 3D aerodynamic forces and moments at the
yawbearinginaccordancewiththemethoddescribedinAppendixB.
3.6.2ProgramStructure
Fig. 3.10 describes the structure of code HAWT_BEM. This code is organized into three
separate modules: the Data Input Module in which the parameters describing the rotor
geometry and operating condition are inputted, together with the required aerofoil data.
The Data Processing Module implements the numerical solution of the BEM equations
described above in section 3.6.1 to determine the spanwise distributions of the various
aerodynamic parameters at different blade azimuth positions. The 3D aerodynamic loads
inducedbytherotorbladesattheyawbearingandtherotoroutputpowerarecalculated
usingthesolutiondescribedinAppendixB.TheDataOutputModuleoutputsthelocalblade
androtorglobalresultsasafunctionofblade/rotorazimuthangle.
43
Chapter3TheBladeElementMomentumTheory
DataInput:
(1)InputRotorGeometryDetails:
NumberofBlades(B)
(2)InputRotorOperatingConditions:
RotorAngularSpeed()
RotorYawAngle()
WindSpeed(U)
(3)InputAerofoilData:
Inputof2Daerofoildata
SelectionofTypeII modelsforstalldelay/unsteady
effects
DataProcessing:
ApplicationofnumericalsolutiondescribedinSection3.6.1
todeterminethespanwise distributionsofthefollowing
uc ,sweep,
parametersatdifferentbladeazimuthangles:,
Re,V ,V ,Vr,Vrel,Cl,Cd,Cm,dA,dA
CalculationofAerodynamicLoadsandOutputPower
inducedatyawbearingusingnumericalsolutiondescribed
AppendixB
DataOutput:
OutputofresultsfrommoduleDataProcessing at
eachblade/rotorazimuthangle()
Figure3.10StructureofcodeHAWT_BEM.
44
Chapter4 AerodynamicAnalysisoftheTUDelftModelTurbine
4.AerodynamicAnalysisoftheTUDelftModel
Turbine
4.1Introduction
ThischapterdescribestheaerodynamicanalysiscarriedoutontheTUDelftmodelwind
turbine. A series of experiments were carried in the Delft University of Technology open
tunnel jet facility. The experiments consisted of detailed hotfilm inflow measurements in
the near wake of the turbine and smoke visualization experiments to trace the tip vortex
paths in the turbine wake. The experiments were carried out in both axial and yawed
conditions. The inflow measurements were limited to one rotor tip speed ratio and blade
pitch setting only that yielded attached flow conditions over the blades. The smoke
visualization experiments were carried out at different tip speed ratios and blade pitch
settings that resulted in both attached and stalled flow at the blades. The measurements
were mainly required for three reasons: (1) to obtain a better understanding of the
aerodynamicsofwindturbinesinyaw;(2)toassessthelimitationsofBEMmodelsinyaw
forattachedflowconditionswheretheuncertaintyintheaerofoildataisnottheissuethat
limitsBEMmodelsfrompredictingloadsaccurately.Thisassessmentwasafirststepbefore
carrying out a more extensive assessment on the NREL rotor in both attached and stalled
conditions (refer to Chapter 6); (3) to use the experimental data to validate the newly
developedfreewakevortexmodelinbothaxialandyawedconditions(refertoChapter5).
Thischapterisorganizedinthreeseparatesections:
A. Section 4.2 will describe the experiments in detail together with the data reduction
procedures that were required to obtain the required experimental data. The main
experimentalresultsarealsopresented.
B.Section4.3implementsamethodforderivingthesteady/unsteadyboundcirculationand
aerodynamicloaddistributionsatthebladesbycouplingtheinflowmeasurementswithan
45
Chapter4 AerodynamicAnalysisoftheTUDelftModelTurbine
unsteady aerofoil model. In this procedure, the angle of attack values at the blades are
estimated from the hotfilm measurements and usedin the unsteady aerofoil model to be
abletodeterminetheliftcoefficients.Thelatterarethenusedtoestimatethedistributions
for the bound circulation and aerodynamic loads at the blades using the bladeelement
theory.Tobeabletocarryoutthesecomputationsinanefficientandorganizedmanner,a
dedicatedcomputerprogram,HAWT_LFIM,wasdeveloped.
C.Section4.4dealswiththeassessmentofatypicalBEMcode(HAWT_BEM,seesection3.6)
usingboththeexperimentalinflowmeasurementsandtheaerodynamicloadresultsfrom
sections4.2and4.3.
RotorDetails
Thewindtunnelmodelrotorwasahorizontalaxiswindturbinewiththespecifications
listedintable4.1below.
Table4.1Windturbinegeometricspecifications
Numberofblades 2
Airfoilsection NACA0012
RotorradiusR 0.6m
Bladerootradius 30%oftipradius
Chordc 0.08m(constant)
Bladelength 0.42m
Bladetwist (r/R)=(6+tip)6.67(r/R),0.3r/R0.9
(r/R)=tip,0.9<r/R1
Figure4.1WindturbinemodelatDelftUniversityofTechnology.
46
Chapter4 AerodynamicAnalysisoftheTUDelftModelTurbine
Therotorshaftanditsbearingsareplacedonanextendedsupportleaving0.75mfreespace
behindtherotorplanetominimizetheinteractionofthedevelopedwakewiththesupport
structure within a distance of about one rotor radius from the rotorplane. The rotor hub
heightis2.33mabovetheground.Avariablepitchmechanismisinstalledintherotorhub,
capableofadjustingthebladepitchwithaccuraciesof0.10.Therotorislinkedtoaconstant
speed drive unit consisting of a 1.5kW motor/generator, with the rotational speed that is
adjustable from 0 to 16Hz. Strain gauges are installed on the rotor shaft and one of the
blades to be able to measure the rotor axial thrust and blade root edgewise and flapping
bending moments. The aerodynamic behavior characteristics of the model turbine when
operatinginaxialconditionsareshowninFigs.4.2and4.3.
CT
Figure4.2VariationofaxialthrustcoefficientwithtipspeedratioandbladetippitchangleTheta
(tip)(deg)(Source:Vermeer[97]).
CP
Figure4.3VariationofpowercoefficientwithtipspeedratioandbladetippitchangleTheta(tip)
(deg)(Source:Vermeer[97]).
47
Chapter4 AerodynamicAnalysisoftheTUDelftModelTurbine
4.2WindTunnelMeasurements
The wind tunnel measurements on the model turbine were performed in the open jet
wind tunnel of Delft University of Technology. Both the wake inflow and smoke
visualization experiments were carried out in close collaboration with Phd colleague
WouterHaans.
4.2.1WindTunnel
Theopenjetwindtunnelconsistedofaflowchannelwithacircularcrosssection,alarge
inletfan,flowstraightenersandgauzes.Fig.4.4isaschematicdiagramofthetunnel.The
flow straightener was a hexagonal shaped honeycomb structure made out of thin
aluminumsheets.Therewerethreeidenticalgauzes:onejustbehindtheflowstraightener
and two spaced by 0.2m at 2m upstream of the tunnel jet exit. The tunnel had an exit jet
diameterof2.24manditscentralaxiswas2.33mabovetheground.Thefanwasdrivenbya
45kWdcmotorinaWardLeonardcircuit.Theexitjetwindvelocitycouldbeadjustedby
controlling the rotational speed of the fan.The tunnel maximum windspeed was equal to
14.5m/s. The turbulence level was equal to 1.20.2% at Ujet =5.5m/s, the speed at which
measurementsweretaken.Theexitjetvelocityprofilewasnotuniformthroughout,asmay
benotedinFig.4.5.Avelocitydipwasobservedatthecentreoftheexitandthisismainly
duetothecentrebodycontainingthemotorfandrivemechanism.
Thetunnelwassituatedinahall(length35m,width20m,height5.5m)andthetunnelexit
wasapproximately11mfromthebackwall.
gauze
wind
gauze
gauze
motor
Figure4.4OpenjetwindtunnelatDelftUniversityofTechnology.
48
Chapter4 AerodynamicAnalysisoftheTUDelftModelTurbine
0.93
0.4 95
0.
0.93
0.3
0.93
0.95
0.9
0.2
0.99
0.99
0.95
0.1
y [m]
0
95
0.
0.
99
0.93
-0.1 0. 9
0.93
1
0. 9
0.95
3
0. 97
-0.2
0.9
5
-0.3
0.97
-0.4
. 93
0
0.9
0.9
1
3
-0.5
99
0.
-0.4 -0.2 0 0.2 0.4 0.6
x [m]
Figure4.5Velocitydistributionmeasuredintheemptytunnelat1mdownstreamfromthetunnel
jetexitwiththepitotreadingssetto5.5m/s.Thevelocitiesarenondimensionalizedwithrespectto
themaximumvelocityrecorded.(x,y)=(0,0)coincideswiththecentralaxisofthetunnel.
4.2.2PartI:InflowMeasurements
ExperimentalSetup
Therotorwasplacedinfrontofthewindtunnelexitwiththehubcentrelocatedinthejet
centreand1mdownstreamfromthejetexitplane.Fig.4.6isaschematicoverviewofthe
rotor and its position relative to the wind tunnel. The conventions for the yaw angle and
azimuthanglearealsoshown.Seenfromabove,apositiveyawangleimpliesthattherotor
is rotated counterclockwise. Standing in between the tunnel exit and rotor while looking
downwind,therotorazimuthangleincreasesastherotorrotatesintheclockwisedirection.
Theazimuthangleiszerowhenthefirstbladeisverticalandpointingupwards.Theaxial
distanceisdefinedwithrespecttotherotoraxis.Thetunnelwindvelocitywasmeasuredat
the jet exit plane using three interconnected pitotstatic tubes that were connected to an
electronicpressuresensor,togetherwithambientpressureandtemperaturereadings.
Thenearwakevelocitiesoftherotormodelweremeasuredusingconstanttemperaturehot
filmanemometry.Thistechniquemakesuseofhotwiresorhotfilmsthatactassensorsand
areveryusefulinobtainingthefastresponsevelocitymeasurements.Usingdifferentprobe
49
Chapter4 AerodynamicAnalysisoftheTUDelftModelTurbine
p
Hot-wire probe
(i, j)
Probe
position
1m
Traversing
system
U
R=1.12m
Front view Top view
Figure4.6Schematicoverviewoftestsetupwithframeofreference.
Figure4.7Themodelwindturbineandopenjetwindtunnelused.
orientations or multisensor probes, it is possible to obtain threedimensional components
of complex flows. The sensor consists of a very fine wire or film that is attached between
twosupportingneedles.Thewireorfilmisusuallymadeoutofaquartzfiberandcoated
with platinum. Current passed through the sensor raises its temperature above the
adiabaticrecoverytemperatureofthegas.Thehotwireorfilmthenrespondstochangesin
total temperature and mass flux. In subsonic applications where the fluid temperature is
lowandconstant,theproblemofheattransferthroughthesupportneedle(endlosses)and
radiationeffectscanbeignoredandthesensorsresponsecanbetakentobeasafunctionof
50
Chapter4 AerodynamicAnalysisoftheTUDelftModelTurbine
theflowvelocityonly.Inaconstanttemperaturesystem,thehotwireorfilmismaintained
at constant temperature. Electronic circuits used in a constant temperature anemometer
include a bridge circuit with a feedback system to maintain the wire or film at constant
resistance. The bridge voltage is a measure of the cooling of the wire and therefore a
measureofvelocity.
In these wind tunnel inflow measurements, two different types of single hotfilm probes
were used: one with the film normal to the probe (TSI 120120) and one with the film
paralleltotheprobe(TSI121120),(refertoFig.4.8).Ineachofthetwoprobes,thehotfilm
consists of a platinum film on a fused quartz substrate. The two probes are very similar,
withtheexceptionforthehotfilmorientation.
Fig(a):Normalprobe(TSI120120)
Fig(b):Parallelprobe(TSI121020)
Figure 4.8 The normal and parallel types of single hotfilm probes used. (Courtesy: TSI
Instruments).
ExperimentalProcedure
The experiments were carried out at a rotor speed and tunnel velocity of 11.65Hz and
5.5m/s, yielding a tip speed ratio of 8. This yielded a Reynolds number equal to about
150,000 at the blades. The blade tip pitch angle was set to 20. This was very close to the
conditionsforpeakpower(seeFig.4.3).Theexperimentswerecarriedoutatdifferentyaw
angles of the rotor (00, 300 and 450). A major disadvantage of using hotfilms in rotor
experiments is that it is physically impossible to measure the inflow directly in the
51
Chapter4 AerodynamicAnalysisoftheTUDelftModelTurbine
rotorplane.Anestimatehadtobemadebytakingwakemeasurementsatdifferentplanes
parallel to the rotor plane both upstream and downstream as shown in Fig. 4.9.
Interpolationwasthenappliedtoderivethewakevelocitiesintherotorplane.Thehotfilm
measurementsweretakenatthefollowingplanes:3.5cm,6.0cmand9.0cmdownstreamof
rotorplane and 6.0cm upstream of rotorplane. In dimensionless form, these distances are
equivalent to Ya/R equal to 0.058, 0.1 and 0.15 downstream and 0.1 upstream of the
rotorplane.Foreachplane,themeasuringpointswerelocatedatradialpositions40,50,60,
70, 80, 90 and 100%R and at azimuth increments of 150. These yielded traces of velocity
against overawholerotorrevolution(i.e.00to3600).TermsiandjinFig.4.6areintegers
thatdenotetheradialandazimuthpositionsrespectivelyofeachmeasurementpointwithin
themeasurementplane. isanintegerrepresentingtheazimuthpositionoftherotor.Note
thattheprobeazimuthangle(p)isdifferentfromtheazimuthangleoftherotor().
Whilsttherotorwasrotating,theprobewaspositionedatdifferentiandjpositionsineach
measuring plane. For each measurement point, the hotfilm readings were taken every 20
increments of rotor azimuth angle. At each point, 54 velocity traces were taken
corresponding to 54 consecutive rotor revolutions and the mean velocity trace was
determined.
Sinceitwasnecessarytomeasurethethreedifferentcomponentsofthewakevelocities,the
readings had to be repeated for different orientations of the hotfilm probes. Six different
hotfilmorientationswererequiredtobeabletoderivetheflowvelocitycomponentsusing
anewmethoddevelopedbyHaans[37].
Measurement planes
Figure4.9Hotfilmmeasurementsatdifferentplanesparalleltotherotorplane.
52
Chapter4 AerodynamicAnalysisoftheTUDelftModelTurbine
DataReduction
Thissectiondescribestheproceduresadoptedtocalibratethehotfilmsanddeducethe3D
componentsofthemeasuredvelocitiesinthenearwakeoftherotor.
HotfilmCoordinateSystemDefinitions
Inordertobeabletomeasure3Dflowcomponentsusinghotfilms,asuitablesystemofco
ordinatesshouldbedefined.Fig.4.10showsthelocalcoordinatesystems(xpypzp)usedfor
thenormalandparallelprobes.Inbothcases,thecoordinateaxesareattachedtothehot
filmandnottotheprobe.Theypaxisisalwaysalignedwiththehotfilm.
yp
zp
yp
xp
xp
zp
Fig.(a):Normalprobe Fig.(b):Parallelprobe
Figure4.10LocalCartesiansystemofcoordinatesforthenormalandparallelprobecalibration.
The calibration procedure of each hotfilm consisted of two steps: (a) a speed calibration
and(b)anangularcalibration.
(a)SpeedCalibrationthehotfilmvoltage,E,wascorrelatedwithaknownwindspeedofthe
tunnel jet (Ujet) with the hotfilm in a normal position to the flow (with the ypzp plane
alignedwiththeflow).Forthecorrelation,thetemperaturecorrected,averagedKingslaw
wasused[14]givenby:
E
2
(
= T f Ta ) ( A + BU jet
n
) (4.1)
whereTfandTaarethepresethotfilmandmeasuredflowtemperaturerespectively.A,B
and n are calibration constants with A set equal to E
2
(T f ) U
Ta
jet = 0
. B and n were
derivedfrommeasuringEatdifferentwindtunnelspeedsandthenapplyingacurvefitting
procedureusingthemethodofleastsquares.Forthisspeedcalibration,Ujetwasdetermined
usingasinglepitotstatictubelocatedinthevicinityofthehotfilm.
53
Chapter4 AerodynamicAnalysisoftheTUDelftModelTurbine
(b)AngularCalibrationthiscalibrationprocedurewasrequiredtoderivethecharacteristics
ofthehotfilmsatdifferentflowdirections.Thebasisforthisprocedureisbasedonthefact
that given that Veff is the velocity measured when correlating with the hotfilm voltage E
whenthelatterisnormaltotheflowasforthespeedcalibration(Eqt.4.1),Veffiscorrelated
tothe3Dvelocitycomponentsinaccordancewith[14]:
Veff
2
= h 2U p 2 + k 2V p 2 + W p 2 (4.2)
Up,VpandWparethe3Dflowvelocitycomponentsinthedirectionofthehotfilmaxisof
Fig.4.10(xp,yp,zp),respectivelywhilehandkaretheangularcalibrationconstantsforthe
particular hotfilm. To determine the values for h and k, each hotfilm was subjected to a
constantaxialvelocitywhilstbeingrotatedinaverticalplaneparalleltotheaxialdirection.
Indoingso,thehotfilmwasorientedindifferentorientationssuchthatUp,VporWpwas
zero. Two orientations where used both for the normal and parallel probes. These
orientationsareshowninFig.4.11.Eqt.4.2wasappliedtoeachdifferentorientationtoyield
the equations given in Fig. 4.11. During the angular calibration, Veff was recorded for the
range900<<900,withincrementsof100.Foreach,10,000samplereadingsweretakenand
theaveragedvalueswerefound.Thecorrespondingstandarddeviationswerefoundtobe
very small (on the order of 0.001% of the averaged values) and therefore their influence
could be neglected. To determine h or k for each probe orientation, a trialanderror
algorithm was used. The method was based on using the experimental values for V and
assumingdifferentvaluesofh(ork)toestimateVeffinaccordancewiththecorresponding
equation from Fig. 4.11. For each , the error between the estimated value of Veff and the
experimental value was found. The assumed value of h (or k) that yielded the minimum
errorwastakentobethecorrectrequiredvalue.Thisminimumerrorwasfoundtobeless
than5%.Theangularcalibrationprocesswasrepeatedfrequentlyinordertominimisethe
uncertaintyduetohotfilmageing.Thevaluesofhandkwerefoundtobeontheorderof
1.1 and 0.25 respectively for both probes. It was found that the variation of these angular
calibration constants only varied minimally with the tunnel wind speed. This simplified
considerably the data reduction process. Fig. 4.12 illustrates typical characteristic curves
derived for the hotfilms using the experimental measurements taken during the angular
calibrationprocedures.ThecurvesfortheestimatedvaluesforVeff usingthederivedvalues
ofhandkarealsoshown.Notethatfororientation1oftheparallelprobe,thecharacteristic
curve is asymmetric since for negative the prong of the probe will disturb the flow
approaching the hotfilm. Consequently for negative , Eqt. (3) of Fig. 4.11 is invalid for
<00.
54
55
Chapter4 AerodynamicAnalysisoftheTUDelftModelTurbine
6 7.2
56
Experiment
7.1
5 Estimated using h = 1.1125
7
V = 6.295 m/s
6.9
4
6.8
3 6.7
Veff (m/s)
Veff (m/s)
6.6
2
V = 5.595 m/s 6.5
6.4
1 Experiment
0 6.2
-100 -80 -60 -40 -20 0 20 40 60 80 100 -100 -80 -60 -40 -20 0 20 40 60 80 100
(deg) (deg)
Fig.(a):Normalprobe,Orientation1 Fig.(b):Normalprobe,Orientation2
6 8
7
5
4
5
3 4
Veff (m/s)
Veff (m/s)
V = 5.214 m/s 3
2
V = 6.165 m/s
2
Experiment Experiment
1
Estimated using k = 0.2978 1
Estimated using k = 0.2902 &
h = 1.0554
0 0
-100 -80 -60 -40 -20 0 20 40 60 80 100 -100 -80 -60 -40 -20 0 20 40 60 80 100
(deg) (deg)
Fig.(c)Parallelprobe,Orientation1 Fig.(d)Parallelprobe,Orientation2
Fig.4.12 Typicalcharacteristiccurvesforthehotfilmprobesasderivedfromtheangularcalibrationprocedures
Chapter4 AerodynamicAnalysisoftheTUDelftModelTurbine
Chapter4 AerodynamicAnalysisoftheTUDelftModelTurbine
Derivationof3Dnearwakevelocitiesfromhotfilmmeasurements
This section describes the technique adopted to derive the 3D near wake velocity
components from the hotfilm effective velocities (Veff) measured with different hotfilm
orientationsatthedifferentpointsintherotornearwakeasdiscussedinsection4.2.2.
Therequired3Dwakevelocitycomponentsatagivenpointweretheaxial,horizontaland
verticalcomponentsdenotedbywh,waandwv,suchthattheywerealignedwiththeglobal
coordinate axes XaYaZa, respectively. Two different methods were adopted to deduce
these components: a traditional method and a new method being proposed by Wouter
Haans[37].
(a) Traditional Method: With this method, three hotfilm orientations were required. Two
weretakenwiththenormalprobe,whilethethirdwastakenwiththeparallelprobe.Inone
ofthenormalprobeorientations,thehotfilmwasalignedwiththeXaaxisoftherotorwhile
inthesecondnormalprobeorientation,thehotfilmwasalignedwiththeZa.Intheparallel
probeorientation,thehotfilmwasorientedwiththeYaaxis.Eqt.4.2wasappliedforeach
orientationtoyieldthefollowingthreeequations:
Veff , X
a
2
= k 2 wh 2 + h 2 wv 2 + wa 2
Veff ,Y
a
2
= h 2 wh 2 + wv 2 + k 2 wa 2 (4.3)
Veff , Z
a
2
= h 2 wh 2 + k 2 wv 2 + wa 2
In the equations above, Veff,Xa, Veff,Ya, Veff,Za are the averaged hotfilm effective velocities
measured during the experimental procedure. It should be emphasised that Eqt. 4.2 only
holds for an instantaneous point in time. Since it was impossible to measure the three
different effective velocities simultaneously, the above system of equations could strictly
speakingbeonlyappliedontheassumptionthatthewakevelocitiesforagivenpointand
rotorazimuthangleareconstantanddonotvarywithtime.Thisassumptionisonlyvalid
whenturbulencelevelsinthewakewassmall.Sincetherotorwasoperatingsuchthatthe
flowwasattached,turbulencelevelsweresmall,evenbehindtherotorplane.Therequired
3D velocity components were solved by reorganizing the above equations in matrix form
andapplyingmatrixinversionasfollows:
1
wh 2 k 2 h2 1 Veff , Xa 2
2 2
= h
wv 1 k 2 Veff ,Ya 2 (4.4)
wa 2 h 2 k2 1 Veff , Za 2
57
Chapter4 AerodynamicAnalysisoftheTUDelftModelTurbine
Amaindisadvantageofthismethodregardsthefactthatthedirectionsoftheflow
velocitiesremainunknown.Componentwawasfoundtobesignificantlylargerthanthe
othercomponents,anditcouldbeeasilyassumedthatitactsinthedownstreamdirection.
(b) New Method: Wouter Haans developed a more advanced approach that is capable of
finding the directions of the wh and wv. The method makes use of six different probe
orientationsoftheasymmetricresponseoftheparallelprobetotheflow angleduetothe
obstructionoftheflowresultingfromtheprobesprongs(seeFig.4.12(c)).Inthismethod,
thetraditionalapproachwas stillusedtofindtheaxialflowcomponentwa.Detailsofthe
methodaregivenin[37].
Thefinalstepindatareductionprocesswastoobtaintheflowcomponentsinlocalrotorco
ordinates(xyzreferenceframe)byusingthefollowingmatrixtransformation:
wr sin( ) cos( ) 0 wh
wt = cos( ) sin( ) 0 wv (4.5)
wa 0 0 1 wa
Thelocalinflowmeasuredvelocityatanypointinaparticularmeasuringplanedependson
thegeometricallocationofthepointandtherotorazimuthangle.Theaxial,tangentialand
(
wr , r , p
R ) .
Forthesakeofthedataprocessingandcalculationusingthedevelopedsoftwarecodes,a
threeletter index notation was used to denote a velocity component at each point. For
instance the axial flow velocity was represented by ( wa ) ,i , j where , i and j are indices
denotingtherotorazimuthangle(),theradiallocation(r/R)andtheazimuthangleofthe
measuringlocation(p).
58
Chapter4 AerodynamicAnalysisoftheTUDelftModelTurbine
InflowResults
DerivedWakeVelocities
Since the measurements were carried out at different yaw angles and different
measurementplanes,theresultingdatabasewasquiteanextensiveonecontainingatleast
24MB of data. Figs. 4.13, 4.14 and 4.15 plot typical measured signals for the flow velocity
components(wa,wtandwr)at=00obtainedatvariouspoints(r/R, p)asafunctionofrotor
azimuthangle().
Tobeabletointerpretthevelocitysignals,itishelpfultounderstandthedistinctaxialand
tangentialvelocitypatternsmeasuredwheneachbladepassesbythehotfilmprobe.Such
patternsareillustratedinFig.4.16andaremainlyinducedbytheboundcirculationofthe
blades.Theaxialvelocitypattern(seeninFig.4.16(a))ischaracterizedbyanincreaseinthe
flow velocity followed by a rapid decrease. The peak to peak velocity difference (wa,max
wa,min) will decrease as the distance of the probe from the rotorplane (i.e. Ya) is increased.
This may be observed in Figs. 4.13(a), 4.14(a) and 4.15(a). Since the rotor has two blades,
then two blade passage signals can be recognized over one whole rotor revolution. The
tangentialvelocitypatternischaracterizedbyUshapedpattern(seeninFig.4.16(b)).The
tangentialvelocityisnormallynegativei.e.oppositetothedirectionoftherotatingblade.
Thisagreeswellwiththelawofconservationofmomentum.Thenegativepeaktangential
velocity decreases as the probe is moved away from the rotorplane, as noted in Figs. 4.13
(b),4.14(b)and4.15(b).
59
Chapter4 AerodynamicAnalysisoftheTUDelftModelTurbine
6.5
6.0
5.5
5.0
4.5
wa (m/s)
4.0
Ya = 6.0cm
3.5 UpStream
3.0 Ya = 3.5cm
DwnStream
2.5 Ya = 6cm
DwnStream
2.0
Fig. (a) Ya = 9cm
DwnStream
1.5
0 60 120 180 240 300 360
(deg)
2.0
1.0
0.0
wt (m/s)
-1.0
Ya = 3.5cm
DwnStream
Ya = 6cm
-2.0
DwnStream
Fig. (b)
Ya = 9cm
DwnStream
-3.0
(deg)
3.0
2.0
1.0
wr (m/s)
0.0
0 60 120 180 240 300 360
Ya = 3.5cm
DwnStream
-1.0 Ya = 6cm
DwnStream
Fig. (c)
Ya = 9cm
DwnStream
-2.0
(deg)
Figure 4.13 Axial, tangential and radial velocities derived from the hotfilm measurements for
=00.Theprobeislocatedat(r/R,p)=(0.4,750).Bladepassageisobservedat=750and2550.
60
Chapter4 AerodynamicAnalysisoftheTUDelftModelTurbine
6.5
6.0
5.5
5.0
4.5
4.0
wa (m/s)
3.5
3.0
Ya = 6.0cm
2.5 UpStream
Ya = 3.5cm
2.0
DwnStream
1.5
Ya = 6cm
1.0 DwnStream
2.0
1.0
0.0
0 60 120 180 240 300 360
wt (m/s)
-1.0
-2.0
Ya = 3.5cm
DwnStream
2.0
1.0
wr (m/s)
0.0
0 60 120 180 240 300 360
Ya = 3.5cm
DwnStream
Fig. (c)
-1.0 Ya = 6cm
DwnStream
Ya = 9cm
DwnStream
-2.0
(deg)
Figure 4.14 Axial, tangential and radial velocities derived from the hotfilm measurements for
=00.Theprobeislocatedat(r/R,p)=(0.7,750).Bladepassageisobservedat=750and2550.
61
Chapter4 AerodynamicAnalysisoftheTUDelftModelTurbine
6.5
6.0
5.5
5.0
4.5
4.0
wa (m/s)
3.5
3.0
Ya = 6.0cm
2.5 UpStream
2.0 Ya = 3.5cm
DwnStream
1.5
Ya = 6cm
1.0 DwnStream
Fig. (a)
0.5 Ya = 9cm
DwnStream
0.0
0 60 120 180 240 300 360
(deg)
2.0
1.0
0.0
wt (m/s)
-1.0
Ya = 3.5cm
DwnStream
-2.0 Ya = 6cm
Fig. (b) DwnStream
Ya = 9cm
DwnStream
-3.0
(deg)
3.0
2.0
wr (m/s)
1.0
Ya = 3.5cm
DwnStream
0.0
Ya = 6cm
0 60 120 180 240 300 360
DwnStream
Fig. (c)
Ya = 9cm
DwnStream
-1.0
(deg)
Figure 4.15 Axial, tangential and radial velocities derived from the hotfilm measurements for
=00.Theprobeislocatedat(r/R,p)=(0.9,750).Bladepassageisobservedat=750and2550.
62
Chapter4 AerodynamicAnalysisoftheTUDelftModelTurbine
Axial direction
Direction of blade motion
wa
Axial flow
wa,min
wt
Fig. (b)
Tangential
flow
Figure4.16Distinctaxialandtangentialflowvelocitypatternsinducedbybladepassage.
Inthewtsignals,someturbulencemayalsobeobservedat positionsotherthanthoseat
the blade passage positions. This turbulence may also be observed in the wa signals and
resultsfromtheeffectofthewakevortexsheetpassingbythehotfilmprobe.
Fig. 4.17 shows typical axial velocity signals at different yaw angles for a given probe
location. It is noted that the flow velocity at this location decreases as the yaw angle is
increased.Foramoredetailedphysicalexplanationoftheseinflowmeasurementsreferto
theworkofHaans[37],Vermeer[98,99]andMast[55].
63
Chapter4 AerodynamicAnalysisoftheTUDelftModelTurbine
6.5
6.0
5.5
5.0
4.5
4.0
wa (m/s)
3.5
3.0
2.5
2.0
1.5 Yaw 0 deg
Determinationofaxialflowvelocityatthebladepassagelocation(wa,c)
From the experimental axial flow velocity signals as shown in Figs. 4.13(a), 4.14(a) and
4.15(a),itwaspossibletoestimatetheaxialflowvelocitiesatthebladepassagelocationsfor
each of the probe measuring locations (i.e. at r = 40, 50, 60, 70, 80, 90%R at azimuthal
incrementsof150).Thesewereobtainedforeachofthefourmeasuringplanes(i.e.atYa=
6cm upstream and 3.5cm, 6cm and 9cm downstream) and were required to be able to
estimate the flow velocities at the blades from which the angle of attack could then be
found.Theaxialflowvelocityatthebladepassagelocation(denotedbywa,c)wasestimated
bytakingittobeequaltotheaverageofthemaximumandminimumvelocities.Referring
toFig.4.16(a),thisisequalto
wa ,max + wa ,min
wa ,c = (4.6)
2
In the index representation, the value of wa,c at a given location (r/R,p) at a particular
measurementplaneisdenotedas(wa,c)i,j.Recallthatindexidenotestheradiallocation(40,
50, 60, 70, 80, 90 or 100%R) of the hotfilm measurement point. Index j denotes the
azimuthallocationofthispoint(00,150,300,.3450).
The results obtained for wa,c at =00 are given in Fig. 4.18(a). The mean values of wa,c
obtainedfromthedifferentprobepositions(atazimuthalincrementsof150)ateachradius
areshowninthisplot.Thecorrespondingstandarddeviationsarealsopresented.Foraxial
64
Chapter4 AerodynamicAnalysisoftheTUDelftModelTurbine
conditions(=00),wa,cshouldideallybeconstantwith p.Butthiswasnotthecasedueto
various sources of error which will be discussed later. For instance Fig. 4.18(b) plots the
variationofwa,cwiththebladeazimuthangleoveronewholerotorrevolutionfor=00.Itis
notedthatthevariationisnotconstantwith.Amajorcontributionforthisnonidealtrend
isthenonuniformityinthetunnelexitjet,asdepictedearlierinFig.4.5.Thereisavelocity
dipforpositionsnear =00whichinfactresultsinadiscontinuousbehaviourinthevelocity
variation between 33003600 and 00300. This discontinuity may be easily observed in Fig.
4.18(b).
Linear interpolation was employed to estimate the spanwise variations of wa,c at the
rotorplane(Ya=0cm)usingthemeanvaluesforthe6cmupstreamand3.5cmdownstream
planes:
(w ) ( wa ,c )
( wa,c ) = ( wa ,c )
a ,c
Ya = 3.5 cm Ya =6 cm
+ ( 0 ( 6) ) (4.7)
Ya = 0 cm Ya =6 cm 3.5 ( 6)
TheseinterpolatedvaluesareincludedinFig.4.18(a).Thesevaluesareanestimateforthe
axial flow velocities at the blade lifting lines and they were used to estimate the angle of
attackfromthemeasurements,aswillbedescribedlateroninsection4.3.
Figs.4.19and4.20presentthevaluesofwa,cobtainedinyawedconditions(=300and450).
Since in yawed conditions, the flow velocities at the blades are unsteady, the results are
plotted as a function of the blade azimuth angle (). For yawed conditions only, the
variations of wa,c versus were also smoothened to damp out the jerky variations
introducedduetouncertaintiesinthemeasurements.Thesmoothedvariationsareincluded
inFigs.4.19and4.20.Inthedatasmoothing,atechniquethatimplementsaGaussiankernel
was used to compute local weighted averages of the input vector wa,c. In the index
representationusing(i,j),thesmoothedvalueofwa,catagivenpoint(i,j)wasfoundfrom
jtot ( p ) j ( p )k
K ( wa ,c )i , k
(w )
k =1
= (4.8)
( ) ( p )k
( smoothed )
p j
a ,c i, j jtot
K
j =1
x
2
1 2( 0.37 )2
K ( x) = e
2 0.37
isthebandwidththatshouldbeprescribed.Inthisanalysis, wassetequalto0.8.This
wasfoundtobereasonable.Thissmoothingmethodwasusefulsincethedataliesalonga
bandofconstantwidth(equalto150).
65
Chapter4 AerodynamicAnalysisoftheTUDelftModelTurbine
5.5
5.0
4.5
4.0
3.5cm DwnStream
3.0
6cm DwnStream
2.5
9cm DwnStream
2.0
RotorPlane (Interp)
4.5
3.5
3
wa,c (m/s)
2.5
2
r/R=0.4
1.5 r/R=0.5
r/R=0.6
1
r/R=0.7
0.5 r/R=0.8
r/R=0.9
0
0 60 120 180 240 300 360
(deg)
Figure 4.18(b) Variation of the axial flow velocity at the blades with azimuth angle for=00 at
3.5cmdownstreamoftherotorplane.
66
Chapter4 AerodynamicAnalysisoftheTUDelftModelTurbine
Usingthesesmootheddistributions,linearinterpolationwasusedtoestimatethevariations
ofwa,cwith attherotorplanefordifferentradiallocationsusingEqt.4.7.Theinterpolated
valueswereanestimatefortheunsteadyaxialflowvelocitiesattheliftinglineoftheblades
in yawed conditions. The interpolated values are included in Figs. 4.19 and 4.20. In these
figures,thediscontinuousbehaviorinthevelocityvariationbetween33003600and00300is
observed,aswasalreadynotedat=00(seeFig.4.18(b)).Thisdiscontinuityismainlydueto
thenonuniformityinthetunnelexitjet,(seeFig.4.5).Yetitisnotexcludedthatthewake
circulation from the skewed wake (tip and in particular root circulation) resulting from
rotor yaw as well as the influences of the centrebody structure of the testrig could have
yielded abrupt changes of wa,c with . These abrupt changes could also contribute to the
discontinuousbehaviorinthevelocityvariationbetween33003600and00300.
Sourcesoferror
Themainsourcesoferrorintheinflowmeasurementsarethefollowing:
(1) Thenonuniformityintunnelexitjet:asalreadyexplainedearlierinsection4.2.1,
Fig.4.5,itwasfoundthatthetunnelexitvelocityvariesfrom5.2to5.8m/swhen
the pitot readings are set to 5.5m/s. Thus it is reasonable to assume that the
uncertainty interval introduced by the nonuniformity in the tunnel exit jet is
5.50.3m/s.
(2) Thehotfilmprobesinmeasuringtheeffectivevelocities(Veff)mainlyduetoerrors
inthespeedcalibrationconstants.
(3) Errorsinthedatareductiontechniquedescribedearlier(seepage54)mainlydueto
errors in the angular calibration constants. Also, the data reduction technique
assumesthatforagivenrotorazimuthposition,thevelocityatanypointdoesnot
change with time and thus the different hotfilm measurements using different
probeorientationsneednotbecarriedoutsimultaneously.Thepresenceofsome
fluctuations(suchasduetoturbulence)invalidatesthisassumption.
(4) The traversing system experienced some inaccuracies in positioning the hotfilm
probe at the required locations. The positioning accuracy was estimated to be
around1cm.
The rotor angular speed and the pitot readings at the tunnel exit could be adjusted very
accuratelyattherequiredsetting(720rpmand5.5m/srespectively).Thereforetheerrorsin
the experimental data due to possible fluctuations in these parameters were negligible.
Haans[37]hascarriedoutanindepthuncertaintyanalysisforthesemeasurementsat =00
estimatingtheuncertaintytoduetoindividualsourcesoferror.Inthisanalysis,itwas
67
Chapter4 AerodynamicAnalysisoftheTUDelftModelTurbine
concludedthattheoveralluncertaintyinwaatthevicinityofbladepassageisontheorder
of10%.
Alternatively, it was possible to use a simple method to derive an estimate for this
uncertaintybyconsideringtheresultsofwa,cobtainedat =00.Asalreadydescribedinthe
previous section, in the ideal situation, wa,c should be independent of at =00. For each
radial location and measurement plane, wa,c was obtained at azimuth increments of 150
yielding a sample size of 24 readings. The mean and standard deviations were computed
and the results were plotted earlier in Fig. 4.18(a). Assuming the velocity distribution of
each sample follows a normal distribution, the uncertainty interval at 95.45% confidence
could be taken as 2 where is the standard deviation. This uncertainty interval was
computed over radial locations 40, 50, 60, 70, 80 and 90%R and the four measurement
planesanditwasfoundtobeintherangeof610%ofthemeanvaluesofwa,c.Thisisof
thesameorderasthatobtainedbyHaans[37].
The above simple method for estimating the uncertainty in the hotfilm measurements
could not be applied for yawed conditions for the simple reason that wa,c is no longer
independenton .However,itwasjustifiedtoassumethesameuncertaintylevelof610%
for =300 and 450 as well. This is because of the fact that the same apparatus and rotor
speedandbladepitchsettingswereusedandalsobecausethehotfilmspeedandangular
calibration constants were very close to those for axial conditions. An important point to
noteregardsthefactthattheuncertaintyinUduetothenonuniformityinthetunnelexit
jet (0.3m/s) is considerably high (on the order of 5% of 5.5m/s) and should contribute a
considerableproportionofthe610%uncertaintyinwa,c.Yet,whentherotorisyawed,the
magnitude of the axial flow velocity component at the measurements planes is less
dependent on the windspeed U than at no rotor yaw. It also follows that in yaw, the
uncertainty contribution due to the tunnel exitjet nonuniformity should be less. It is
possiblethatthiscouldresultinaloweroveralluncertaintylevelinyawthanfornoyaw.
Butamoreelaborateuncertaintyanalysiswouldbenecessarytoconfirmthis.
ComparisonwithpreviousdatabyVermeer
Inyear1998,Vermeer[96]carriedoutsimilarinflowmeasurementsusingthesamewind
tunnel and rotor with hotwire probes (see also Schepers [69]). The measurements were
performedforthesamewindspeedandtipspeedratioandat =00,300,450and600.The
measurementswerecarriedoutatradiallocations50,60,70and80%Randwerelimitedto
one measurement plane only (at 6cm downstream of the rotorplane). Using this data, the
distributionsofwa,ccouldbeobtainedforthisplaneusingtheproceduredescribedonpage
64. These distributions were compared with those from the new measurements. The
comparisonat =300isshowninFig.4.21.Theagreementisreasonablygoodandthisadds
to confidence in the measurements. The maximum discrepancy is seen at blade azimuth
angles00<<900.Thesamelevelofagreementwasalsoobtainedfor=00and450.
68
Chapter4 AerodynamicAnalysisoftheTUDelftModelTurbine
4.5
4.0
3.5
6cm UpStream
3.5cm DwnStream
3.0
3.5cm DwnStream_Smooth
6cm_DwnStream
6cm DwnStream_Smooth
2.5
9cm DwnStream
9cm DwnStream_Smooth
RotorPlane (Interp)
2.0
0 60 120 180 240 300 360
( deg )
Fig.(a)r/R=0.4
4.5
4.0
3.5
6cm UpStream
wa,c (m/s)
6cm UpStream_Smooth
3.5cm DwnStream
3.0
3.5cm DwnStream_Smooth
6cm_DwnStream
6cm DwnStream_Smooth
2.5
9cm DwnStream
9cm DwnStream_Smooth
RotorPlane (Interp)
2.0
0 60 120 180 240 300 360
( deg )
Fig.(b)r/R=0.5
Fig. 4.19 Variation of axial flow velocity at blades with blade azimuth angle at different radial
locationsat=300.
69
Chapter4 AerodynamicAnalysisoftheTUDelftModelTurbine
4.5
4.0
3.5cm DwnStream
3.0
3.5cm DwnStream_Smooth
6cm_DwnStream
6cm DwnStream_Smooth
2.5
9cm DwnStream
9cm DwnStream_Smooth
RotorPlane (Interp)
2.0
0 60 120 180 240 300 360
( deg )
Fig.(c)r/R=0.6
4.5
4.0
3.5
6cm UpStream
wa,c (m/s)
6cm UpStream_Smooth
3.5cm DwnStream
3.0
3.5cm DwnStream_Smooth
6cm_DwnStream
6cm DwnStream_Smooth
2.5
9cm DwnStream
9cm DwnStream_Smooth
RotorPlane (Interp)
2.0
0 60 120 180 240 300 360
( deg )
Fig.(d)r/R=0.7
Fig.4.19contd.frompreviouspage(=300).
70
Chapter4 AerodynamicAnalysisoftheTUDelftModelTurbine
4.5
4.0
3.5
6cm UpStream
3.5cm DwnStream
3.0
3.5cm DwnStream_Smooth
6cm_DwnStream
6cm DwnStream_Smooth
2.5
9cm DwnStream
9cm DwnStream_Smooth
RotorPlane (Interp)
2.0
0 60 120 180 240 300 360
( deg )
Fig.(e)r/R=0.8
4.5
4.0
3.5
6cm UpStream
wa,c (m/s)
6cm UpStream_Smooth
3.5cm DwnStream
3.0
3.5cm DwnStream_Smooth
6cm_DwnStream
6cm DwnStream_Smooth
2.5
9cm DwnStream
9cm DwnStream_Smooth
RotorPlane (Interp)
2.0
0 60 120 180 240 300 360
( deg )
Fig.(f)r/R=0.9
Fig.4.19contd.frompreviouspage(=300).
71
Chapter4 AerodynamicAnalysisoftheTUDelftModelTurbine
4.5
4.0
3.5
3.5cm DwnStream
2.5
3.5cm DwnStream_Smooth
6cm_DwnStream
2.0
6cm DwnStream_Smooth
9cm DwnStream
1.5
9cm DwnStream_Smooth
RotorPlane (Interp)
1.0
0 60 120 180 240 300 360
( deg )
Fig.(a)r/R=0.4
4.0
3.5
3.0
6cm UpStream
wa,c (m/s)
3.5cm DwnStream
3.5cm DwnStream_Smooth
2.0
6cm_DwnStream
6cm DwnStream_Smooth
9cm DwnStream_Smooth
RotorPlane (Interp)
1.0
0 60 120 180 240 300 360
( deg )
Fig.(b)r/R=0.5
Fig. 4.20 Variation of axial flow velocity at blades with blade azimuth angle at different radial
locationsat=450.
72
Chapter4 AerodynamicAnalysisoftheTUDelftModelTurbine
4.0
3.5
3.0
6cm UpStream
3.5cm DwnStream
3.5cm DwnStream_Smooth
2.0
6cm_DwnStream
6cm DwnStream_Smooth
9cm DwnStream_Smooth
RotorPlane (Interp)
1.0
0 60 120 180 240 300 360
( deg )
Fig.(c)r/R=0.6
4.0
3.5
3.0
6cm UpStream
wa,c (m/s)
3.5cm DwnStream
3.5cm DwnStream_Smooth
2.0
6cm_DwnStream
6cm DwnStream_Smooth
9cm DwnStream_Smooth
RotorPlane (Interp)
1.0
0 60 120 180 240 300 360
( deg )
Fig.(d)r/R=0.7
Fig.4.20contd.frompreviouspage(=450).
73
Chapter4 AerodynamicAnalysisoftheTUDelftModelTurbine
4.0
3.5
3.0
6cm UpStream
3.5cm DwnStream
3.5cm DwnStream_Smooth
2.0
6cm_DwnStream
6cm DwnStream_Smooth
9cm DwnStream_Smooth
RotorPlane (Interp)
1.0
0 60 120 180 240 300 360
( deg )
Fig.(e)r/R=0.8
4.5
4.0
3.5
6cm UpStream_Smooth
3.5cm DwnStream
2.5
3.5cm DwnStream_Smooth
6cm_DwnStream
2.0
6cm DwnStream_Smooth
9cm DwnStream
1.5
9cm DwnStream_Smooth
RotorPlane (Interp)
1.0
0 60 120 180 240 300 360
( deg )
Fig.(f)r/R=0.9
Fig.4.20contd.frompreviouspage(=450).
74
Chapter4 AerodynamicAnalysisoftheTUDelftModelTurbine
4.2 3.8
4 3.6
3.8
3.6 3.4
3.4 3.2
wa,c (m/s)
wa,c (m/s)
3.2 3
3 2.8
2.8
2.6
2.6 Vermeer
Vermeer
present data 2.4 present data
2.4
2.2
2.2
2 2
0 60 120 180 240 300 360 0 60 120 180 240 300 360
(deg) (deg)
Fig.(a)r/R=0.5 Fig.(c)r/R=0.7
3.8 3.8
3.6 3.6
3.4 3.4
3.2 3.2
3
wa,c (m/s)
wa,c (m/s)
3
2.8 2.8
2.2 2.2
2 2
0 60 120 180 240 300 360 0 60 120 180 240 300 360
(deg) (deg)
Fig.(b)r/R=0.6 Fig.(d)r/R=0.8
Figure4.21Comparisonofdistributionsofwa,cwith withthosefromVermeer[96]at =300and
6cmdownstreamoftherotorplane(Ya/R=0.1),seealsoSchepers[69].
75
Chapter4 AerodynamicAnalysisoftheTUDelftModelTurbine
4.2.3PartII:SmokeVisualizationMeasurements
Thefollowingsectionsdescribethewindtunnelsmokevisualizationexperimentstotrack
thetipvortexpathsofthewakeinaxialandyawedconditions.
ExperimentalSetup
Fig.4.22illustratestheschematicdiagramoftheapparatususedinthesmokevisualization
experiments.Therotorpositionwithrespecttothetunnelexitwaskeptthesameasforthe
wakeinflowmeasurements(i.e.withtherotorhub1mdownstreamfromthetunnelexit).
A smoke jet was created using a generator that uses oil to produce smoke. The jet was
injectedintothetunneljetstreamthroughanozzlethatwaslocatedwellupstreamfromthe
rotor.Thenozzlepositionwithrespecttothebladetip(at900or2700azimuthpositions)had
tobeadjustedatdifferentyawanglesoftherotorforoptimalvisualizationofthetipvortex.
Duringtheexperimentsthewindtunnelhalllightswhereswitchedofftocreateacomplete
dark environment. A stroboscope was synchronized with the rotor to flash when the
azimuthangle ofthefirstbladewas900.Inthiswayitwaspossibletocapturethesmoke
flowpatternsinahorizontalplanepassingthroughtherotorhubandobservetheposition
ofthewaketipvorticesrelativetothebladetip.Adigitalcamera,withitslensfocusedon
this plane, was used to record multiple images of these smoke flow patterns. A reference
gridconsistingofequallyspacedwireswasconstructed.Thiswasinstalledinahorizontal
planeontopoftherotorsuchthatthewireswereparallelandperpendiculartothewind
tunnelaxis(refertoFig.4.22).
Sideview Topview
7
6
6
2 1 2 1
7 5 3
3
4 4
1: Model rotor 2: Wind tunnel 3: Video camera 4: Smoke generator 5: Smoke jet 6: Wire grid 7: Stroboscope
Figure4.22Schematicdiagramofexperimentalsetupusedforsmokevisualization.
76
Chapter4 AerodynamicAnalysisoftheTUDelftModelTurbine
Experimentalprocedure
Throughouttheexperiments,thewindtunnelexitjetvelocityUjetwasmaintainedconstant
at 5.5m/s. The smoke visualizations were performed at different yaw angles, tip speed
ratiosandbladetippitchangles:Fivedifferentrotoryawangles:00,150,300,450and+450.
For each of these yaw angles, the tip vortex core positions were recorded at nine
combinationsofthefollowingtipspeedratiosandtippitchangles: = 6,8 and10and tip=
00,20 and40.Thesesettingswereselectedsothattherotorwouldoperateinbothattached
and stalled flow conditions over the blades. The smoke was injected on both sides, at
azimuthanglesequalto900and2700(refertoFig.4.23).Theverticaldistancesbetweenthe
camera,rotorandgridweremeasuredforeachrotorsetting.Thesedistanceswererequired
fortheparallaxcorrectionwhichwillbedescribedlateron.
The measurement campaign employed ensured repeatability and randomization of data.
Foreachrotorsetting,twosmokeinjectionswerecarriedoutoneveryside,eachyieldingat
least 75 photos. The tip vortex measurements for = 8, tip = 00 and = 6, tip = 20 were
repeated.Furthermore,twosymmetrycheckswereaccomplished:
(1)at=00,werethetipvortexcorelocationsat900and2700werecompared;
(2)at =450and+450,werethetipvortexcorelocationsat900and2700werecompared
forthecorrespondingupwindanddownwindsides.
Tipvortexpitchp
ys (DownwindSide)
Tipvortexpath
Rotorbladetip
Tipvortexcore
(DownwindSide)
xs
Axisofskewed
wake
O
s
Yawaxis
Rotoraxis
Rotorbladetip
(UpwindSide) Linesjoiningvortexcores
havingsameage
xs
Tipvortexpitchp Tipvortexpath
ys (UpwindSide)
Figure4.23Definitionsusedinthewakegeometry.
77
Chapter4 AerodynamicAnalysisoftheTUDelftModelTurbine
Inaseparatemeasurementcampaign,theaxialthrustforceontherotorwasmeasuredfor
yawanglesgoingfromchangethis=450and+450withstepsof150.Foreachyawangle,
the axial thrust was measured at each of the nine combinations of and tip. These
measurementsaredescribedinfurtherdetailinreference[36].
DataProcessing
ThePhotos
Figure4.24displaystypicalphotosderivedfromthevisualisations.Duetothestroboscopic
lightsourceandthecameraorientation,thephotosappeartobeinstantaneouspicturesof
theunsteadyhelicaltipvortexstructureinahorizontalplanethatcrosssectstherotorhub.
Thetipvorticesareclearlyidentifiablebytheswirlingsmokepattern,withthecentreofthe
vortexindicatedbythecentreofthesmokefreearea.Thebladetipcanbeidentifiedonthe
photos,togetherwiththegridinthebackground.
The presence of wake expansion is evident from the photos. For yawed conditions, wake
expansionislargeronthedownwindsidethanontheupwindsideandthisleadstowake
skew.Usingphotoeditingsoftware,thedistancesofeachvortexcorerelativetotheblade
tipsweremeasuredagainstthelengthofgridssquares.Knowingthesizeofthesesquares,
0.1mx0.1m,thepositionsofthetipvortexcorescouldbeestablishedafterapplicationof
cameraparallaxcorrectiontothephotomeasurements.Figure4.25isaschematicusedfor
parallaxcorrection.Here,handHaretheperpendiculardistancesbetweengridandsmoke
visualisationplaneandbetweengridandcamera,respectively.Listhemeasurementtaken
from the photo while d is the corrected (actual) distance. Applying similar triangles,
distancedisgivenby
H h
d = L (4.9)
H
Other factors were thought to influence parallax effects, such as lens and camera sensor
misalignment.HoweverthesewereconsideredtobenegligiblesincethevaluesofhandH
werelarge(about1.28mand2.80m,respectively)andthereforetheireffectwasignored.
Usingtheabovemethod,thelocationofeachvortexcentrerelativetothebladetipscould
be found. The position of each vortex core was expressed using local coordinates (xs, ys).
These coordinates are attached to the blade tip and lie in the horizontal plane passing
through the rotor hub centre (see Fig. 4.23). The xsaxis lies along the undisturbed flow
direction,whiletheysaxisisperpendiculartotheundisturbedflow.
78
Chapter4 AerodynamicAnalysisoftheTUDelftModelTurbine
Bladetip
Tipvortexcore
Fig.(a): = 00, = 8 tip= 20 Fig.(b): = 150, = 8 tip= 20;Upwindside
Measuring Grid
h
d
H
Smoke visualisation plane
Camera
Figure4.25Schematicdiagramusedforparallaxcorrection.
79
Chapter4 AerodynamicAnalysisoftheTUDelftModelTurbine
Thetipvortexpitchpwasdeterminedbymeasuringthedistancealongthexsaxisbetween
twosuccessivevortexcoresoriginatingfromthesameblade.Anupwindanddownwindp
hasbeendefined,forthetipvorticesshedontheupwindanddownwindsideoftherotor
plane, respectively. Since in this study, the flow visualisation was limited to a small
distancedownstreamoftherotorplane(lessthan2R),onlythefirstpitchisquoted.
Thewakeskewangleswasfoundasfollows:astraightlinewasdrawnjoiningthevortex
cores on the upwind side to the ones having the same age on the downwind side. This
couldbedone,sinceatwobladedmodelhasbeenused.Themidpointofeachlinewasthen
found.Theaxisoftheskewedwakewasdeterminedbyconstructingabestfitstraightline
originating at the yaw axis and passing through these midpoints (see Fig. 4.23). The wake
skew angle could then be taken as the angle between the rotor axis and the axis of the
skewedwake.
An analysis into the uncertainties of the estimated vortex centre locations has been
performed.Asstated,thecontributionofparallaxbiasuncertaintiestothetotaluncertainty
has been ignored. It is estimated namely, that its effect on the error in the vortex centre
location is substantially smaller than 1.5 cm. Averaging of the measured vortex centre
locations is the main contributor to the random uncertainty. Enclosed within the
contribution from averaging is the effect of the vertical drift of the smoke, out of the
measurement plane. For each visualisation measurement, six photos have been used for
averaging.Themaximumintervalofrandomuncertaintyintheaveragetipvortexlocation
has been estimated at 1.5 cm and 0.5 cm in the xs and ysdirection, respectively. It is
estimated that the random uncertainty in average tip vortex location translates in an
uncertaintyintervalof1.5ins.
AxialForce
A measurement campaign solely dedicated to the measurement of axial thrust (T) on the
rotoratalldifferentsettingswascarriedout.Thethrustmeasurementshadtobecorrected
due to considerable structural interference. Details of this correction are explained in
reference [36]. The corrected values for the thrust were azimuthally averaged and non
dimensionalizedusingthestandardequation:
C =
T (4.10)
T
1 AU 2
2
whereAistherotorsweptareaequaltoR2.
ForeachmeasurementofboththetipvortexlocationsandCT,wastakentobeequaltothe
average over 36 rotor revolutions. The maximum standard deviation over the 9 different
settings of (, tip) was equal to 0.12. CT was also taken to be equal to the average over 36
80
Chapter4 AerodynamicAnalysisoftheTUDelftModelTurbine
rotorrevolutions.Themaximumstandarddeviationoverthe9differentsettingsof(, tip)
wasequalto0.018.
ResultsandDiscussion
Axialthrustcoefficients
Asalreadyearlier,CTwasmeasuredovertherangeofyawangles450to450atincrements
of150.TheCTvalues(averagedoveronewholerevolution)areshowninFig.4.26.Ideally,
atagiven(, tip),thevariationofCTwithshouldbesymmetricalabout=00.However,it
canbeobservedthatthereisaconsistentlackofsymmetrywiththeCTvaluesfornegative
yaw being slightly lower than the corresponding values at positive yaw. Probably, the
reasonfortheasymmetryisthenonuniformityinvelocitydistributionofthetunnelexitjet.
ReferringtoFig.4.5,thevelocitydipregionisnotlocatedcentrallyatthetunnelaxis,butis
shiftedslightlytotheright.ThedifferencesinCTarelargestfor =150.At =300and
=450, the symmetry is better. The maximum difference is 0.076 at =150 and (, tip) =
(10,20).
Tipvortexlocations
Fig.4.27illustratesthetipvortexlocationsobtainedfromthesmokevisualizationphotosat
=00. The tip vortex trajectory determines the wake boundary and hence the wake
expansion. Recall that ys and xs are the distances relative to the blade tip in a direction
perpendicularandparalleltotheundisturbedwindspeed(seeFig4.23).Forthisaxialflow
condition,onlythetrajectoryononesideisplottedsincethewakeexpansionappearedto
besymmetrical.Itmaybenotedthatthewakeexpansionishighlysensitivetoboth and
tip.Figs.4.28(a)and(b)displaysthetipvortexlocationsat =300 atboththeupwindand
downwindside.Fig.(a)showstheeffectofchanging tip,whilemaintaining constantat8.
Fig.(b)showstheeffectofchanging ,whilemaintaining tipconstantat20.Inbothcases,
thewakeexpansiononthedownwindsideissignificantlylargerthanthatontheupwind
side.Thisisduetothefactthatinayawedrotor,thereisacomponentofthefreestream
velocity(equaltoUSin())thatactsinadirectionparalleltotherotorplane(refertoFig.3.1,
Chapter3).
81
Chapter4 AerodynamicAnalysisoftheTUDelftModelTurbine
1.2
0.8
= =6,6,tip==00 0
= =6,6,tip==22 0
CT
0.6
= =6,6,tip==44 0
= =8,8,tip==00 0
0.4
= =8,8,tip==22 0
= =8,8,tip==44 0
= =10,
10,tip==0
0
0
0.2
= =10,
10,tip==2
2
0
= =10,
10,tip==4
4
0
0
-45 -30 -15 0 15 30 45
( deg)
Figure4.26Variationofaxialthrustcoefficientwithyawangleatallcombinationsofandtip.
18
==6,6,tip==000
16
==6,6,tip==440
==8,8,tip==000
14
==8,8,tip==440
12 ==10,
10,tip==000
==10,
10,tip==44
0
10
ys (cm)
0
0 15 30 45 60 75 90
x s (cm)
Figure4.27Tipvortexlocationsat = 00 atdifferentcombinationsofandip.
82
Chapter4 AerodynamicAnalysisoftheTUDelftModelTurbine
Tipvortexpitch
FromthetipvortexlocationsasshowninFigs.4.27and4.28,itwaspossibletoextractthe
tipvortexpitch(p)atdifferent(,tip).AsshowninFig.4.23,pisthepitchalongthexsaxis.
Theresultsfor =00aregiveninFig.4.29.Inthisplotthepitchesonboththerightandleft
handsideareincluded.Notethatthedifferencesinpbetweentheleftandrighthandsides
aresmall.Higher syieldhighervaluesforp,butitmaybeobservedthat tipalsohasa
considerableinfluenceoverp.Fromknowledgeoftherotationalspeedandp,itispossible
to obtain an estimate the transport velocity of the vortex cores. It was observed that this
velocityisgenerallynotmuchsmallerthanthefreestreamvelocity(ontheorderof7095%
ofU).
Fig.4.30showsthetipvortexpitchvariationsat=300forboththeupwindanddownwind
sides. p is larger on the upwind side than on the downwind side. Consequently, the tip
vortexcorevelocitiesontheupwindanddownwindsidearedifferentandthisreflectsthe
asymmetryofthewakeasaresultofrotoryaw.
Wakeskewangle
A parameter often used in modelling the skewed wakes of yaw rotors is the wake skew
angle s. It is used in several engineering models of BEM codes as already outlined in
chapter3(section3.5).InFig.4.31,thevariationofswithisplottedforthedifferentrotor
settings. swasderivedfromthetipvortexlocationsusingtheproceduredescribedearlier
inpage80.Itisobservedthat||>||atalltipspeedratiosandbladepitchsettings.Thisis
due to the fact that the wake expansion is larger on the downwind side than that on the
upwindside.Forthesubjectrotor,smaybeasmuchas80largerthan.
To investigate the influence of the rotor axial thrust on the wake skew angle, |s| was
plotted against CT for the different yaw angles, see Fig. 4.32. It is noted that for each yaw
angle,therelationof|s|withCT isquitelinear.Higherthrustvaluesresultinlargerwake
skewangles.
In BEM engineering models for yaw, the wake skew angle is usually modelled to be a
function of the axial induction factor (or velocity) (for example see Eqt. 3.24, Chapter 3).
The linearity found in Fig. 4.32 and also the fact that generally smoke visualization
measurements are much cheaper to perform than inflow measurements suggest that it
couldbesimplertodevelopengineeringmodelsthatrelate|s|withCTthan|s|witha1.
83
Chapter4 AerodynamicAnalysisoftheTUDelftModelTurbine
18
16
14
12
10
ys (cm)
8
Theta =00, Upwind
tip=0 Upwind
6
=2=02,
Theta
tip
Upwind
Upwind
tip=4=04,
Theta Upwind
Upwind
4
tip=0=00,
Theta Downwind
Dwnwind
2 tip=2= 2,
Theta
0
Downwind
Dwnwind
tip=4= 4,
Theta
0
Downwind
Dwnwind
0
0 10 20 30 40 50 60
x s (cm)
Figure(a)Effectofchangingtipwhilemaintaining constantat8.
14
12
10
8
ys (cm)
6
Lambda = 6, Upwind
= 6, Upwind
= 8,=Upwind
Lambda 8, Upwind
4
= 10,
Lambda = 10, Upwind
Upwind
= 6,=Downwind
Lambda 6, Dwnwind
2
= 8,=Downwind
Lambda 8, Dwnwind
= 10,
Lambda Downwind
= 10, Dwnwind
0
0 10 20 30 40 50 60
x s (cm)
Figure(b)Effectofchanging whilemaintainingtipconstantat20.
Figure 4.28 Tip vortex locations at = 300 at upwind and downwind side for different
combinationsofandtip.
84
Chapter4 AerodynamicAnalysisoftheTUDelftModelTurbine
60
55
50
45
p (cm)
40
35
tip=0
Theta =00, Right
Right
tip=2
Theta =02, Right
Right
30 tip=4
Theta =04, Right
Right
tip=0
Theta =00,
Left
Left
25 tip=2
Theta =02,
Left
Left
tip=4
Theta
0 Left
= 4, Left
20
5 6 7 8 9 10 11
Figure 4.29 Variation of tip vortex pitch with tip speed ratio at = 00 at right & left side for
differenttip.
55
50
45
40
p (cm)
35
tip=0
Theta = 00,Upwind
Upwind
tip=2
Theta = 02,Upwind
Upwind
30
tip=4
Theta = 04,Upwind
Upwind
tip=0
Theta = 00,Downwind
Downwind
25
tip=2
Theta
0
= 2.Downwind
Downwind
tip=4
Theta
0
= 4,Downwind
Downwind
20
5 6 7 8 9 10 11
Figure4.30Variationoftipvortexpitchwithtipspeedratioat = 300 atupwind&downwind
sidefordifferenttip.
85
Chapter4 AerodynamicAnalysisoftheTUDelftModelTurbine
60
45
|s| (deg)
==6,
6, tip
= =0
0 0
30
==6,
6, tip
= =2
2 0
==6,
6, tip
= =4
4 0
==8,
8, tip
= =0
0 0
==8,
8, tip
= =2
2 0
15
==8,
8, tip==4
40
==10,
10, tip
= =00
0
= =10,
10, tip==2
2
0
==10,
10, tip
= =44
0
0
0 15 30 45 60
| | (deg )
Figure4.31Variationofwakeskewanglewithrotoryawanglefordifferentcombinationsof and
tip.
55
50
45
40
35
|s| (deg)
30
25
20
15 ==- 15
150
10 ==- 30
300
5 ==- 45
450
0
0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.0 1.1 1.2
CT
Figure4.32Variationofwakeskewanglewithrotoraxialthrustcoefficientforthedifferentyaw
angles.
86
Chapter4 AerodynamicAnalysisoftheTUDelftModelTurbine
4.3EstimatingtheAerodynamicLoadsattheBladesfromWind
TunnelInflowMeasurements
4.3.1Methodology
As already outlined earlier in section 4.1, one of the objectives of the experiments on the
TUDelftwindturbinewastoprovideexperimentaldatatobeabletocarryoutanindepth
investigationofthelimitationsoftheBEMtheoryforyawedconditions.Asecondobjective
wastousetheexperimentaldatatovalidateanewlydevelopedfreewakevortexcode(see
Chapter5).Inordertoreachtheseobjectives,itwasvitaltohavetheunsteadyaerodynamic
loading distributions along the blades apart from the near wake inflow measurements.
Since the experimental setup used in this study was incapable of measuring such
distributions, unsteady aerofoil theory that accounts for shed vorticity effects in the wake
had to be employed in order to derive the aerodynamic loads from the inflow
measurements. The unsteady lift coefficient could be predicted with reasonable accuracy
sincetheinflowexperimentaldataavailablewasforturbineoperatingconditionsatwhich
theflowbehaviouroverthebladesremainedattached,evenatayawangleof450.
Fig. 4.33 illustrates the sequence of steps used in deriving the aerodynamic loads
distributions on the blades from the inflow measurements taken in the wind tunnel
experiments for operating conditions with =8, tip=20 and yaw angles 00, 300 and 450. In
section 4.2.2, it was explained how the axial flow velocities (wa,c) at the blade lifting lines
wereestimatedbyinterpolatinglinearlythemeasuredvaluesatthe3.5cmdownstreamand
6cm upstream planes (refer to Figs. 4.18(a), 4.19 and 4.20). The former were then used to
findthelocalangleofattackandrelativeflowvelocityatradiallocations40,50,60,70,80,
90%Rusingthefollowingbladeelementtheoryequations:
wa , c
= tan
1
r (4.11)
(w )
2
Vr = a ,c
+ ( r ) 2 (4.12)
Theaboveequationswereappliedforthedifferentazimuthanglesalongthewholeblade
revolution.Tangentialflowvelocitieswereignoredsincethesewerefoundtobeverysmall
compared to absolute velocities of the blades. The 2D unsteady lift coefficient (Cl,2D) was
calculatedusingtheunsteadyaerofoilmodeldescribedlateronsection4.3.2.Thiswasused
to determine the bound circulation at the blades with the KuttaJoukowski theorem in
accordancewith:
B ,2 D = 1 cVr Cl ,2 D (4.13)
2
Itwasfoundthatatthe40%Rand90%Rlocations,theboundcirculationwasunrealistically
high,eventhoughtheangleofattackwasdetermineddirectlyfromthemeasurements.This
87
Chapter4 AerodynamicAnalysisoftheTUDelftModelTurbine
wasbecausea2Dunsteadyaerofoilmodelwasbeingusedtodeterminetheliftcoefficients.
Thusatip/rootlosscorrectionhadtobeemployedto B,2Dsuchthattheboundcirculation
decreases gradually to zero at the blade tip and root. The corrected bound circulation
distribution is being denoted by B,3D. For each blade azimuth angle, the bladeelement
theory equations 3.16 were applied to determine the chordwise and normal aerodynamic
load distributions along the blades between 40%R and 90%R. For these calculations, a lift
coefficientcorrectedfortip/rootlosswasused.Thiswasderivedusingthecorrectedbound
circulationtogetherwiththeKuttaJoukowskitheorem:
2 B ,3 D
Cl ,3 D = (4.14)
cVr
The drag coefficient was obtained from 2D wind tunnel static data for the NACA0012
aerofoilatlowReynoldsnumber.Nocorrectionwasappliedtothestaticdragcoefficientfor
unsteady effects in the aerodynamic loads. This is acceptable since for attached flow
conditionstheinfluenceofdragonthevortexwakestructureisminimal.Notip/rootloss
correctionwasappliedtothedragcoefficient.Thiswasjustifiedsincetheanglesofattack
wereverysmallandthusthedragwasverysmallcomparedwiththelift.
The global aerodynamic loads and the output power of the turbine were computed by
numerically integrating the loading distributions at eachtime stepin accordance withthe
numericalmethoddescribedinAppendixB.Tocheckthevalidityofthesecalculations,the
axial thrust derived from the inflow measurements were checked against their
corresponding measured values. Finally, the inflow measurements were assessed for
possible uncertainties due to tunnel blockage. This assessment was carried out using a
newlydevelopedprescribedwakevortexcode(namedHAWT_PVC).
ThisChapterisorganizedintofiveseparatesectionsasfollows:
A. Section4.3.2describestheunsteadyaerofoilmodel.
B. Section4.3.3describesthesoftwaretoolsdevelopedforthiswork.Theseinclude:
(1)HAWT_LFIMwhichisbasicallyabladeelementtheorycodeimplementingthe
sequence of steps in Fig. 4.33 and (2) HAWT_PVC which is the prescribed wake
vortexmodelusedtoassesstunnelblockageeffects.Thiscodewasalsofoundvery
usefulinderivingatip/rootlosscorrectionfor=00.
C. Section 4.3.4 presents the methods used to assess whether the hotfilm
measurementswereinfluencedbytunnelblockageeffects.
D. Section4.3.5presentsasimplifiedanalyticalapproachforquantifyingerrorsinthe
derivedbladeloadingresultingfromerrorsintheinflowmeasurements.
88
Chapter4 AerodynamicAnalysisoftheTUDelftModelTurbine
E. Section4.3.6presentstheresultsincludingthedistributionsfortheangleofattack
(),flowvelocitiesrelativetotheblades(Vr)andaerodynamicloading(dT,dQ).
Inputexperimentaldataforwa,c atrotorplane
(Figs.4.18(a),4.19and4.20)
FindCl,2D usingunsteadyaerofoilmodel(section4.3.2)
FindCd,2D fromstaticwindtunnel2Ddata
FindB,2D (Eqt.4.13)
CorrectB,2D fortip/rootlosstofindB,3D
FindCl,3D fromB,3D(Eqt.4.14)
Comparerotorthrustderivedfrom
Calculategloballoadsand
inflowmeasurementswiththose
outputpowerusingnumerical
measureddirectlyusingstrain
methodofAppendixB
gauges
Assessfortunnelblockage
usingprescribedvortexcode
Figure4.33Sequenceofstepsusedinderivingtherotoraerodynamicloadsfromtheinflow
measurements.
89
Chapter4 AerodynamicAnalysisoftheTUDelftModelTurbine
4.3.2ATheoreticalMethodforFindingtheUnsteadyLiftCoefficientin
AttachedFlowforaRotatingBladeinYaw
Thissectiondescribesatheoreticalapproachforderivingtheunsteadyliftcoefficientsat
thebladesofayawedrotorfromtheinflowmeasurements.Themethodisonlyapplicable
for attached flow conditions, at which the angles of attack are small. The method was
adapted from Leishman [49, pgs 333340] to evaluate the unsteady lift coefficient as a
functionoftime.Inthisapproach,theinstantaneousliftcoefficientisconsideredtobethe
sumoftwocomponents,thenoncirculatoryandcirculatorylift.Theformeristheliftthat
arisesfromtheaccelerationeffectsoftheflowwhilethelatterisduetothecirculationabout
theaerofoil.
A.NonCirculatoryLiftCoefficient
For an almost rigid blade (experiencing minimal flapping), the noncirculatory lift
coefficientisfoundfrom[49]:
b . ..
C lnc (t ) = V ab (4.15)
V 2
whereaisequalto1/2forapitchaxisatthequarterchordlocation.bisequaltohalfthe
chordlength.
B.CirculatoryLiftCoefficient
Anindicialresponsemethodmaybeusedtoevaluatethecirculatoryliftcoefficient.Ifthe
indicial response function is known, then the unsteady loads due to arbitrary changes in
angle of attack can be obtained through the superimposition of the indicial aerodynamic
responsesbyapplyingDuhamelsintegral.Considerageneralsysteminresponsetoatime
dependent forcing function F(t, t>0). Given that the indicial response function F of the
system is known, then the system response u(t) to the forcing can be mathematically
expressedintermsofDuhamelsintegralas
t
dF
u (t ) = F (0) (t ) + (t ) d (4.16)
0
dt
By analogy with the equation above, the circulatory lift coefficient, Clc ,in response to an
unsteadyangleofattackcanbeexpressedas
C c (t ) = 2 (0) ( s ) + d ( ) ( s ) d (4.17)
s
l
0 dt
90
Chapter4 AerodynamicAnalysisoftheTUDelftModelTurbine
where is the indicial response function derived by Wagner [102] for the lift on a thin
aerofoilundergoingastepchangeinangleofattackinincompressibleflow.sisthereduced
timegivenby
t
2
c 0
s = Vdt (4.18)
whichrepresentstherelativedistancetravelledbytheaerofoilthroughtheflowintermsof
aerofoilsemichordsduringtimeintervalt.
A major difficulty in solving Duhamels intergral in Eqt. 4.17 deals with the Wagner
function . Although the Wagner function is known exactly, its evaluation is not in a
convenient analytic form. Therefore, the function is usually approximated by a simply
exponential or algebraic approximation. One approximation to the Wagner function,
attributedtoR.T.Jonesiswrittenintheformofanexponentialdecayfunctionasfollows:
where A1, A2, b1 and b2 are taken as 0.165, 0.335, 0.0455 and 0.3 respectively. Eqt. 4.17
includesthetimehistoryeffectsoftheshedwakeonthelift.
C.TimeVaryingIncidentVelocity
Eqts. 4.15 and 4.17 only consider the situation in which the local freestream velocity
relativetotheaerofoil,V,isconstant.Inyawedflow,thebladeelementofthewindturbine
blades will encounter a timevarying incident velocity. Consequently, the shed wake
vorticity leaves the aerofoil at a nonuniform velocity. A pictorial representation of this
phenomenonisgiveninFig.4.34.
Leishman [49] modifies Eqts. 4.15 and 4.17 to account for the time variation in incident
velocityasfollows:
b d (V ) ..
(t ) = ab (4.20)
nc
C
l
V
2
dt
B
S
Vs1Vs2
Figure 4.34 timedependent freestream velocity causes shed vorticity to be convected at a non
uniformspeed(Vs).
91
Chapter4 AerodynamicAnalysisoftheTUDelftModelTurbine
Cl s
d (V )
Cl ( s ) = V ( s ) V ( s0 ) ( s0 ) ( s ) + ds (4.21)
c
( ) ( s ) d
s0
= Cl e ( s )
Clistheliftcurveslopeequalto2/radianforincompressibleflow. e(s)canbeviewedas
aneffectiveangleofattackinthatitcontainsallthetimehistoryinformationrelatedtothe
unsteadycondition.
D.RecursiveSolutionfortheCirculatoryLiftusingtheDuhamelsIntegral
Leishman [49, pg 337] presents a solution for solving Eqt. 4.17 recursively. This section
presentsthissolutionhowevermodifiedtosolveEqt.4.21tobeabletocaterforanunsteady
flowvelocity.
FromEqt.4.21,theeffectiveangleofattackmaybeexpressedas
1 d (V )
s
( s ) =
e V ( s0 ) ( s0 ) ( s ) + ( ) ( s )d (4.22)
V ( s ) s0
ds
SubstitutingEqt.4.19in4.22,
1 d (V )
s
e (s) = V ( s0 ) ( s0 ) (1 A1e 1 A2 e 2 ) + ( ) (1 A1e b1 ( s ) A2 e b2 ( s ) ) d
b s b s
V ( s ) s0
ds
V ( s0 ) ( s0 ) V ( s0 ) ( s0 ) A1 e V ( s0 ) ( s0 ) A2 e
b 1 s b 2 s
=
V (s) V (s) V (s)
1
s
d (V ) A1 s
d (V ) A2 s
d (V )
+ ( ) d ( )e b1 ( s )
d ( )e b2 ( s )
d
V (s) s 0
ds V (s) s 0
ds V (s) s 0
ds
Terms V(so)(so)A1eb1s /V(s) and V(so)(so)A2eb2s /V(s) containing the initial values of V and
areshorttermtransientsandcanbeneglected.Consequently,theDuhamelintegralmay
beexpressedas
1
( s ) =
e [V (s) ( s) X (s) Y ( s)] (4.23)
V (s)
whereX(s)andY(s)areequalto
92
Chapter4 AerodynamicAnalysisoftheTUDelftModelTurbine
s
d (V )
X ( s ) = A1 ds
( )e b 1 ( s ) d (4.24a)
s0
d (V )
s
Y ( s ) = A2 ( )e b2 ( s ) d (4.24b)
s0
ds
Assumeacontinuoussystemwithtimestep swhichmaynotbeconstantandthats0=0.
Forthenexttimesteps+s,
s +s
d (V )
X ( s + s ) = A1 ( ) e b ( s +s ) d
1
0
ds
s
d (V ) s +s
d (V )
= A1 e b1 ( s )
( ) e b1 ( s )
d + A1 ( ) e b ( s +s ) d
1
0
ds s
ds
d (V )
s +s
X ( s + s ) = X (s) e b1 s
+ A1 e b1 ( s +s )
( ) e b d
1
s
ds
= X (s)e b1s
+ I (4.25)
where
d (V )
s +s
I = A1 e b ( s +s )
1
( ) e b d
1
s
ds
ToevaluateI,asimplifiedestimateford(V)/dsisusedbyapplyingabackwarddifference
approximationattimes+s:
d (V ) V ( s + s ) ( s + s ) V ( s ) ( s )
=
ds s
Thus
s +s
V ( s + s ) ( s + s ) V ( s ) ( s )
I = A1 e b ( s +s )
1
s eb d
1
V ( s + s ) ( s + s ) V ( s ) ( s ) 1 e
b1 s
= A1
b (4.26)
s 1
Expandtermeb1sintheformofapowerseries
93
Chapter4 AerodynamicAnalysisoftheTUDelftModelTurbine
b1 s b12 s 2 b13 s 3
e = 1 b1s + + + ... (4.27)
2! 3!
Substituting the above equation in Eqt. 4.26 and neglecting terms b12s2 and higher since
b1sisverysmall,wewillendupwiththefollowingsimplerelationforintegralI
[ ]
I = A1 V ( s + s ) ( s + s ) V ( s ) ( s ) (4.28)
PuttingEqt.4.28inEqt.4.25,weendupwitharecursiveequationgivenby
( )
X ( s + s ) = X s e 1 + A1 [V ( s + s ) ( s + s ) V ( s ) ( s ) ]
b s
or
( )
X ( s ) = X s s e 1 + A1 [V ( s ) ( s ) V ( s s ) ( s s ) ] (4.29)
b s
ArecursivesolutionmaybederivedforY(s)inEqt.4.24busingasimilarmethod.
Insummary,therecursivesolutionfortheindicialEqt.4.22consistsofthefollowingthree
onestepformulas:
1
( s ) =
e [V ( s) ( s ) X ( s) Y ( s )] (4.30a)
V (s)
X ( s ) = X ( s s ) e b s + A1 [V ( s ) ( s ) V ( s s ) ( s s ) ] (4.30b)
1
Y ( s ) = Y ( s s ) e b s + A2 [V ( s ) ( s ) V ( s s ) ( s s ) ] (4.30c)
2
E.NumericalSolutionsAlgorithmsforConditionsofSteadyYaw
In this work, a new approach is presented for solving the noncirculatory and circulatory
liftcoefficients(Eqts.4.20and4.21,respectively)asafunctionoftimeforarotorinsteady
yawed conditions. The main advantage of this approach is that the solution is based on
matrixinversionandthusiscomputationallyefficient.
Wenowconsiderawindturbineinafixedyawangle,rotatingatconstantangularspeed
in a steady and uniform wind flow field. Under such operating conditions, the unsteady
wakemaybeassumedtobeperiodic.Consideronewholerevolutionoftherotorbladethat
isdividedintoagivennumberofequallyspacedazimuthalpositions,asshowninFig.3.7
(section3.6).Thenumberofazimuthalpositionsisequalto tot. denotesthenumberofthe
timestep.Thetimeelapseduringonetimestepisgivenby
= 2 (4.31)
* tot
94
Chapter4 AerodynamicAnalysisoftheTUDelftModelTurbine
NumericalSolutionfortheNonCirculatoryLiftCoefficient
Given that V and are known at each blade azimuth angle, (i.e. at each rotor time step),
Eqt.4.20maybesolvedasfollows:
Itisfirstnecessarytoestablishamethodforevaluatingd/dt,d(V)/dtandd2/dt2ateach
blade azimuth angle. A simple way of doing so is to apply a forward difference
approximationateachtimestep.Ford/dt,thiswouldbe
d +1
= (4.32)
dt
The global error in this method is only in the order O(2). In this work, a more accurate
method was used based on the AdamBashforth multistep numerical integration
technique. This technique makes use of data at previous time steps inorder to predict a
solution at the next. In the case of a fourthorder AdamBashforth method, four previous
datapointsarerequiredandtheequationtakestheformof
= 1 + [55& 1
59& 2 + 37& 3 9& 4 ] (4.33)
24
Since we have a periodic wake then tot = 0 and the above equation may be written at
eachbladetimesteptoformasystemoftotequationswhichmaybewritteninmatrixform
as
0 55 0 0 0 0 9 37 59 & 0
59 55 0 0 0 0 9 37 &1
1
2 37 59 55 0 0 0 0 9 & 2
3 = 9 37 59 55 0 3
0 0 0 &
o o o o o o o o o
(4.34a)
o
0 0 9 37 59 55 0 0 &
o o o o o o o o o o
tot 1
0 0 0 0 9 37 59 55
& tot 1
[ ] [ ]
or = A * & (4.34b)
24
where = (
1 ) (4.34c)
95
Chapter4 AerodynamicAnalysisoftheTUDelftModelTurbine
TheglobalerrorinthismethodisonlyintheorderO(5).Anidenticalapproachisadopted
to find d(V)/dt and d2/dt2 using the same matrix A. After these time derivatives are
evaluatedateachtimestep,theyaresubstitutedinEqt.4.20toyieldthenoncirculatorylift
coefficientatallbladetimesteps.
NumericalSolutionfortheCirculatoryLiftCoefficient
Given that V and are known at each blade azimuth angle, (i.e. at each rotor time step),
Eqts.4.21maybesolvedrecursivelyasfollows:
ApplyingEqts.4.30a,b,cforeachtimestep,
e =
1
[V X Y ] (4.35)
V
X p X 1 = P (4.36a)
Y q Y 1 = Q (4.36b)
where
p = e b1s q = e b2 s (4.37)
P = A1 (V V 1 1 ) (4.38a)
Q = A2 (V V 1 1 ) (4.38b)
UsingEqt.4.18,thevaluesforasmallincrementalchangeinsbecomes
s = [V + V 1 ](4.39)
c
ThevalueofXateachtimestep issolvedasfollows:Eqt.4.36(a)isappliedforeachtime
stepinordertogivetotequationswhichcanbewritteninmatrixformas
96
Chapter4 AerodynamicAnalysisoftheTUDelftModelTurbine
1 0 0 0 0 0 0 p0 X 0 P0
p1 1 0 0 0 0 0
0 X1 P1
0 p2 1 0 0 0 0 0 X 2 P2
0 0 p 1 . 0 0 0 0 . .
=
0 0 0 p . 0 0 0 X P
1 1
0 0 0 0 . 1 0 0 X P
0 0 0 0 0 . 1 0 . .
0 0 0 0 0 0 p tot 1 1 X tot 1 P tot 1
or
r r
G X = P (4.40)
r
The values of X , [X0 X1 X ...Xtot1]T , at each time step are then evaluated using matrix
inversionsuchthat
r r
X = G 1 * P (4.41)
r
Thevaluesof Y arecalculatedinasimilarfashionhoweverreplacingpandPwithqandQ
respectively. Once the values of X and Y at each time step are obtained, Eqt. 4.35 is
employed to obtain the equivalent circulatory angle of attack to be used to find the
circulatoryliftcoefficient.
SummaryofNumericalSolutiontoEstimatetheUnsteadyLiftCoefficientfromtheInflow
Measurements
From inflow measurements, the values of and Vr at each time step may be evaluated
(ParameterVrinarotatingbladeisequaltoV).Thesecanthenbeusedtofindp,P,qandQ
usingEqts.4.37,4.38and4.39,andthusfindthecirculatoryliftcoefficient.Thelatteristhen
addedtothenoncirculatoryliftcoefficient(computedusingEqt.4.20)togivethetotallift
coefficientasafunctionoftime.
97
Chapter4 AerodynamicAnalysisoftheTUDelftModelTurbine
Divideonewholebladerevolutionintoequallyspacedtimesteps asinFig.3.7
ComputeNoncirculatoryLiftCoefficient ComputeCirculatoryLiftCoefficient
Procedure: Procedure:
(3)SolveforXandY
(4)ApplyEqt.4.35tofinde
(5)FindClc from Cl = 2 e
c
Addnoncirculatoryandcirculatorycomponentsofliftcoefficient:
Cl = Cl nc + Cl c
Figure4.35Summaryofmethodusedtodeterminetheunsteadyliftcoefficient.
4.3.3DevelopedSoftwareTools
A.DescriptionofProgramHAWT_LFIM
Program HAWT_LFIM (LFIM meaning Loads from Inflow Measurements) was specifically
developed to derive the aerodynamic loads at the rotor blades from the inflow
measurements using the procedure of Fig. 4.33. This code was written using MathCad
version11andisapplicableforbothnonyawedandyawedconditions.Fig.4.36describes
the structure of this code. This code is organized into three separate modules:in the Data
Input Module in which the experimental parameters describing the rotor geometry and
operatingconditionareinputted,togetherwiththeaxialflowvelocities(wa,c)ofFigs.4.18(a),
4.19 and 4.20. The Data Processing Module implements the various bladeelement theory
equationsandtheunsteadyaerofoiltheorytobeabletoestimatetheaerodynamicloading
98
Chapter4 AerodynamicAnalysisoftheTUDelftModelTurbine
distributionsattheblades.Thelatterareintegratednumericallytobeabletofindtherotor
globalaerodynamicloadsandoutputpower.TheDataOutputModuleoutputstheresultsas
afunctionofrotorazimuthangle.
DataInput:
(1)InputRotorGeometryDetails:
NumberofBlades(B)
BladeTipandRootRadii(Rt & Rr)
Bladechordandtwistdistributions(c& )
(2)InputRotorOperatingConditions:
RotorAngularSpeed()
RotorYawAngle()
WindSpeed(U)
(3)InputExperimentalData:
AxialInflowVelocitiesatbladeliftinglines
(wa,c) estimatedfromhotfilmmeasurements
DataProcessing:
Calculationofthefollowing:
(6) globalloads&outputpower
DataOutput:
OutputofresultsfrommoduleDataProcessing at
eachbladeazimuthangle()
Figure4.36StructureofcodeHAWT_LFIM.
99
Chapter4 AerodynamicAnalysisoftheTUDelftModelTurbine
B.DescriptionofProgramHAWT_PVC
Asalreadydescribedearlierinsection4.3.1,theprescribedwakevortexmodelwasmainly
developedtobeabletoestimatequantitativelytheinfluenceofwindtunnelblockage.This
codewasalsohelpfulinapplyingatip/rootlosscorrectionat=00.Themainfunctionofthe
code is to estimate the induced velocities in the rotor wake using a known bound
circulationdistribution.Thecode,namedHAWT_PVC(PVCmeaningPrescribedVortexCode)
is also written using MathCad version 11 and is applicable for both nonyawed and
yawedconditions.Fig.4.37describesthestructureofthisvortexmodel.Theinputstothe
modelarethebladegeometryandtherotoroperatingparameterstogetherwiththeknown
boundcirculationdistributionsattheblades.
DataInput:
(1)InputRotorGeometryDetails:
NumberofBlades(B)
(2)InputRotorOperatingConditions:
RotorAngularSpeed()
RotorYawAngle()
WindSpeed(U)
(3)InputAerodynamicParameters:
BoundCirculationatBlades(B)
WakeGeometryDetails(s,p,Rt,w1 &Rt,w2)
DataProcessing:
(1)ModelWakeGeometry
(2)BiotSavart Computations
DataOutput:
InducedVelocitiesatPlaneParallelto
Rotorplane
Figure4.37:StructureofcomputercodeHAWT_PVC.
100
Chapter4 AerodynamicAnalysisoftheTUDelftModelTurbine
Since in a yawed rotor the bound circulation at the blades becomes timedependent, then
this circulation is prescribed to the code as a function of blade azimuth angle. The wake
parameters derived from the smoke visualization experiments (refer to section 4.2.3) are
also inputted and used by the code to model the prescribed wake geometry. From the
known bound circulation distributions, the trailing circulation distribution is derived and
used in conjunction with the BiotSavart law (refer to Appendix C) to calculate the 3D
inducedvelocitydistributionsatanyrequiredplaneparalleltotherotorplane.
BladeandWakeModel
Fig. 4.38 givesdetails for theadopted model for the rotorblades and wake. Each blade is
modelled using a liftingline with a single lumped vortex located along the quarterchord
point of the blade sections. The lifting line consists of a fixed number (n) of piecewise
constantspanwisesegments,eachofequallength.
The wake consists of helical vortex sheets, one per blade. Each vortex sheet consists of
heliceslocatedatdifferentradiallocations.Eachhelixissegmentedintostraightlinevortex
filaments to represent trailing circulation in the wake (refer to Fig. 4.38). For yawed
conditions,thehelicesareskewed,dependingontheinputtedvalueofs.Themodellingof
shed circulation was not included, since for the operating conditions of the TUDelft rotor
beingconsideredinthisstudy,thiscirculationcomponentwasfoundtobeverysmall.
Z, Za
k+1,b,i
(T )k,b,i
Ya [Xa,Ya,Za]k, b, i
O
Y
k-1,b,i
wake
nodes
X
Rotor Axis Xa
Blade Tip
Figure4.38:Modellingofbladesandwakeinprescribedwakevortexmodel.
101
Chapter4 AerodynamicAnalysisoftheTUDelftModelTurbine
Thetrailingfilamentnodesareinterconnectedbynodes.Todeterminethe3DCartesianco
ordinatesofeachwakenode,thefollowingmathematicalmodelisused:
Eachrotorrevolutionisdividedintoafixednumber(tot)ofequallyspacedazimuthsteps
asinFig.3.7.Theazimuthalstepofthewakehelicesismaintainedequaltothatoftherotor
(i.e.2/tot).Thusthenumberoffilamentsusedforeachrevolutionineachhelixisalsoequal
to tot.Thewakeextendsdownstreamdependingontheprescribednumberofrevolutions
(nwRev)andthevalueofthepitchp.Sinceinthesmokevisualizationexperiments,aslight
variationwasobservedbetweenthetipvortexpitchvaluesontheupwindanddownwind
sides(someresultsareplottedinFigs.4.29and4.30),theaveragevalueofthetwovalues
forpisusedinthemodel.Itisassumedthatthevortexsheetpitchisequaltothatofthetip
vortex at all radiallocationsin the wake. As shown in Fig. 4.23 pitch p is taken alongan
axisparalleltothefreewinddirection.Inthewakemodel,apitchpwactingalongtheYaaxis
isusedandthisisgivenby
pw = p cos( ) (4.42)
WakeGeometry
Eachwakenodeisdenotedbyindicesk,b,i(seeFig.4.38),wherebandidenotetheblade
number and the radial location from which the helix is originating. i is equal to 0 for all
wakenodesthatlieonthehelixthatisoriginatingfromthebladeroot.iisequalton1for
the wake nodes that lie on the helix originating from the blade tip. b=0 denotes the first
blade.kdenotesthenodenumberonthehelix,wherek=0isforthestartingnodethatlieson
the respective blade lifting line. The total number of vortex filaments used to represent a
singlehelixisdenotedbyktotandthisisequaltonwRev*tot.
Figs.4.39(a),(b)and(c)describethegeometryusedtodefinethewakeouter(tip)andinner
(root)boundaries.Foryawedconditions,bothoftheseareskewedwithrespecttotherotor
axisandtheircentralaxisisthewakeskewaxis.Thusthehelicaltrailingvorticeshavethe
wakeskewaxisastheircentralaxisandnottheYaaxis.Theouterwakeboundaryisdefined
using the prescribedterms Rt,w1 and Rt,w2 that are derivedfrom the smoke visualization of
tipvortexcores(refertosection4.2.3).Thesedefinetheexpansionofthewakeinthenear
fieldoftherotor.Itisassumedthatthecrosssectionofthewakeinaplaneparalleltothe
rotorplaneisannular,asshowninFig.4.39(c).AtanydistanceYafromtherotorplane,the
radiioftheouterandinnerwakeboundariesaredenotedbyRt,wandRr,wrespectively.
Fortheouterwakeboundary,aquadraticfitisappliedforthevariationofRt,wwithYafor
Ya<pwsuchthat
Rt , w (Ya ) = a0Ya 2 + a1Ya + a2 (4.43)
where constants a0, a1 and a2 are determined by applying the following three boundary
conditions:(i)atYa=0(i.e.attherotorplane),Rt,w=0.99Rt;(ii)atYa=pw/2,Rt,w=Rt,w1and(iii)at
102
Chapter4 AerodynamicAnalysisoftheTUDelftModelTurbine
Ya=pw,Rt,w=Rt,w2(refertoFig.4.39(a).Inboundarycondition(i),itisbeingassumedthatthe
outboardedgevortexsheetisreleasedfromthebladeat0.99Rt.
Applying the above three boundary conditions to Eqt. 4.43 leads to three equations that
maybewritteninmatrixformas
0 0 1
a0 0.99 Rt
( ) ( )
1 * a1 = Rt , w1 (4.44)
2
w
p pw
2 2
p 2 a R
w pw 1 2 t , w 2
Usingmatrixinversion,constantsa0,a1anda2maybeevaluated,asfollows:
1
0 0 1
a0 0.99 Rt
( ) ( )
2
a1 = w 1 * Rt , w1 (4.45)
p pw
2 2
a2 R
pw
2
pw 1 t , w 2
For Ya>pw, Rt,w is set constant and equal to Rt,w2. For the inner wake boundary no wake
expansionistakenintoconsideration.Thisboundarytakestheformofaskewedcylinder
havingaconstantradiusequaltotherotorbladerootradiusRr.
Tosummarize,thewakeboundariesaredefinedbythefollowingequations:
Rt , w (Ya ) = a0 Ya 2 + a1Ya + a2 if Ya pw
= Rt , w 2 if Ya > pw (4.46)
Rr , w (Ya ) = Rr Ya
Thecoordinatesofeachwakenodeinpolarcoordinatesisdenotedby[(rw)k,b,i,(w)k,b, (Ys)k]
wherethecentralaxisisalignedwiththewakeskewaxisYs(refertoFig.4.39(a)).rwdenotes
the radial distance and is equal to Rt,w for the wake nodes lying on the outer edge of the
vortexsheets(i=n1),i.e.thewakenodesthatlieonthehelixoriginatingfromthebladetip
(at r=0.99Rt). rw is equal to Rr,w for the wake nodes lying on the inner edge of the vortex
sheets(i=0),i.e.thewakenodesthatlieonthehelixoriginatingfromthebladeroot(atr=Rr).
For Ya pw, the outermost helix (from the blade, i=n1) will experience an expansion in
accordance with Eqt. 4.46 while the innermost helix (from the blade root, i=0) does not
expand.
103
Tipvortexcore
104
Outerwakeboundary
pw
Ya
Rt
s
Wakeskewaxis
Y
Ys
R t,w1
R t,w
2
R t,w
(Y a
)
X
Xa
Outerwakeboundary
Fig.(a):Geometryforouterwakeboundaryofprescribedwake
Za
Outerwakeboundary
Innerwakeboundary
Ya w
a)
Innerwakeboundary
(Y
t,w
R
Rr
s Wakeskewaxis
Y
Xa
R r,w
Ys
(Y a
)=
Rr
X
Innerwakeboundary
Wakeskewaxis
Rr,w (Ya)
Xa
Fig.(c):CrosssectionthroughprescribedwakeatgivenYa
Fig.(b):Geometryforinnerwakeboundaryofprescribedwake
Figure4.39 Schematicdiagramsdescribinghowprescribedvortexwakewasmodelled
Chapter4 AerodynamicAnalysisoftheTUDelftModelTurbine
Chapter4 AerodynamicAnalysisoftheTUDelftModelTurbine
The rest of the helices (i=1,2..n2) are allowed to expand in accordance with the following
linearinterpolatingformula
Rt , w ( Ya ) Rr , w ( Ya )
rw (Ya ) = Rr , w (Ya ) + ( r Rr ) (4.47)
Rt Rr
Sinceafteronewholehelicalrevolution,eachhelixwilladvancebyonepitch(pw),thenthe
axialdistancealongtheYaaxismaybewritteninindexform(Ya)kas
k pw
(Ya )k = (4.48)
tot
andEqt.4.47inindexformisexpressedby(rw)k,b,ias
( R ) ( Rr , w ) k ,b ,i
( rw ) k ,b ,i = ( Rr , w ) k ,b ,i + t , w k ,b ,i ( ri Rr ) (4.49)
Rt Rr
Theanglewateachwakenodeisdeterminedfrom
2 b
(w )k ,b = ( ) + k (4.50)
B
The position of each wake node on the vortex sheets is expressed in 3D Cartesian co
ordinates[(Xa)b,k,i,(Ya)k,(Za)b,k,i]usingthefollowingtransformationequations:
Finally, the wake node coordinates are transformed into the global fixed XYZ frame of
referenceusingthefollowingtransformation:
105
Chapter4 AerodynamicAnalysisoftheTUDelftModelTurbine
( X )k ,b ,i cos ( ) sin ( ) 0 1 ( X a ) k ,b ,i
Y
( Y ) k ,b ,i = sin ( ) cos ( ) 0 ( ) (4.53)
a k
1 ( Z a )k ,b ,i
( Z )k ,b ,i
0 0
WakeTrailingCirculationDistribution
T represents the trailing circulation in the wake due to a spanwise variation in bound
circulation.Eachtrailingvortexfilamentinthewakeisdenotedby(T)k,b,iandisconnected
to nodes (k,b,i) and (k+1,b,i), refer to Fig. 4.38. When the turbine is operating in steady
conditions with its axis parallel to the windspeed direction, the bound circulation at the
blades isconstant. Consequently, eachhelixofthewake vortexsheetwillhaveauniform
circulation.Fornonyawedconditionshowever,theboundcirculationbecomesafunction
ofthebladeazimuthangle().Thus,thecirculationateachhelixisvariedfromfilamentto
filament, depending on its azimuth position of the filament, i.e. (w)k,b. The assignment of
the circulation at each trailing filament is carried out in accordance with the following
relation:
(T ) k ,b ,i = ( B ) ,b ,i 1 ( B ) ,b ,i (4.54a)
where
2 b
(w )b , k = + (4.54b)
B
Eqt.4.54(b)isusedsothateachfilamentisassignedacirculationcalculatedusingthebound
circulationwhenthebladesareatthesameazimuthangleasthatofthefilament.
NumericalSolution
106
Chapter4 AerodynamicAnalysisoftheTUDelftModelTurbine
=2
(n1,p) /
to
t
(ip,p1) (ip+1, p )
(ip,p)
(ip1,p)
(0,p) (ip,p+1)
Figure4.40NotationusedforcalculationspointswithinplanedistantYapfromrotorplane.
Foreachrotortimestep ,thefollowingstepsarecarriedout
(1) thegeometry ofthewakevortexsheetduetoeachbladeismodelledusingEqts.
4.464.53.
(2) thetrailingcirculationateachwakefilamentisfoundusingEqts.4.54(a),(b).
(3) theBiotSavartequations(AppendixC,Eqts.C.24..32)areappliedtocalculatethe
3D induced velocity components at each calculation point (ip,p) due to all bound
vorticesatthebladesandtrailingvortexfilamentsinthewake,asfollows:
B 1 n 1 ktot B 1 n 1
( u X )ip , p = ( B ) ,b,i * GXBip , p , ,b,i + ( T )k ,b,i * GXTip , p ,k ,b,i
b=0 i =0 k =0 b =0 i =0
B 1 n 1 ktot B 1 n 1
( uY )ip , p = ( B ) ,b,i * GYBip , p , ,b,i + ( T )k ,b,i * GYTip , p ,k ,b,i
b=0 i =0 k =0 b=0 i =0
B 1 n 1 ktot B 1 n 1
( uZ )ip , p = ( B ) ,b,i * GZBip , p , ,b,i + ( T )k ,b,i * GZTip , p ,k ,b,i
b =0 i =0 k =0 b=0 i =0
(4.55)
where GXB, GYB and GZB are the geometric influence coefficients due to the bound
circulation vortices while GXT, GYT and GZT are the geometric influence coefficients for
the trailing circulation vortices. The discretization equations for these coefficients are
presentedinTables4.2and4.3.Asimplenumericalcutoffmethodisusedtodesingularise
107
108
( r1 ) + r Y Y Z Z Z Z Y Y
ip , p , ,b ,i ( 2 )ip , p , ,b ,i ( P )ip , p ( B ) ,b ,i +1 ( B ) ,b ,i +1 ( B ) ,b ,i ( P )ip , p ( B ) ,b ,i +1 ( B ) ,b ,i +1 ( B ) ,b ,i
GXBip , p , ,b ,i = 2 2
2 ( r1 )ip , p , ,b ,i ( r2 )ip , p , ,b ,i ( r1 )ip , p , ,b ,i + ( r2 )ip , p , ,b ,i ( L ) ,b ,i
r + r Z Z X X X X Z Z
( 1 )ip , p , ,b ,i ( 2 )ip , p , ,b ,i ( P )ip , p ( B ) ,b ,i +1 ( B ) ,b ,i +1 ( B ) ,b ,i ( P )ip , p ( B ) ,b ,i +1 ( B ) ,b ,i +1 ( B ) ,b ,i
GYBip , p , ,b ,i = 2 2
2 ( r1 )ip , p , ,b ,i ( r2 )ip , p , ,b ,i ( r1 )ip , p , ,b ,i + ( r2 )ip , p , ,b ,i ( L ) ,b ,i
r + r X X Y Y Y Y X X
( 1 )ip , p , ,b ,i ( 2 )ip , p , ,b ,i ( P )ip , p ( B ) ,b ,i +1 ( B ) ,b ,i +1 ( B ) ,b ,i ( P )ip , p ( B ) ,b ,i +1 ( B ) ,b ,i +1 ( B ) ,b ,i
GZBip , p , ,b ,i = 2 2
2 ( r1 )ip , p , ,b ,i ( r2 )ip , p , ,b ,i ( r1 )ip , p , ,b ,i + ( r2 )ip , p , ,b ,i ( L ) ,b ,i
2 2 2
( r1 )ip, p, ,b,i = X X + Y Y + Z Z
( P )ip , p ( B ) ,b ,i ( P )ip , p ( B ) ,b ,i ( P )ip , p ( B ) ,b ,i
2 2 2
( r2 )ip , p , ,b ,i = X X + Y Y + Z Z
( P )ip , p ( B ) ,b ,i +1 ( P )ip , p ( B ) ,b ,i +1 ( P )ip , p ( B ) ,b ,i +1
2 2 2
( L ) ,b,i = ( X ) X + Y Y + Z Z
B ,b ,i +1 ( B ) ,b ,i ( B ) ,b ,i +1 ( B ) ,b ,i ( B ) ,b ,i +1 ( B ) ,b ,i
Table4.2:Discretization equationsforinfluencecoefficientsforboundcirculation
Chapter4 AerodynamicAnalysisoftheTUDelftModelTurbine
( r1 ) + r Y Y Z Z Z Z Y Y
ip , p ,k ,b ,i ( 2 )ip , p ,k ,b ,i ( P )ip , p ( )k +1,b ,i ( )k +1,b ,i ( )k ,b ,i ( P )ip , p ( )k +1,b ,i ( )k +1,b ,i ( )k ,b ,i
GXTip , p ,k ,b ,i = 2 2
2 ( r1 )ip , p ,k ,b ,i ( r2 )ip , p ,k ,b ,i ( r1 )ip , p ,k ,b ,i + ( r2 )ip , p ,k ,b ,i ( L )k ,b ,i
( r1 ) + r Z Z X X X X Z Z
ip , p ,k ,b ,i ( 2 )ip , p ,k ,b ,i ( P )ip , p ( ) k +1,b ,i ( )k +1,b,i ( )k ,b,i ( P )ip , p ( ) k +1,b,i ( )k +1,b,i ( )k ,b ,i
GYTip , p ,k ,b ,i = 2 2
2 ( r1 )ip , p ,k ,b ,i ( r2 )ip , p ,k ,b ,i ( r1 )ip , p ,k ,b ,i + ( r2 )ip , p ,k ,b ,i ( L ) k ,b ,i
( r1 ) + r X X Y Y Y Y X X
ip , p ,k ,b ,i ( 2 )ip , p ,k ,b ,i ( P )ip , p ( )k +1,b ,i ( )k +1,b ,i ( )k ,b ,i ( P )ip , p ( )k +1,b ,i ( )k +1,b ,i ( )k ,b ,i
GZTip , p ,k ,b ,i = 2 2
2 ( r1 )ip , p ,k ,b ,i ( r2 )ip , p ,k ,b ,i ( r1 )ip , p , k ,b ,i + ( r2 )ip , p ,k ,b ,i ( L )k ,b ,i
2 2 2
( r1 )ip , p,k ,b,i = X X + Y Y + Z Z
( P )ip , p ( )k ,b ,i ( P )ip , p ( )k ,b ,i ( P )ip , p ( )k ,b ,i
2 2 2
( r2 )ip, p ,k ,b,i = X X + Y Y + Z Z
( P )ip , p ( )k +1,b ,i ( P )ip , p ( )k +1,b ,i ( P )ip , p ( )k +1,b ,i
2 2 2
( L ) k ,b ,i = ( X )
k +1,b ,i
( X )k ,b ,i + (Y )k +1,b ,i (Y )k ,b ,i + ( Z )k +1,b ,i ( Z )k ,b ,i
Table4.3:Discretization equationsforinfluencecoefficientsfortrailingcirculation
109
Chapter4 AerodynamicAnalysisoftheTUDelftModelTurbine
Chapter4 AerodynamicAnalysisoftheTUDelftModelTurbine
theBiotSavartequations.NotethatinEqts.4.55above,theinductioncomponents(uX,uY,
uZ) are in the XYZ reference frame. These are finally transformed into the moving xyz
frameofreferenceusingthefollowingtransformation:
(4.56)
whereux,uyanduzarethetangential,axialandradialinductioncomponentsrespectively.
The above three steps are repeated for all rotor time steps ( = 0..tot1). This yields a 3D
induction distribution at the calculation plane at all azimuth steps of one whole rotor
revolution.
4.3.4AssessmentofWindTunnelBlockageEffects
Fromtheinflowmeasurements(refertoFigs.4.18(a),4.19and4.20,seesection4.2.2),itwas
required to derive the axial induced velocity components at the rotorplane by applying
equation
ua = wa U cos( ) (4.57)
Amajordifficultyinfindingtheaxialinducedvelocitieswasinestablishingthetruefree
windspeedaccurately(U)tobeabletouseitinEqt.4.57.Duetotheinfluencesoftunnel
blockage,theexitjetvelocitymeasuredatthreepointsusingthepitotstaticprobesmaybe
slightlydifferentfromthetruefreewindspeed.Itwillbeshownlateroninthissectionthat
asmalldifferenceinthevalueofUmayyieldtolargeerrorsinthederivedaxialinduced
velocities.
Whenawindturbineisoperatinginarealopenairenvironment,thenthewakeisallowed
to expand freely, as illustrated in Fig. 4.41(a). Outside the wind turbine wake, the flow
velocity may be considered to be uniform and equal to the freewind speed. The flow
characteristics across a wind turbine in a wind tunnel may therefore be considerably
differentthanthosethatwouldbeexperiencedifthesameturbineisoperatinginanopen
air environment. This is because in a wind tunnel, the flow field is limited to a restricted
flowchannel.Suchdifferencesgiverisetowindblockageeffects.Amajorprerequisitein
windtunneltestingistoachieveaflowenvironmentthatisascloseaspossibletothatofan
110
Chapter4 AerodynamicAnalysisoftheTUDelftModelTurbine
open air environment. Precautions should therefore be taken to ensure that blockage
influences are minimal. If not, then the measured results should be corrected. Physically,
tunnelblockageisrelatedtothemomentumreductioninthewake.Theamountofblockage
willdependonthewindtunnelconfigurationused.Inanopenjetwindtunnel(similarto
theoneusedinthisstudy),blockageeffectsarelessthanthatinaclosedsection,sincethe
turbine wake is less restricted from expanding (see Figs. 4.41(b) and (c)).Also, for a given
tunnelconfiguration,tunnelblockagedependsontheratioofthecrosssectionalareaofthe
rotortothatofthetestsection.Thelargerthisratio,thelargertunnelblockageeffectstend
tobe.Thisratioisoftenreferredtotheblockagefactor(denotedherebyBF).Foranopenjet
tunnel,BFistakentobeequaltoArotor/Ajet.ForthisstudyontheTUDelftrotor,BFwasequal
to0.29.
Basically,twomainblockagephenomenatakeplaceinanopenjetwindtunnel:
(1) BlockageduetoWakeBoundary
If we consider a closedsection wind tunnel (see Fig. 4.41(b)), wake expansion is
constrainedbythetunnelwalls.Whentherotorisoperatinginawindturbinestate,the
flow velocity wa would be less than the freestream velocity U. But since the same
volume of air that passes any section upstream of the turbine must pass any section
behind it, it follows that the velocity bypassing the rotor wb (outside the slipstream)
wouldbegreaterthanU.Inanopenairenvironment,thiswb wouldbeequaltoU(Fig.
4.41(a)).Fortheclosedtestsection,thestaticpressureoftheairbypassingtheturbine
would therefore be less than that of the undisturbed stream having velocity U. This
influencestheturbinesothatitdevelopsathrustlargerthanwouldbedevelopedinan
unrestrictedflowofthesamespeedwiththesamerotorangularspeedandbladepitch.
Alternatively, it can also be argued that the thrust developed would be equal to that
produced when the turbine is operating in an openair environment at a higher
windspeed.
Inanopenjetwindtunnel,theflowfieldislimitedbythesizeofthejet.However,the
jet is not constrained by walls and thus the turbine wake is allowed to expand more
freelythaninclosedtestsection(seeFig.4.41(c)).Howeverifwbatanypointinthejet
flowbypassingtheturbineisnotequaltoU,thensomewakeboundaryblockagewill
stillbepresent.Asimplefirstorderrelationforwbmaybederivedfrombasicactuator
disktheoryforaturbineinaxialconditionsinaninviscidandincompressibleflow:
Consider two sections p and q in the fluid stream such that p is located far upstream
fromtherotorwherethetunnelspeedisequaltoUwhileqislocatedatadownstream
distancefromtherotor(seeFig.4.42).LetDjbethediameterofthetunneltubewhileDj
be the diameter of the tunnel jet at section q. Let the diameter of the rotor wake at
sectionqbeD.Theflowvelocityatsectionqisdenotedbywa.ReferringtoFig.4.42
111
Chapter4 AerodynamicAnalysisoftheTUDelftModelTurbine
BypassVelocitywb=U
U
WakeFlowVelocitywa<U
BypassVelocitywb=U
Fig(a) Openairenvironment
BypassVelocitywb>U
Tunnelwall
U WakeFlowVelocitywa<U
BypassVelocitywb>U
Fig(b) Windtunnelwithclosedtestsection
BypassVelocitywb
Jetboundary
U
WakeFlowVelocitywa<U
BypassVelocitywb
Fig(c) Windtunnelwithopentestsection
Figure4.41:Wakedevelopmentsinopenairandintwodifferentwindtunneltypes(openandclosed
testsection).
112
Chapter4 AerodynamicAnalysisoftheTUDelftModelTurbine
wb
D
U
Dj wa Dj
wb
p
q
Figure4.42Nomenclatureforsimpleanalysisofwindturbineinanopenjetwindtunnel.
andapplyingthevolumecontinuityequationbetweensectionspandq,
D j 2U = D '2 wa '+ ( D j '2 D '2 ) wb
D '2
U 2
wa '
Dj (4.58)
wb = '2
D
1
D j '2
Leta1bethemagnitudeoftheaxialinductionfactoratsectionq
wa ' U
where a1
'
=
U
DividingbothsidesofEqt.4.58byUandsubstitutingfora1yields
D '2
1
Dj 2 (1 + a ) 1
'
wb
= (4.59)
U D '2
1
D j '2
ItcanbeeasilyshownfromEqt.4.59thatforthewindturbinewaketoreachtheidealwake
boundaryconditionasthatforanopenairenvironment(i.e.wb/U=1),thejetdiameteratany
distancefromtherotorplaneshouldexpandinaccordancewith
Dj
D j
'
= (4.60)
1 + a1'
Sinceforthenormaloperatingstateofawindturbinea1<0,thenDj>Dj.Alsofromsimple
axialmomentumconsiderations,a1willincreasenegativelydownstreamuntilitreachesa
113
Chapter4 AerodynamicAnalysisoftheTUDelftModelTurbine
2.4
2.2
1.8
Dj'/Dj
1.6
1.4
1.2
1
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9
- a 1'
Figure4.43Requiredopenjetexpansionatdifferentlocalaxialinductionfactorsaspredictedby
Eqt.4.60tohavetheidealsituationwherenowakeblockageispresent.
valuethatisdoublethatattherotorplaneinthefarwake.Thisimpliesthatthetunneljet
diametershouldincreasegraduallydownstreaminaccordancewithEqt.4.60.Fig.4.43plots
the required wake expansion with the local axial induction factor. In reality such a
condition would be very difficult to obtain and consequently some wake blockage will
alwaysbepresent.Yet,theaboveanalysisrevealsthatbyhavinganopenjettunnelinstead
ofatunnelwithaclosedtestsection(whereDjhastobemaintainedconstantandequalto
Dj),wakeboundaryblockagewouldbelesssignificant.Foryaw,therequiredjetexpansion
wouldbemorecomplex(duetowakeskew)andevenmoredifficulttohave.
(2) BlockageduetoproximityofRotortoTunnelExitJet
Whenawindturbineisplacedinthejetofthewindtunnel,itspresencewillbealsofelt
upstream of the rotorplane. In fact, the retardation of the flow starts way upstream
fromtherotor,asshowninFig.4.44.Thusiftherotorisplacedclosetothetunnelexit
wherethepitotreadings(Ujet)aretaken,thelatterwillnotmeasurethetruefreestream
velocity.Inaddition,therotorwillcausethetunnelexitvelocitydistributiontobecome
nonuniform. The nonuniformity becomes more complex when the rotor is yawed.
ThiseffectisillustratedpictoriallyinFigs.4.45(a)and(b).Thisseverityofthistypeof
blockagedependsontheblockagefactor(BF),thedistanceoftherotorfromthetunnel
exit,therotorgeometryandtheoperatingcondition.Therotorgeometryandoperating
conditionwilldeterminethethrustexertedbytherotoronthefluidstream.Ahigher
operatingthrustwillmakethepitotreadingsmoresusceptibletothistypeofblockage.
Inyawedconditions,thiseffectcouldbemoreseverefortworeasons:firstbecausethe
114
Chapter4 AerodynamicAnalysisoftheTUDelftModelTurbine
Flow
Error Rotordisk
U Ujet
Axialdistance
Figure4.44DiscrepancybetweenUjetandUduetoproximityofrotortopitottubes.
upwind blade will be closer to the tunnel exit. Secondly, the skewed wake causes an
uneven induction at the tunnel exit. The skewed wake induction at the blades causes
thelocalaxialthrustatthebladestobehigherbetweenazimuthangles1800and3600,
than over the rest of the rotorplane. This may be observed in Fig. 4.46 where a
predicteddistributionofthelocalbladeaxialthrust(indimensionlessform)isshown
asafunctionofbladeazimuthanglefortheTUDelftrotoratayawangleof300.This
distributionwaspredictedusinganaccelerationpotentialcodebyvanBussel[15].
Ideally,therotorshouldbeplacedwelldownstreamofthetunnelexitsuchthatitwill
not influence the jet velocity distribution at the tunnel exit. However, doing so will
causetheinflowattherotor tohaveahigherturbulencelevel.Thisisduetothefact
thatasthetunneljetexpandsdownstream,theturbulencelevelgenerallyincreases.In
thisstudytherotorwasplaced1mdownstreamofthetunnelexit.
Ineffect,theinfluencesoftunnelblockageduetowakeboundaryandthatduetothe
proximity of the rotor to the exit jet cannot be considered in isolation. In fact it is
possiblethatthecombinedactionofthetwomayreducetheuncertaintyintakingUjet
equal to U. One physical explanation would be the following: Referring to Fig. 4.42,
blockageduetowakeboundarymaycausewbtobehigherthanUifthejetexpansion
remainssmall.ThismaycausethelocalflowatthepitottobecomehigherthanU.On
theotherhand,tunnelblockageduetorotorproximitytothetunnelexitmaycausethe
localflowatthepitottobecomelessthanU.Inthiscasethecombinedactionwillhelp
inreducinguncertainty.However,itmayhappenthattheproximityoftherotortothe
jetmayalsohelptospeeduptheflowatthepitot.Asaresult,bothblockagetypeswill
increaseuncertaintylevels.
115
Chapter4 AerodynamicAnalysisoftheTUDelftModelTurbine
ua
U
wa
Pitot readings
Fig(a) AxialConditions
Ujet
U ua wa
Pitot readings Ujet
Fig(b) YawedConditions
Figure4.45Influenceofrotorontunnelexitjetvelocitydistribution.
00
2700 900
1800
Figure4.46Variationofthelocalaxialthrustcoefficientwithbladeazimuthanglefor=300forthe
TUDelftrotoraspredictedbyanaccelerationpotentialmethod.(Source:vanBussel[15]).
116
Chapter4 AerodynamicAnalysisoftheTUDelftModelTurbine
QuantificationofErrorintheComputedAxialInducedVelocityduetoErrorinUjet
As already outlined before, tunnel blockage makes it difficult to determine the true free
streamvelocity.Thetruefreestreamvelocityatanypointinthecrosssectionofthetunnel
tubemaydifferfromthatmeasuredusingthepitotstatictubesatthetunnelexitjet.Usinga
simple actuator disk model,it may be shown mathematically that very large errors in the
derivedaxialinducedvelocitymayresult whenthelatteriscalculatedontheassumption
thatthefreewindspeed(U)isequaltothatmeasuredbythepitotstatictubesatthetunnel
exit(Ujet):
Let kc be equal to the ratio of the exit jet velocity to the true free stream velocity (i.e.
kc=Ujet/U).Letwa,expbeequaltothemeasuredaxialflowvelocityattherotordisk.Letuabe
theaxialinducedvelocityattherotordiskderivedusingtheassumptionthatUjetisequal
thetruefreestreamvelocity(U).Letua*betheaxialinducedvelocityderivedusingthetrue
freestreamvelocity(U).Then
ua = wa ,exp U jet cos( ) (4.61a)
ua
*
= wa ,exp U cos( ) (4.61b)
U jet
= wa ,exp cos( ) (4.61c)
kc
ThepercentageerrorintroducedbyfindingtheaxialinducedvelocityusingUjetinsteadof
Uwillthenbeequalto:
ua ua *
kc = 100 (4.62)
ua *
SubstitutingEqt.4.61(c)inEqt.4.62yields
Since for a wind turbine the axial flow velocity decreases continuously downstream then
kc wa ,exp < U jet cos( ) where kc>0. If the tunnel blockage causes the local velocity at the
pitotprobestobelessthanthetruefreewindspeed,thenkc<1.Ontheotherhandiftunnel
blockage causes a speedup in the flow local to the probes, then kc>1. Fig. 4.47 plots the
variation of error kc with kc in accordance with Eqt. 4.63 for Ujet=5.5 m/s and for different
yaw angles and axial flow velocities (wa). The error in the axial induced velocity will be
larger at lower values of kc, larger yaw angles and at higher measured values of wa. Note
thatforkc=1,theerrorisequaltozerosinceforthisconditiontheexitjetvelocityisequalto
thetruefreestreamvelocity.ItcanbeeasilyobservedfromFig.4.47thatasmalldeviation
of exit jet velocity from the true freewind speed could possibly yield a large error in the
117
Chapter4 AerodynamicAnalysisoftheTUDelftModelTurbine
estimationoftheaxialinducedvelocity.Forinstanceatyaw450,a10%deviationofUjet(i.e.
kc=0.9) from the true freestream velocity may easily result in a 30% error in ua. This
providesevidencethatspecialcareshouldbetakenwhenassumingUjettobeequaltothe
freestream velocity and that it is crucial to know the freestream velocity accurately if
accurate prediction of the induced velocity is to be estimated. A very important note to
makeisthat,apartfromtunnelblockage,thenonuniformityinvelocitydistributionofthe
tunnelexitjet(showninFig.4.5,page49)alsocausethetruefreewindspeedtobedifferent
fromthatmeasuredbytheprobes.ItcouldbeeasilyobservedfromFig.4.5,thatakcvalue
of0.93canbeeasilypresentedduetotunnelexitjetnonuniformity.Ifadditionallytunnel
blockage is present such that it contributes to a kc value of 0.95, then the total effective kc
value would be equal to 0.93X0.95 = 0.884. From Fig. 4.47, it is seen that the later value
yieldsapercentageerrorinuaofaround2025%,whichisconsiderable.
100
50
kc (%)
0
=00 0deg,
Yaw , w wa=2
a=2m/sm/s
0.5 0.6 0.7 0.8 0.9 1 1.1 1.2
=30
Yaw 30 0deg,
, w wa=2
=2m/s
a
m/s
=45
Yaw 45 0deg,
, w wa=2
=2m/s
a
m/s
-50
=00 0deg,
Yaw , w wa=3
a=3m/sm/s
=30
Yaw 30 0deg,
, w wa=3
=3m/s
a
m/s
=45
Yaw
0, w =3m/s
45 deg, wa=3 m/s
a
-100
kc
Figure4.47Variationof kcwithkcfordifferentyawanglesandmeasuredinflowvelocities(wa,exp).
TheresultsarecalculatedforUjet=5.5m/ssinceduringallmeasurements,thetunnelspeedwasset
suchthatthepitotreadingsmeasuredthiswindspeed.
ProceduresforAssessingInfluencesduetoTunnelBlockage
Two separate computational procedures were adopted to assess the extent to which the
inflow measurements on the TUDelft rotor could have been affected by tunnel blockage.
The procedures are illustrated in Fig. 4.48. Both procedures make use of the prescribed
118
Chapter4 AerodynamicAnalysisoftheTUDelftModelTurbine
vortex model HAWT_PVC on the assumption that this model is reasonably accurate for
modellingthewakeinflowdistributionsforbothaxialandyawedconditions.Recallthatin
thiscodeaboundcirculation(whichmayvarywithbladeazimuthangle)isprescribedto
calculatetheinducedvelocitydistributionatagivenplane.
( )
B r ,
PrescribetoHAWT_PVC
R
Procedure1 Procedure2
( )
Calculateinducedvelocity ua , c r ,
Calculateinducedvelocity
( )
R Ya =3.5 cm
,
' r
attunnelexit, ua ,exit R
fromexperimentalinflowdataand
usingHAWT_PVC assumingUjet =UusingEqt.4.61(a)
(
Calculateinducedvelocity ua ,c r R ,
'
) Ya =3.5 cm
( )
usingHAWT_PVC
ke r ,
FindusingEqt.4.64
R
( )
a r R ,
FindusingEqt.4.65
(
ke r , )
Is<0.05?
R Yes Yes
(
a r R , )
Is<10%?
No
No
TunnelBlockageMinimal
MeasurementDataValidandUsable
TunnelBlockageSignificant
MeasurementDataInvalidandUnusable
Figure4.48Twoproceduresusedtoassesstheinflowmeasurementsforinfluencesduetotunnel
blockage.
119
Chapter4 AerodynamicAnalysisoftheTUDelftModelTurbine
Sinceinthiscode,themodelledvorticityisonlyduetotherotor,thenitmaybearguedthat
thecomputedinducedvelocitiesareequivalenttothosewhichwouldhavebeenobtained
shouldtherotorwasoperatinginanopenairenvironment(i.e.noblockage)withthesame
boundcirculationattheblades.
In both assessment procedures, the bound circulation derived directly from the inflow
measurementsusingHAWT_LFIMisprescribedtovortexcodeHAWT_PVC.InProcedure1,
the vortex code computes the tunnel exit axial velocity distribution, ua ,exit
'
( r R , ) . The
ratioofthisvelocitytothatmeasuredusingthepitotprobesisevaluatedusing
(
ua ,exit ' r , ) (4.64)
ke ( r
R )
, =
U jet
R
Inidealconditions(noblockage),thisratioshouldbeequaltozero.Butthiswouldnotbe
thecaseifblockageduetorotorproximitytothetunnelexitispresent(refertoFig.4.44).
Also tunnel nonuniformity may cause the velocity at some locations in the jet exit to be
higherthanUjet.Thiswouldleadtoahigherwakecirculationwhoseinductionisfeltmore
atthetunnelexit.Inthisstudy,itwasdecidedtolimitketo0.05.Ifitwashigher,thenthe
measurement data would be invalid and unusable or a correction to the data would be
required.
InProcedure2,theaxialinducedvelocitiesatthe3.5cmdownstreamplanearederivedfrom
theexperimentalmeasurementsusingtheassumptionthatthepitotUjetisequaltothefree
stream velocity (using Eqt. 4.61(a)). These are then compared with that computed by
HAWT_PVC for the same plane, ( )
ua ,c r , . The two induced velocities are compared
R
alongthedifferentpointsinthe3.5cmplaneandtherelativeerrorisfoundusing
(
ua ,c r , ) (
ua , c ' r , )
( )
R = R Ya =3.5 cm
a r R , =
Y 3.5 cm
'
(
ua , c r ,
R )
(4.65)
Ya = 3.5 cm
Assumingthattheresultsfor BderivedinaccordancewithFig.4.33andHAWT_PVCare
reasonablyaccurate,theerrorintheequationaboveshouldmainlybeinusingEqt.4.61(a)
tofind ua , c ( r R , ) .Consideringthatthevaluesforw areaccurate,itfollowsthatthe
a,exp
mainsourceoferrorin aisduetotheUjet(i.e.thepitotreading).Inidealconditions,this
should be zero. Note that in this assessment method, error a is based on the induced
velocitiesat3.5cmdownstreamandnotthoseattherotorplanetoreducetheuncertainties
duetothelinearinterpolationthatwasrequiredtoestimatewa,cattherotorplane(Eqt.4.7).
120
Chapter4 AerodynamicAnalysisoftheTUDelftModelTurbine
4.3.5QuantificationofErrorsinDerivingBladeLoadingduetoErrors
inInflowMeasurements
Asalreadyoutlinedbefore,inthisworkrelatedtotheTUDelftturbine,themainobjective
was to investigate an approach for deriving the aerodynamic loading distributions at the
blades from hotfilm inflow measurements in attached flow conditions. In this analysis, it
was therefore very important to investigate the uncertainty in the derived blade loading
resultingfromerrorsinwa,cwhichisdirectlyobtainedfromthehotfilmmeasurements.It
was also necessary to investigate how this uncertainty is influenced by other parameters
suchasthebladeradiallocationandpitchangle.
Considerabladeelementatagivenbladeazimuthpositionandyawangle.Assumethatthe
localangleofattackissmallanddoesnotexceedthestallangle.Neglectingtheeffectsof
unsteadyflowovertheblades,theliftcoefficientmaybeapproximatedtoCl=2.Thelocal
liftloadinginN/mmaythenbeexpressedas
dL
fL = = cV 2 (4.66)
dr
Taking wa,c as the axial flow velocity at the blade lifting line obtained from the
measurements using linear interpolation (see Eqt. 4.7 and Figs. 4.18(a), 4.19 and 4.20) and
substitutingEqts.4.11and4.12inEqt.4.66,
wa ,c
(
f L = c wa ,c 2 + r 2 2 tan 1 ) (4.67)
r
Relation tan
1
(w a ,c
r ) isequaltothelocalinflowangle() andthiscanbeexpandedin
seriesformasfollows
3 5 7
wa ,c wa ,c 1 wa ,c 1 wa ,c 1 wa ,c
= tan
1
= r 3 r + 5 r 7 r + .... (4.68)
r
Since for the TUDelft turbine, wa,c<<r at all radial locations, then we can neglect terms
(w r ) andhigher.ThusEqt.4.67becomes
3
a ,c
w2 w
f L cr 2 2 1 + r (4.69)
r
2 2
Neglecting the drag on the blades (this assumption is reasonable for 1<<12 deg for the
NACA0012 aerofoil), the axial thrust (N/m) and torque (Nm/m) loading at a given radial
locationonthebladesasequalto
dT
fT = = f L cos ( ) (4.70a)
dr
121
Chapter4 AerodynamicAnalysisoftheTUDelftModelTurbine
dQ
fQ = = r f L sin ( ) (4.70b)
dr
where istheinflowangle.Termscos()andsin()mayalsobeexpandedinseriesform
as
2 4
1 wa ,c 1 wa ,c
cos ( ) = 1 + 4! r ...
2! r
(4.71)
3 5
wa ,c 1 wa ,c 1 wa ,c
sin ( ) = + 5! r ...
r 3! r
(w r ) andhigher,theaxial
3
SubstitutingEqts.4.71inEqts.4.70andneglectingterms a ,c
thrustandtorqueloadingcanthenbeapproximatedto
wa ,c wa ,c 2
fT cr 2 2 (4.72)
2 2
r 2r
wa ,c
f Q cr wa ,c (4.73)
2
The percentage error in fT due to an error in measured wa,c at a given radial location on a
bladecouldbesimplyfoundfrom
fT wa ,c + wa ,c
fT
T =
wa ,c
100 (4.74)
fT wa ,c
Alternatively,forsmallvaluesinwa,conly,thepercentageerrorTisgivenby
fT w
T
a ,c
wa ,c (4.75)
wa ,c fT
where w isthepercentageerrorinwa,cgivenby
a ,c
wa ,c
w = 100 (4.76)
a ,c
wa ,c
FromEqts.4.72and4.75,itfollowsthat
1 wa ,c
r 2 2 w wa ,c
r a ,c
T (4.77)
wa ,c wa ,c 2
2 2
r 2r
122
Chapter4 AerodynamicAnalysisoftheTUDelftModelTurbine
For the condition in which the local blade pitch angle, , is equal to zero, it can be
concludedthat T wa,c .Thisimpliesthatwhenusingtheinflowmeasurementstofindthe
axial thrust loading, the error in the loading remains approximately equal to that in wa,c.
Thisonlyholdswhenthelocalpitchangleiszero.Forotherpitchangleshowever,Eqt.4.77
shows that for constant speed , the percentage error in the thrust loading, T , is
dependentonparameterswa,c,rand .Fig.4.49(a)showsthatvariationof T with w at
a ,c
differentvaluesof andwa,cwhilekeepingrconstant.Itisnotedthat,atagivenvalueof
w a ,c
,error T isverymuchdependentonthevaluesof andwa,c.
Amoreimportantobservationconcernsthefactthatinmanysituations,asmallerrorinwa,c
mayyieldasignificantlylargererrorinthederivedloading.Whenwa,c/rbecomescloseto
, T becomesverylarge,evenwhen w issmall.Infact,thisistheconditionwhenthe
a ,c
localangleofattackissmallandthusthethrustloading(fT)approacheszero.Thisisavery
important issue that needs serious consideration when deriving blade loads from inflow
measurements.
Therefore one can conclude that despite the fact that for attached flow conditions (low
anglesofattackattheblades),theuncertaintyintheaerofoildataissmall,theuncertainty
in the derived loading due to measurement errors in wa,c may be large. On the contrary,
whendealingwithhighanglesofattack,theuncertaintyduetowa,cissmall,butthatdueto
theaerofoildatamaybelarge(especiallywhendealingwithhighlystalledflows).
and speed () equal to those of the TUDelft rotor in the hotwire measurements (i.e.
=720rpm and tip=20). The results are shown for two values of wa,c (equal to 2 and 3m/s)
whichareontheorderofmagnitudeofthosederivedinthewindtunnel.Itcanbeeasily
noted that a small variation in wa,c has a significant effect on T and may cause it to be
the exact equation 4.74 and the approximate equation 4.77. But the results from the two
equationsveryfoundtobenearlyequal.
123
Chapter4 AerodynamicAnalysisoftheTUDelftModelTurbine
80
T
eT Exact, (=00, w wac=2m/s)
Exact, (q=0deg, =2m/s)
a,c
70
T Approx,
eT (=00, wa,c
Approx, (q=0deg, wac=2m/s)
=2m/s)
60 T Exact,
eT (=2 , wa,c
0
Exact, (q=2deg, wac=2m/s)
=2m/s)
40
T Approx,
eT (=40, wa,c
Approx, (q=4deg, =3m/s)
wac=3m/s)
30
20
10
0
0 2 4 6 8 10 12
wa,c (%)
Figure 4.49(a) Variation of error due to axial thrust loading against the error in wa,c. r/R=0.5,
=720rpm.
90
T Exact, (=0wac=2m/s
eT_Exact, 0, w =2m/s)
a,c
80
T Approx, (=0wac=2m/s
eT_Approx, 0, w =2m/s)
a,c
70
T Exact, (=0wac=3m/s
eT_Exact, 0, w =3m/s)
a,c
60 T Approx, (=0wac=3m/s
eT_Approx, 0, w =3m/s)
a,c
50
T (%)
40
30
20
10
0
0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
r/R
Figure 4.49(b) Spanwise variation of the error due to axial thrust loading for w =8%.
a ,c
=720rpm.Spanwisevariationof isequaltothatoftheTUDelftrotorwithtip=20.
124
Chapter4 AerodynamicAnalysisoftheTUDelftModelTurbine
ApplyingasimilarapproachtothetorqueloadingfQ,thepercentageerrorinfQduetoan
errorinmeasuredwa,catagivenradiallocationandbladecouldsimplybefoundfrom
fQ fQ
wa ,c + wa ,c
Q =
wa ,c
100 (4.78)
fQ
wa ,c
Alternatively,forsmallvaluesinwa,conly,thepercentageerrorTisgivenby
fQ w
Q
a ,c
wa ,c (4.79)
wa ,c fQ
FromEqts.4.73and4.79,itfollowsthat
2 wa ,c
r w
a ,c
Q (4.80)
wa ,c
r
Itcanbeeasilyobservedfromtheequationabovethat Q isalwayslargerthan w which
a ,c
concludesthattheerrorinthederivedtorqueloadingwouldalwaysbelargerthanthatin
theinflowmeasurements.Fortheconditioninwhichthelocalbladepitchangle, ,isequal
tozero,itcanbeconcludedthat Q 2 wa ,c .Thustheerrorisdoubled.Foragivenradial
locationandoperatingcondition,theerrorintorqueloadingdueto w isthereforehigher
a ,c
than that in the axial thrust loading. This result indicates more difficulties in maintaining
accuracy when deriving the rotor torque and power coefficients from the inflow
measurements.Figs.4.50(a)and(b)areplotssimilartoFigs.4.49(a)and(b)but Q isshown
instead. It can be observed that the values of Q are larger than those of T . In these
figures, the results for Q are computed both with the exact equation 4.78 and the
approximateequation4.80.Theresultsfromthetwoequationsdiffermarginally,especially
atlargervaluesof w .FromEqt.4.80,itcanbeshownthataswa,c/r approaches ,then
a ,c
Q becomesmuchlargerthan w a ,c
.
To summarise, from the above analysis it can be concluded that when deriving the axial
thrust and torque loading distributions from measured inflow values (wa,c), the
uncertainties in these derived distributions due to errors in the inflow measurements is
largestatlowanglesofattack.Foragiveninflowmeasurementerror,theresultingerrorin
the torque loading is larger than that for axial thrust. In a yawed rotor conditions, the
valuesofwa,catafixedwindspeedandrotorspeedaresmallerthanfornonyawed
125
Chapter4 AerodynamicAnalysisoftheTUDelftModelTurbine
100
90 Q
eQ Exact, (=00, w wac=2m/s)
Exact, (q=0deg, =2m/s)
a,c
eQQ Approx, (=00, wa,c
Approx, (q=0deg, wac=2m/s)
=2m/s)
80
Q Exact,
eQ (=20, wa,c
Exact, (q=2deg, wac=2m/s)
=2m/s)
70
Q Approx,
eQ (=20, wa,c
Approx, (q=2deg, wac=2m/s)
=2m/s)
60 Q Exact,
eQ (=4 , wa,c
0
Exact, (q=4deg, wac=3)
=3m/s)
40
30
20
10
0
0 2 4 6 8 10 12
wa,c (%)
Figure 4.50(a) Variation of error due to torque loading against the error in wa,c. r/R=0.5,
=720rpm.
120
eQ_Exact,
Q Exact, (=0 , wa,c=2m/s)
wac=2m/s
0
100 eQ_Approx,
Q Approx, (=0
0, w =2m/s)
wac=2m/s
a,c
eQ_Exact,
Q Exact, (=0
0, w =3m/s)
wac=3m/s
a,c
80 eQ_Approx,
Q Approx, (=0
0, w =3m/s)
wac=3m/s
a,c
Q (%)
60
40
20
0
0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
r/R
Figure 4.50(b) Spanwise variation of the error due to axial thrust loading for w =8%.
a ,c
=720rpm.Spanwisevariationof isequaltothatoftheTUDelftrotorwithtip=20.
126
Chapter4 AerodynamicAnalysisoftheTUDelftModelTurbine
conditions. This results in a smaller angle of attack at some blade azimuth positions and
hence the uncertainties in the derived loading due to errors in the inflow measurements
wouldbelarger.Forthisreason,itismoredifficulttoderivetheloadingdistributionsfor
inflowmeasurementswhentreatingyawedconditions.Followingthisanalysis,onemight
askwhythehotfilmmeasurementscarriedoutinthisstudywerenotaccomplishedwitha
lowerrotortipspeedratio(insteadof =8)oratasmallertippitchangle(insteadof tip=20)
soastoincreasethelocalanglesofattackatthebladesandhencereducetheuncertainties
inthederivedloadingduetothemeasurementerrors.Thereasonsfornotdoingsoaretwo:
firstofalllargeranglesofattackcouldeasilyresultinstallinsomeradiallocationsonthe
blades.Thiswouldgeneratehighturbulencelevelsinthenearwakewithwouldotherwise
increasetheerrorininflowmeasurementsthemselves.Secondly,itisverydifficulttomodel
theaerofoildataaccuratelyinstalledconditions,especiallywhentherotorisyawed.
127
Chapter4 AerodynamicAnalysisoftheTUDelftModelTurbine
4.3.6ResultsandDiscussion
This section presents the results for aerodynamic flow parameters and blade loads that
werederivedfortheTUDelftrotorusingHAWT_LFIM.Thewindtunnelinflowvaluesfor
wa,c at the rotorplane estimated directly from the hotfilm measurements by linear
interpolation, (see Figs. 4.18(a), 4.19 and 4.20) were used in conjunction with the unsteady
aerofoil model (described in section 4.3.2) in accordance with the procedure described in
section 4.3.1. The parameters were computed for different azimuth positions (at 150
increments)overonewholerevolutionoftherotor.
A.UncertaintyAnalysis
Thefirstorderanalysisofsection4.3.5hasshownthattheuncertaintyinthederivedblade
loading is very sensitive to the uncertainty in the inflow measurements used. In section
4.2.2 (page 67), the sources of error in the hotfilm measurements were outlined and the
uncertainty in these measurements was estimated to be in the orderof 610% for =00. In
the calculations using HAWT_LFIM, it was necessary to introduce a more elaborate
uncertainty analysis to be able to quantify the resulting errors in each derived parameter
(e.g. the derived angle of attack or aerodynamic thrust loading) due to errors in the
measureddata.Alsothepossiblepresenceofsignificanttunnelblockagecouldonlyaddto
theuncertaintyinthedata.Tunnelblockageeffectswerehoweverassessedand,asitwillbe
explainedlateron,theywerefoundtobenegligible.
UncertaintyAnalysisinAxialConditions
Foraxialconditions,anuncertaintyanalysiswascarriedoutbyderivingtheparametersat
thedifferentbladeazimuthanglesoveronewholerevolution(at150increments)takinginto
account the deviations in wa,c. The mean value and standard deviation of each parameter
acrossthewholerevolutionwerethenestimatedforthedifferentradiallocations(40,50,60,
70,80and90%R).Thestandarddeviationwastakenasameasureoftheuncertaintydueto
measurementerrors.
UncertaintyAnalysisinYawedConditions
Foryawedconditions,theuncertaintyanalysisismuchmoredifficulttoaccomplishthanin
axial conditions because the parameters are known to be a function of . However, as
alreadyremarkedinsection4.2.2(page68),itisreasonabletoassumethattheuncertainty
inwa,cforyawedconditionswouldbeequaltothatin=00,i.e.intherangeof610%.
ForthecalculationswithHAWT_LFIM,anuncertaintyof0.26m/sand0.21m/sinwa,cwas
assumedat =300and450,respectively.Thisisequivalenttoabout8%oftheazimuthally
averaged value of wa,c at each radial location. The calculations were performed for three
128
Chapter4 AerodynamicAnalysisoftheTUDelftModelTurbine
differentsetsofwa,cattherotorplane:(1)withthevaluesofwa,cshowninFigs.4.19and4.20;
(2)sameas(1)butwitheachwa,cvalueincreasedby0.26m/sand0.21m/sat =300and450 ,
respectively; (3) same as (1) but with each wa,c value decreased by 0.26m/s and 0.21m/s at
=300and450,respectively.
B.HAWT_LFIMResultsforAxialConditions
Thissectionpresentstheresultsthatwereobtainedforthebladeaerodynamicparameters
inaxialconditions(i.e.=00).Themeanvaluesobtainedoveronewholerevolutionare
plottedtogetherwitherrorbarsdenotingthecorrespondingonestandarddeviationfrom
themean.
B.1Derivationofspanwisedistributionsofangleofattack,flowrelativevelocityand
boundcirculation
Fig. 4.51 illustrates the angle of attack () distribution obtained for axial conditions using
Eqt.4.11computedatradiallocations40,50,60,70,80and90%R.Themeanangleofattack
is small at each radial location and is much smaller that the stalling angle for the NACA
0012 aerofoil (which is equal to 110 at a Reynolds number of 150,000). This provides
evidence that the flow over the blades is fully attached. The corresponding flow relative
velocity (Vr) distribution computed using Eqt. 4.12 is shown in Fig. 4.52. For Vr, the
standarddeviationwasfoundtobenegligibleduetothefactthatforthisstudyr>>Uatall
radiallocations.
TheangleofattackvaluesofFig.4.51wereusedintheunsteadyaerofoilmodelpresented
insection4.3.2.Foraxialconditions,theangleofattackwasalmostconstantwithblade
5
(deg)
0
0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
r/R
Figure4.51Distributionofangleofattack(U=5.5m/s,=700rpm,tip=20).
129
Chapter4 AerodynamicAnalysisoftheTUDelftModelTurbine
azimuthangle,theslightfluctuationsmainlybeingduetothenonuniformityinthetunnel
exit jet. The lift coefficient could be approximated to the incompressible equation for
attached flow: Cl=2. The bound circulations at 40, 50, 60, 70, 80 and 90%R were then
foundfromtheKuttaJoukowskilawEqt.4.13.TheseresultsareplottedinFig.4.53where
theyarereferredbycurveYa=0cm.
Asalreadyoutlinedbefore,theseresultswerecalculatedusingthevaluesforwa,cestimated
at the rotorplane using linear interpolation of those at Ya=6cm and at Ya=3.5cm. To
investigate the sensitivity of B,2D due to changes in wa,c, the same calculations were also
performed using the measured wa,c values at Ya=3.5cm. The resulting B,2D values are also
includedinFig.4.53wheretheyarereferredtobyYa=3.5cm.Considerabledifferences
45
40
35
30
25
Vr (m/s)
20
15
10
0
0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
r/R
Figure4.52Distributionofflowvelocityrelativetomovingblades(U=5.5m/s,=700rpm,tip=20).
0.7
0.6
0.5
,2D (m /s)
0.4
2
0.3
Ya=0 cm
0.2
Ya=3.5 cm
0.1
0
0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
r/R
Figure 4.53 Distribution of bound circulation distribution estimated using 2D lift coefficient
(U=5.5m/s,=700rpm,tip=20).
130
Chapter4 AerodynamicAnalysisoftheTUDelftModelTurbine
were found, especially towards the outboard sections. This indicates the importance of
estimating the inflow at the rotorplane using both upstream and downstream
measurements.Fortheoperatingconditionbeingtested,thechangein B,2Dresultingfrom
theuseofwa,cvaluesatYa=3.5cminsteadofthoseestimatedforYa=0cmisnotmainlydueto
changesinVr(sincer>>wa,c),butduetotheangleofattackthatissensitivetochangesin
wa,c.
B.2ExtrapolatedBoundCirculationDistributionCorrectedforTipandRootLoss
Amajordifficultyencounteredwastoderiveanextrapolatedboundcirculationdistribution
acrossthewholebladespan(from30%to100%R)fromthepointvaluesat40,50,60,70,80
and 90%R. The derived bound circulation at locations 40% and 90%R was found to be
rather unrealistically high (see Fig. 4.53), even though the angle of attack was derived
directlyfromtheinflowmeasurements.Thisisbecausethesetwolocationsareverycloseto
thebladetipandrootandthehighly3Dnatureoftheflowherewillcausetheliftcoefficient
to be less than 2 . This phenomenon has been observed by Johansen et al. [40] when
deriving lift and drag coefficients from detailed CFD computations and by Tangler [90]
whenderivingaerofoildatafrompressuremeasurementsonarotatingturbinebladeusing
avortexmodel.ThesamewasobservedinthisprojectwhenanalyzingtheNRELPhaseVI
rotor(refertoChapter6,Fig.6.19(a)).
To derive an extrapolated bound circulation distribution that accounts for tip/root loss
(B,3D),twodifferentmethodswereused:
In method 1, the estimated values of the bound circulation at 40% and 90%R were
discarded and a cubic variation of bound circulation was prescribed between 30% and
40%R and between 80% and 100%R. The bound circulation at 30% and 100%R was set to
zero. A spline interpolation was then applied to obtain a continuous bound circulation
distributionacrossthewholebladeusingthetechniquedescribedinAppendixD.
Inmethod2,atrialanderrorapproachwasusedinconjunctionwiththeprescribedwake
vortex model HAWT_PVC described in section 4.3.3. In this method, the axial induced
velocityattheblades(ua,c)atYa=3.5cmwasinitiallyfoundbyassumingthatthefreestream
velocitywasequaltotheidealfreewindspeedof5.5m/s(i.e.usingresultsforwa,c inFigs.
4.18(a)inEqt.4.57).Thiswasjustifiedsince,asitwillbeexplainedlateroninsectionB.3,
blockage effects for axial conditions were found to be small. Then, a large number of
different spanwise distributions for bound circulations were assumed and each was
prescribed to HAWT_PVC to calculate the spanwise distributions ua,c at Ya=3.5cm. Using a
trialanderroralgorithmembeddedinthecode,theprescribedvortexmodelHAWT_PVC
determinedtheboundcirculationthatyieldedaninducedvelocityvariationthatisclosest
tothatderiveddirectlyfrommeasurementsatYa=3.5cm.Thevortexmodelparameterswere
set as shown in table 4.4. Parameters Rt,w1, Rt,w2 and p were obtained from the smoke
visualizationexperiments(refertosection4.2.3).Thecalculationswerecarriedoutforthree
131
Chapter4 AerodynamicAnalysisoftheTUDelftModelTurbine
different values of n: n=15, 21 and 31. The modelled helical wake geometry used for the
vortex model is depicted in Fig. 4.54. Fig. 4.55 shows the derived distributions for B,3D
obtained using the two methods. For method 2, the distributions for B,3D obtained for the
three different values of n are shown. In fact, the difference between the resulting
distributionsisnegligible.TheinducedvelocitiesatYa=0.035cmforsuchdistributionsusing
HAWT_PVC are shown in Fig. 4.56. These agree very well with those obtained for the
measurementsusingEqt.4.57.ItisobservedinFig.4.55thattheresults frommethods1and
2 agree very well, thus providing confidence in the derived bound circulation along the
blades. It should be noted that in method 2, no knowledge of aerofoil data for the lift
coefficientisused.
Table4.4:ParametersusedfordeterminingtheboundcirculationusingHAWT_PVCfor=00.
nwRev 10
tot 36
100
Rt,w1 0.6398
Rt,w2 0.668
p 0.37
ReleaseTip 0.99
ReleaseRoot 0.305
0.5mm
Figure4.54Wakegeometryfordeterminingboundcirculationdistributionatbladesusing
prescribedvortexmodel(HAWT_PVC)inmethod2.(n=21).
132
Chapter4 AerodynamicAnalysisoftheTUDelftModelTurbine
0.7
0.6
0.5
B,3D (m /s)
2 0.4
0.3
0.2 Method 1
Method 2, n = 15
0.1
Method 2, n = 21
Method 2, n = 31
0
0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
r/R
Figure4.55Derivedboundcirculationdistributionsusingmethods1and2.
0.5
0.0
0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
-0.5
ua,c (m/s)
-1.0
-1.5
Exp
-2.0
Method 2, n = 15
Method 2, n = 21
Method 2, n = 31
-2.5
r/R
Figure4.56ComparisonofinducedvelocitiesatbladesatYa=3.5cmcomputedusingmethod2
withthatderivedfromexperimentalmeasurements(Exp)usingEqt.4.57.Theerrorbarsinplot
Expdenotetheonestandarddeviationsduetotheuncertaintyinthehotfilmmeasurements.
133
Chapter4 AerodynamicAnalysisoftheTUDelftModelTurbine
B.3AssessmentoftheEffectsofTunnelBlockageinAxialConditions
Toassessforblockagethetwoproceduresdescribedinsection4.3.4,Fig.4.48wereapplied.
Inbothcases,themeanboundcirculationderivedusingmethod1(refertosectionB.2)was
prescribedtovortexmodelHAWT_PVCtogetherwiththeparametersoftable4.4.
Inprocedure1,HAWT_PVCcalculatedtheaxialinducedvelocitiesinlinewiththebladesbut
at the tunnel exit (i.e. finding ua,c at Y=1m) and the factor ke were evaluated at all radial
locations(usingEqt.4.64).TheresultsarepresentedinFigs.4.57(a)and(b).Notethatsince
we are dealing with axial conditions, the calculations were carried out at only one rotor
azimuth angle (at =00). It can be observed that the induced velocity at the tunnel exit is
verysmall.ThehighestvalueforkebeingpredictedbyHAWT_PVCisonly0.029atr/R=0.3
(i.e.2.9%ofUjet).kedecreasescontinuouslyto0.022atthebladetip.Thesevaluesforkeare
very small and are only on the order of magnitude of the standard deviations caused by
tunnel jet nonuniformity. This provides evidence that tunnel blockage due to rotor
proximitytothetunnelexitcanbeneglected.
Toassessfortunnelblockageusingprocedure2(seesection4.3.4,Fig.4.48),theaxialinduced
velocities at Y=3.5cm were initially found by assuming that the freestream velocity was
equal to the ideal freewindspeed of 5.5 m/s (i.e. using results for wa,c of Fig. 4.18 in Eqt.
4.57). Actually these are the same values as those of Fig. 4.56 (referred by Exp).
HAWT_PVCthencalculatedtheinducedvelocityua,catY=3.5cmresultingfromthebound
circulation estimated using method 1. The two induced velocities were compared and the
percentagediscrepancy(a)wasfoundinaccordancewithEqt.4.65.Theresultsareshown
in Figs. 4.58(a) and (b). The induced velocities derived using the two different methods
agree very well and this implies that it is justified to assume that Ujet = U when applying
Eqt.4.57.Thusitmaybearguedthatblockageeffectsaresmall.However,oneshouldkeep
in mind that in the blockage assessment, it is being assumed that the vortex model
HAWT_PVCissuitablyaccurateincalculatingtheinductionforagivenboundcirculation.
134
Chapter4 AerodynamicAnalysisoftheTUDelftModelTurbine
0
0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
-0.02
-0.04
-0.08
-0.1
-0.12
-0.14
-0.16
-0.18
r/R
Figure4.57(a)Spanwisevariationofua,ccalculatedatthetunnelexitusingHAWT_PVC(n=21).
0.035
0.03
0.025
0.02
ke
0.015
0.01
0.005
0
0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
r/R
Figure4.57(b)SpanwisevariationofkecalculatedusingHAWT_PVC(n=21).
135
Chapter4 AerodynamicAnalysisoftheTUDelftModelTurbine
0.5
0.0
0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
-1.0
-1.5
Exp
-2.0
HAWT_PVC, n = 15
HAWT_PVC, n = 21
HAWT_PVC, n = 31
-2.5
r/R
Figure 4.58(a) Comparison of axial induced velocities at 3.5cm downstream of rotorplane
calculatedusingmeasurementsandassumingUjet=UwiththosecalculatedfromHAWT_PVC.
15
10
5
a(%)
0
0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
-5
-10
-15
r/R
Figure4.58(b)SpanwisevariationofpercentagediscrepancycalculatedusingEqt.4.65.
136
Chapter4 AerodynamicAnalysisoftheTUDelftModelTurbine
B.4AxialInductionFactorDistributions
Since the blockage effects were found to be small, then U can be taken to be equal to Ujet
(5.5m/s).Theaxialinductiondistributionsforua,cateachmeasuringplane(Ya=6,3.5,6and
9cm) could be found using the measured values of wa,c in Fig. 4.18(a) and Eqt. 4.57. The
azimuthally averaged induced velocities (i.e. the average axial induced velocity over an
annulus) at each radial location for a given rotor azimuth angle could be computed as
follows:By knowing the axial flow velocity wa ateach probe position (denoted by indices
(i,j)) and rotor azimuth angle (denoted by index ) (refer to section 4.2.2, page 58), the
azimuthallyaveragedaxialflowvelocityatagivenradiallocationandrotorazimuthangle
couldbefoundfrom:
j = jtot
1
( wa ,aver )i , =
jtot
( w )
j =0
a ,i , j
(4.81)
Theazimuthallyaveragedaxialinducedvelocitiescouldthenbefoundfrom
ua = wa ,aver U (4.82)
Theresultsforua,canduaatthedifferentmeasuringplanesareshowninFigs.4.59(a)and
(b).Theaxialinductionfactorsattherotorplane(a1,canda1)werethenbefoundbydividing
the interpolated induced velocities at the rotorplane by 5.5m/s. The results are shown in
Figs. 4.60(a) and (b). The bars indicated the +/one standard deviations in the data across
onewholerevolution.
It can be observed from Figs. 4.60 and 4.61 that the standard deviations in a1 are smaller
than those in a1,c. This is a result of the fact that averaging tends to damp out any
fluctuationsinvelocitythattendtooccurlocally.
137
Chapter4 AerodynamicAnalysisoftheTUDelftModelTurbine
0.3
0.0
0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
-0.3
-0.6
-0.9
ua,c (m/s)
-1.2
3.5cm DwnStream
-1.8
6cm DwnStream
-2.1
9cm DwnStream
-2.4
RotorPlane
(Interpolated)
-2.7
r/R
Figure4.59(a)Spanwisevariationofaxialinducedvelocityatbladeliftingline.
0.3
0.0
0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
-0.3
-0.6
-0.9
ua (m/s)
-1.2
3.5cm DwnStream
-1.8
6cm DwnStream
-2.1
9cm DwnStream
-2.4
RotorPlane
(Interpolated)
-2.7
r/R
Figure4.59(b)Spanwisevariationofazimuthallyaveragedaxialinducedvelocityatrotorplane.
138
Chapter4 AerodynamicAnalysisoftheTUDelftModelTurbine
0.00
0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
-0.05
-0.10
-0.15
a1,c
-0.20
-0.25
-0.30
-0.35
r/R
Figure4.60Spanwisevariationofaxialinductionfactoratbladeliftingline.
0.00
0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
-0.05
-0.10
-0.15
a1
-0.20
-0.25
-0.30
-0.35
r/R
Figure4.61Spanwisevariationofazimuthallyaveragedaxialinductionfactoratrotorplane.
139
Chapter4 AerodynamicAnalysisoftheTUDelftModelTurbine
B.5BladeLoadDistributions
Usingthedistributionfor B,3Dcalculatedusingmethod1(seesectionB.2,page131),thelift
coefficientcorrectedfortip/rootloss(Cl,3D)wascalculatedfromEqt.4.14atradiallocations
40,50,60,70,80and90%R.Thedragcoefficientwasfoundfromthe2DdatafortheNACA
0012 aerofoil. The drag coefficient was not corrected for tip/root loss. The error incurred
wassmall,sincethedragcoefficientwasverysmall.TheresultsareshowninFigs.4.62and
4.63.UsingHAWT_LFIM,thebladechordwiseandnormalloadingvalues(dAanddA)at
40,50,60,70,80and90%RwerefoundusingthebladeelementtheoryequationsEqts.3.16
andthelift/dragcoefficientsofFigs.4.62and4.63.Eachrotorbladewasdiscretizedinto22
equallyspacedelements.Theloadingvaluesatthebladetipandrootweresettozeroanda
splineinterpolation(asinAppendixD)wasappliedtoestimatetheloadingvaluesateach
of the 22 blade sections. The results are displayed in Figs. 4.64 and 4.65. The blade thrust
and torque loading distributions (dT3D and dQ3D) were estimated using a similar method,
however using Eqts. 3.20 (refer to Figs. 4.66 and 4.67). In Figs. 4.62 4.67, the error bars
denote the one standard deviations resulting from the uncertainties in the inflow
measurements. It was found that the maximum errors in dT3D and dQ3D due to the
uncertaintiesintheinflowmeasurements(wa,c)wereequalto7.4%and13.6%.Recallfrom
section 4.2.2 (page 68) that the uncertainty in wa,c was found to be in the range 610% at
=00.ItcanthereforebeconcludedthattheerrorindT3Dremainedinthesameorderofthat
oftheinflowmeasurementsfromwhichitwasderived.HowevertheerrorindQ3Dishigher
than that for dT3D and this consolidates what was found earlier in section 4.3.5 that
Q > T .
0.8
0.7
0.6
0.5
Cl,3D
0.4
0.3
0.2
0.1
0
0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
r/R
Figure4.62Distributionofliftcoefficient(correctedfortip/rootloss).
140
Chapter4 AerodynamicAnalysisoftheTUDelftModelTurbine
0.025
0.02
0.015
CD,2D
0.01
0.005
0
0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
r/R
Figure4.63Distributionofdragcoefficient(notcorrectedfortip/rootloss).
1.6
1.4
1.2
1
dA (N/m)
0.8
0.6
0.4
0.2
0
0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
r/R
Figure4.64Distributionofchordwiseaerodynamicloading.
141
Chapter4 AerodynamicAnalysisoftheTUDelftModelTurbine
30
25
20
dA (N/m)
15
10
0
0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
r/R
Figure4.65Distributionofnormalaerodynamicloading.
30
25
20
dT3D (N/m)
15
10
0
0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
r/R
Figure4.66Distributionofaxialthrustaerodynamicloading.
142
Chapter4 AerodynamicAnalysisoftheTUDelftModelTurbine
1.2
0.8
0.4
0.2
0
0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
r/R
Figure4.67Distributionoftorqueaerodynamicloading.
B.6ComparisonofRotorPowerandAxialThrustCoefficients
HAWT_LFIMintegratedtheloadingdistributionsofFigs.4.64and4.65usingtheprocedure
ofAppendixBtodeterminetheaxialthrustandpowercoefficientsfortherotorfor =8and
tip=20.Thesewerecomparedwiththosemeasuredduringthewindtunnelexperimentsby
means of strain gauges (refer to Figs. 4.2, 4.3 and 4.26). The comparisonis shown in table
4.5.Ideallythetworeadingsshouldbethesame.Thepercentagediscrepancyismainlydue
to various sources of error associated with the hotfilm measurements. Another source of
erroristheuncertaintyinthetip/losscorrectionused.Thetip/rootcorrectionappliedusing
the cubic extrapolation and using HAWT_PVC (methods 1 and 2, refer to section B.2) may
underestimatethethrustloadingatipandrootregion,thusresultinginalowervaluefor
CT. One should also remark that the lifting line model used in HAWT_LFIM and
HAWT_PVCisa2Dflowmodelandisratherlimitedinrepresentingthe3Deffectsatthe
rectangularbladetipandrootregions.
Table4.5:ComparisonofaxialthrustandpowercoefficientsderivedbyHAWT_LFIMusinghotfilm
nearwakemeasurementswiththosemeasuredusingstraingaugetechniques.
143
Chapter4 AerodynamicAnalysisoftheTUDelftModelTurbine
C.HAWT_LFIMResultsforYawedConditions
Thissectionpresentstheresultsthatwereobtainedforthebladeaerodynamicparameters
inyawedconditions(=300and450)usingHAWT_LFIM.Sinceinyawtheparametersare
unsteady,theresultsareplottedasafunctionofbladeorrotorazimuthangle()overone
whole revolution. The error bars in the plotted results denote the uncertainty bounds
resulting from the estimated errors of 0.26m/s and 0.21m/s in wa,c at =300 and 450,
respectively.Theseerrorsareapproximatelyequivalentto8%oftheazimuthallyaveraged
valuesofwa,c.
C.1 Derivation of spanwise distributions of angle of attack, flow relative velocity and
boundcirculation
Figs.4.68and4.69showthevariationsofthelocalangleofattack()withbladeazimuth
angle()derivedusingEqt.4.11foryawangles300and450.Thevariationsareperiodicand
the inboard blade sections experience the highest variations of both the mean and cyclic
componentsoftheangleofattack.Themaximumangleofattackisabout100andoccursat
=00,r/R=0.4and =300.Thisisslightlylessthanthe2D stallinganglefortheNACA0012
aerofoil which is about 110 at a Reynolds number of 150,000. Thus, at these yawed
conditionstheflowoverthebladesisfullyattached.Theuncertaintyinresultingfromthe
assumederrorsinwa,cisintherangeof0.80atr/R=0.4to0.40atr/R=0.9.Adiscontinuity
in the variation of with is observed between 3300<<3600 and 00<<300 at almost all
radial locations. This results from the discontinuity observed in the inflow measurements
which was already mentioned earlier in section 4.2.2, page 67 (see Figs. 4.19 and 4.20).
Consequently this discontinuity has resulted in a discontinuity of the other parameters
derivedfromtheinflowmeasurements(e.g.liftanddragcoefficients,boundcirculationand
aerodynamic loading), as will be noted later on. As already described, this behaviour
between3300<<3600and00<<300 ismainlyduetothenonuniformityinthetunnelexitjet
andpossiblyalsofrominfluencesfromthecomplexcirculationinskewedflowcausedby
theyawedrotor.Itcouldalsobethecasethattheinfluencesfromthecentrebodystructure
contributetosuchabehaviour.
FromtheangleofattackvariationsofFigs.4.68and4.69,itwaspossibletocalculatetherate
of change of the local angle of attack ( & ) with time as a function of using the inverse
AdamBashfortmethoddescribedinsection4.3.2(page95).TheresultsareplottedinFigs.
4.70 and 4.71. In general, & is negative for 00<<1800 and positive for 1800<<3600. The
highest rates of change of angle of attack are observed at the inboard blade sections and
consequentlyhigherlocalunsteadyaerodynamiceffectsareexperiencedhere.
Figs. 4.72 and 4.73 show the variations of the local flow relative velocities (Vr) with blade
azimuthangle()usingEqt.4.12.Asinaxialconditions,theuncertaintyinthisparameter
duetoerrorsinthemeasuredflowvelocities(wa,c)isverysmallbecauser >>wa,c.
144
Chapter4 AerodynamicAnalysisoftheTUDelftModelTurbine
12 5
10 r/R = 0.4
r/R = 0.7
4
8
3
(deg)
(deg)
6
2
4
1
2
0 0
0 30 60 90 120 150 180 210 240 270 300 330 360 0 30 60 90 120 150 180 210 240 270 300 330 360
( deg ) ( deg )
8 5
7 r/R = 0.5
4
6
5
3
(deg)
(deg)
4
2
3
r/R = 0.8
2
1
1
0
0
0 30 60 90 120 150 180 210 240 270 300 330 360 0 30 60 90 120 150 180 210 240 270 300 330 360
( deg ) ( deg )
6 5
5 r/R = 0.6
4
3
(deg)
(deg)
3
2 2
r/R = 0.9
1
1
0 0
0 30 60 90 120 150 180 210 240 270 300 330 360 0 30 60 90 120 150 180 210 240 270 300 330 360
( deg ) ( deg )
Figure4.68Variationofangleofattackwithbladeazimuthangleat=300.
145
Chapter4 AerodynamicAnalysisoftheTUDelftModelTurbine
9 4
8
r/R = 0.4
3.5
r/R = 0.7
7
3
6
2.5
5
(deg)
(deg)
2
4
1.5
3
1
2
0.5
1
(deg)
4 2
(deg)
3 1.5
2 1
r/R = 0.8
1 0.5
0 0
0 30 60 90 120 150 180 210 240 270 300 330 360 0 30 60 90 120 150 180 210 240 270 300 330 360
( deg ) ( deg )
4.5 3.5
4 r/R = 0.6
3
3.5
3 2.5
(deg)
2.5 2
(deg)
2
1.5
1.5 r/R = 0.9
1
1
0.5 0.5
0
0 30 60 90 120 150 180 210 240 270 300 330 360 0
0 30 60 90 120 150 180 210 240 270 300 330 360
( deg )
( deg )
Figure4.69Variationofangleofattackwithbladeazimuthangleat=450.
146
Chapter4 AerodynamicAnalysisoftheTUDelftModelTurbine
200 200
150 150
r/R = 0.7
100 100
50 50
dot deg/s
(deg/s)
(deg/s)
dotdeg/s
0 0
0 30 60 90 120 150 180 210 240 270 300 330 360 0 30 60 90 120 150 180 210 240 270 300 330 360
-50 -50
-100 -100
-200 -200
( deg ) ( deg )
200 200
150 150
100 100
r/R = 0.8
50
50
dot deg/s
(deg/s)
(deg/s)
dotdeg/s
0 0
0 30 60 90 120 150 180 210 240 270 300 330 360 0 30 60 90 120 150 180 210 240 270 300 330 360
-50 -50
-100 -100
r/R = 0.5
-150 -150
-200 -200
( deg ) ( deg )
200 200
100 100
50 50
(deg/s)
(deg/s)
dotdeg/s
dotdeg/s
0
0 30 60 90 120 150 180 210 240 270 300 330 360
0
0 30 60 90 120 150 180 210 240 270 300 330 360
-50 -50
-100 -100
r/R = 0.6
-150 -150
-200 -200
( deg ) ( deg )
Figure4.70Variationoftimerateofchangeofangleofattackwithbladeazimuthangleat=300.
147
Chapter4 AerodynamicAnalysisoftheTUDelftModelTurbine
200
200
150
100 100
50
(deg/s)
(deg/s)
dotdeg/s
dotdeg/s
0 0 30 60 90 120 150 180 210 240 270 300 330 360 0
0 30 60 90 120 150 180 210 240 270 300 330 360
-100 -50
r/R = 0.4 -100
r/R = 0.7
-200
-150
-300 -200
( deg ) ( deg )
200 200
150 150
100
100
50 50
(deg/s)
(deg/s)
dotdeg/s
dotdeg/s
0 0
0 30 60 90 120 150 180 210 240 270 300 330 360 0 30 60 90 120 150 180 210 240 270 300 330 360
-50 -50
-200 -200
( deg ) ( deg )
200 200
150 150
100 100
50
50
(deg/s)
(deg/s)
dotdeg/s
dotdeg/s
0 0
0 30 60 90 120 150 180 210 240 270 300 330 360 0 30 60 90 120 150 180 210 240 270 300 330 360
-50 -50
-100 r/R = 0.6 -100 r/R = 0.9
-150 -150
-200
-200 ( deg ) ( deg )
Figure4.71Variationoftimerateofchangeofangleofattackwithbladeazimuthangleat=450.
148
Chapter4 AerodynamicAnalysisoftheTUDelftModelTurbine
45
40
35
30
25
Vr (m/s)
20
15
r/R = 0.4
r/R = 0.5
10
r/R = 0.6
r/R = 0.7
5
r/R = 0.8
r/R = 0.9
0
0 30 60 90 120 150 180 210 240 270 300 330 360
( deg )
Figure4.72Variationofflowrelativevelocitywithbladeazimuthangleat=300.
50
45
40
35
30
Vr (m/s)
25
20
15 r/R = 0.4
r/R = 0.5
10 r/R = 0.6
r/R = 0.7
5 r/R = 0.8
r/R = 0.9
0
0 30 60 90 120 150 180 210 240 270 300 330 360
( deg )
Figure4.73Variationofflowrelativevelocitywithbladeazimuthangleat=450.
149
Chapter4 AerodynamicAnalysisoftheTUDelftModelTurbine
Figs. 4.74 and 4.75 plot the resulting spanwise bound circulation distributions (B,2D) for
different blade azimuth angles (00, 900, 1800 and 2700) as derived in accordance with the
KuttaJoukowski law (Eqt. 4.13). The error bars for this parameter due to the estimated
errorsinwa,carenotincludedintheseplotsforthesakeofclarity.Buttheerrorin B,2Dwas
found to be on the order of 0.071 and 0.055m2/s at yaw angles 300 and 450 respectively.
Unlikeaxialconditions,theboundcirculationisafunctionofthebladeazimuthangleand
isthereforeunsteady.ComparingFigs.4.53,4.74and4.75,itmaybenotedthatthebound
circulation level decreases as the yaw angle is increased, but the unsteadiness actually
increases.
0.7
0.6
0.5
B,2D (m /s)
0.4
2
0.3
0.2
== 000
== 90
900
0.1 == 180
1800
== 270
270
0
0
0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
r/R
Figure4.74Variationofboundcirculationwithbladeazimuthangleat=300.
150
Chapter4 AerodynamicAnalysisoftheTUDelftModelTurbine
0.5
0.45
0.4
0.35
0.3
B,2D (m /s)
2
0.25
0.2
0.15 == 000
0.1 == 90
900
0.05
== 180
1800
== 270
2700
0
0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
r/R
Figure4.75Variationofboundcirculationwithbladeazimuthangleat=450.
C.2ExtrapolatedBoundCirculationDistributionsCorrectedforTipandRootLoss
The section describes how the extrapolated distributions for the unsteady bound
circulationsforthewholebladespanwerederivedfromthepointvaluesat40,50,60,70,80
and 90%R. Recall that in axial conditions the bound circulation values at 40% and 90%R
werefoundtobeunrealisticallyhigh(refertosectionB.1,Fig.4.53),eventhoughtheangle
ofattackwasbeingderivedfromtheinflowmeasurements.Thispresentedadifficultyfor
deriving a bound circulationdistribution across the whole blade span. The same problem
was also expected to occur at yaw angles 300 and 450, for the simple reason that the
unsteady aerofoil model employed to derive the lift coefficient was a 2D model and does
not cater for blade tip/root 3D effects. In axial conditions, the prescribed vortex model
HAWT_PVC could easily be employed to derive the extrapolated bound circulation
distributions corrected for tip/root loss using a trialanderror approach (This approach is
referred to as method 2 in section B.2). This was possible because this vortex model was
foundtobeveryaccuratewhentreatingaxialconditions.However,whenitcametoyawed
conditions,itwasfoundthatthepredictedresultsbyHAWT_PVCforua,cdidnotagreevery
well with the corresponding experimental results. Consequently the trialanderror
approach was very difficult to apply at both 300 and 450 yaw. The following alternate
method was applied to obtain the unsteady extrapolated bound circulation distributions
151
Chapter4 AerodynamicAnalysisoftheTUDelftModelTurbine
thataccountfortip/rootlosseffects(B,3D):theestimatedvaluesoftheboundcirculationat
40%and90%Rwerediscarded(astheywereexpectedtobehigh).Theboundcirculationat
the blade root and tip (30% and 100%R) was set to zero. A spline interpolation was then
employedtoobtainacontinuousboundcirculationacrossthewholebladespanusingthe
methoddescribedinAppendixD.Theresultsforsuchdistributionsat300and450yaware
showninFigs.4.76and4.77.Intheseplots,thedistributionsofthe2Duncorrectedvalues
(B,2D)ofFigs.4.74and4.75arealsoincludedforthepurposeofcomparison.Thedifference
between the B,2D and B,3D distributions is at the regions between 30%R and 50%R and
between 80%R and 100%R, where the tip/root loss correction is applied together with the
extrapolation.Figs.4.78and4.79showthesameresultsfor B,3Dbutplottedasafunctionof
thebladeazimuthangle.
Toderiveanaccuratetip/rootlosscorrectionforyawedconditionswerenobladepressure
measurementscouldbeperformedisnotaneasytaskandrequiresathoroughmodellingof
3DflowsatthebladetipssuchasCFD.Thiswasbeyondthescopeoftheproject.Thefactof
notcarryingoutthisindepthanalysisandinsteadapplyingthesimplesplineextrapolation
method described above introduces some level of uncertainty in the unsteady bound
circulation values at the blade tip and root regions. It should be kept in mind that these
derivedboundcirculationswererequiredtovalidatethefreewakevortexmodelusingthe
procedure described in Figs. 2.4 and 5.10. A high uncertainty due to the tip/root loss
correctioncouldhavebeendetrimentaltothisvalidationexercise.Therefore,itwasvitalto
assess the significance of this uncertainty. To do so, the vortex model HAWT_PVC was
used. The uncorrected bound circulation distributions (B,2D) of Fig. 4.74 and 4.75 were
extrapolatedusingthesamemethoddescribedabovebutthecirculationvaluesat40%and
90%R were not discarded (i.e. no tip/root loss correction was applied). The two different
extrapolatedcirculations, B,2Dand B,3D,werethenprescribedtoHAWT_PVCtocalculate
theaxialinducedvelocitiesattheblades(ua,c atYa=0).Thelatterwerecompared,asshown
in Figs. 4.80 and 4.81. Table 4.6 gives the parameters used in HAWT_PVC. In this table,
parameters Rt,w1, Rt,w2, p, and s were obtained from the smoke visualization experiments
(refertosection4.2.3).Thecalculationswereperformedatdifferentvaluesofn(n=11,21
and 31) to show that the numerical errors due to blade discretization are negligible. Figs.
4.82 and 4.83 illustrate the prescribed skewed wakes modelled by HAWT_PVC. The
descrepancy between the induced velocities resulting from the two different circulations,
B,2D and B,3D, gives a first order indication of how significantly this uncertainty in the
tip/root correction would influence the spanwise induced velocity distribution along the
blades. Figs. 4.80 and 4.81 show that this uncertainty mainly influences only induced
velocities at the blade tip and root region. The region between 50% and 80%R experience
onlyaminorinfluence.
152
Chapter4 AerodynamicAnalysisoftheTUDelftModelTurbine
0.7
0.6
B,2D, =
GB,2D f =270
2700
B,3D, f==90
GB,3D 900
0.1
B,3D, f==180
GB,3D 1800
B,3D, =
GB,3D f =270
2700
0
0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
r/R
Figure4.76Comparisonofextrapolatedboundcirculationdistributionscorrectedfortip/rootloss
(B,3D)withuncorrecteddistributions(B,2D)at=300.
0.7
0.6
0.5
B,2D and B,3D (m /s)
B,2D, f==000
GB,2D
2
B,2D, f==180
GB,2D 180
0
0.3
B,2D, =f =
GB,2D 270
270 0
B,3D, f==90
GB,3D 0
90
0.1 B,3D, f==180
GB,3D 1800
B,3D, =f =
GB,3D 270 0
270
0
0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
r/R
Figure4.77Comparisonofextrapolatedboundcirculationdistributionscorrectedfortip/rootloss
(B,3D)withuncorrecteddistributions(B,2D)at=450.
153
Chapter4 AerodynamicAnalysisoftheTUDelftModelTurbine
0.5 0.6
0.45
0.4 0.5
0.35 0.4
0.3
,3D (m /s)
,3D (m /s)
2
2
0.25 0.3
0.2
0.15 0.2
0.05
0 0
0 30 60 90 120 150 180 210 240 270 300 330 360 0 30 60 90 120 150 180 210 240 270 300 330 360
( deg ) ( deg )
0.7 0.6
0.6 0.5
0.5
0.4
,3D (m /s)
,3D (m /s)
0.4
2
0.3
0.3
0.2
0.2
r/R = 0.8
0.1 r/R = 0.5 0.1
0 0
0 30 60 90 120 150 180 210 240 270 300 330 360 0 30 60 90 120 150 180 210 240 270 300 330 360
( deg ) ( deg )
0.6 0.5
0.45
0.5 0.4
0.35
0.4 0.3
,3D (m /s)
,3D (m /s)
0.3
2
2
0.25
0.2
0.2
0.15
r/R = 0.6
0.1 r/R = 0.9
0.1 0.05
0 0
0 30 60 90 120 150 180 210 240 270 300 330 360 0 30 60 90 120 150 180 210 240 270 300 330 360
( deg ) ( deg )
Figure4.78Variationoftheextrapolatedboundcirculation(correctedfortip/rootloss)withblade
azimuthangleat=300.
154
Chapter4 AerodynamicAnalysisoftheTUDelftModelTurbine
0.4 0.45
0.35 0.4
0.3 0.35
0.3
0.25
,3D (m /s)
,3D (m /s)
0.25
0.2
2
2
0.2
0.15
0.15
0.1
0.1 r/R = 0.7
0.05 r/R = 0.4
0.05
0 0
0 30 60 90 120 150 180 210 240 270 300 330 360 0 30 60 90 120 150 180 210 240 270 300 330 360
( deg ) ( deg )
0.5 0.5
0.45 0.45
r/R = 0.5
0.4 0.4
0.35 0.35
0.3 0.3
,3D (m /s)
,3D (m /s)
0.25
2
2
0.25
0.2 0.2
0.15 0.15
r/R = 0.8
0.1 0.1
0.05 0.05
0 0
0 30 60 90 120 150 180 210 240 270 300 330 360 0 30 60 90 120 150 180 210 240 270 300 330 360
( deg ) ( deg )
0.45 0.4
0.4 0.35
0.35 0.3
0.3 0.25
,3D (m /s)
,3D (m /s)
0.25
2
0.2
0.2 0.15
0.15
0.1 0.1
r/R = 0.9
r/R = 0.6
0.05 0.05
0 0
0 30 60 90 120 150 180 210 240 270 300 330 360 0 30 60 90 120 150 180 210 240 270 300 330 360
( deg ) ( deg )
Figure4.79Variationoftheextrapolatedboundcirculation(correctedfortip/rootloss)withblade
azimuthangleat=450.
155
Chapter4 AerodynamicAnalysisoftheTUDelftModelTurbine
B,2D n = 11
GB,2D 11
GB,2D
B,2Dnn == 21
21 GB,2D n = 31
B,2D n = 31
B,2D n = 11
GB,2D 11
GB,2D
B,2Dn == 21
21 GB,2D n = 31
B,2D n = 31
0.2
GB,3D n = 11
B,3D n = 11 GB,3D
B,3D n = 21 n = 21 GB,3D
B,3D n = 31 n = 31 GB,3D
B,3D n = 11
1
n = 11GB,3D
B,3D n = 21 n = 21
GB,3D
B,3D n = 31 n = 31
0
0.5
-0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
-0.4 0
0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
-0.6 -0.5
ua,c
ua,c
-0.8
-1
-1
-1.2 -1.5
-1.4
-1.6
-2
-2.5
-1.8
r/R r/R
GB,2D
B,2D
n = 11
n = 11 GB,2D
B,2D
n = 21 n = 21 GB,2D
B,2D
n = 31 n = 31 GB,2D
B,2D
n = 11 GB,2D
n = 11B,2D
n = 21 GB,2D
n = 21
B,2D
n = 31 n = 31
0.2
GB,3D n = 11
B,3D n = 11 GB,3D
B,3D n = 21 n = 21 GB,3D
B,3D n = 31 n = 31 GB,3D
B,3D n = 11
0.6
n = 11
GB,3D
B,3D n = 21 n = 21
GB,3D
B,3D n = 31 n = 31
0 0.4
ua,c
-0.8
-1
-0.6
-0.8
-1.2
-1
-1.4 -1.2
-1.6 -1.4
-1.8
r/R
-1.6
r/R
Figure4.80SpanwisedistributionsofaxialinducedvelocityatbladesaspredictedbyHAWT_PVC
n = 11
GB,2D n = 11
B,2D B,2D n = 21
GB,2D n = 21 n = 31
GB,2D n = 31
B,2D n = 11
GB,2D n = 11
B,2D
GB,2D n = 21
B,2D n = 21 n = 31
GB,2D n = 31
B,2D
n = 11
GB,3D
B,3D
n = 11 GB,3D
B,3D n = 21
n = 21 GB,3D
B,3D n = 31
n = 31 GB,3D n = 11
B,3D n = 11 GB,3D n = 21 n = 21
GB,3D n = 31
B,3D n = 31
0.4 0.6
B,3D
0.2 0.4
0.2
0
-0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 0
-0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
-0.4
ua,c
ua,c
-0.4
-0.6 -0.6
-0.8
-0.8
-1 -1
-1.2 -1.2
-1.4 -1.4
r/R r/R
Fig. (a): = 00 Fig. (b): = 900
n = 11
GB,2D n = 11
B,2D B,2D n = 21
GB,2D n = 21
B,2D n = 31
GB,2D n = 31 GB,2D
B,2D n = 11
n = 11 GB,2D
B,2D
n = 21 n = 21
GB,2D
B,2D
n = 31 n = 31
n = 11
GB,3D
B,3D n = 11 GB,3D n = 21
B,3D n = 21 GB,3D n = 31
B,3D n = 31 n = 11
GB,3D n = 11 GB,3D n = 21 n = 21
GB,3D n = 31 n = 31
0.4
B,3D
0.6
B,3D B,3D
0.2
0.4
0
-0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 0.2
-0.4
0
ua,c
ua,c
-1
-0.4
-1.2
-1.4
-0.6
-1.6 -0.8
r/R r/R
Figure4.81SpanwisedistributionsofaxialinducedvelocityatbladesaspredictedbyHAWT_PVC
156
Chapter4 AerodynamicAnalysisoftheTUDelftModelTurbine
Table4.6:ParametersusedforthecomputationsusingHAWT_PVCfor=300and450.
=300 =450
nwRev 10 10
tot 36 36
100 100
Rt,w1 0.61 0.60
Rt,w2 0.62 0.61
p 0.32 0.28
ReleaseTip 0.99 0.99
ReleaseRoot 0.305 0.305
s 350 480
0.5mm 0.5mm
Figure4.82PrescribedhelicalwakeasmodelledbyHAWT_PVCat=300.n=21.
157
Chapter4 AerodynamicAnalysisoftheTUDelftModelTurbine
Figure4.83PrescribedhelicalwakeasmodelledbyHAWT_PVCat=450.n=21.
C.4AssessmentoftheEffectsofTunnelBlockageinYawedConditions
To assess for tunnel blockage effects, Procedure 1 described in Fig. 4.48 was applied. The
boundcirculationdistributionsshowninFigs.4.78and4.79wereprescribedtoHAWT_PVC
to calculate the induced velocities at the wind tunnel exit (ua tunnel exit) over a circular
region having a diameter equal to that of the rotor. The parameters used for HAWT_PVC
werethesameasintable4.6.Theinducedvelocitywascomputedinadirectionparallelto
the tunnel jet. To obtain a conservative estimate, the upper limit of the circulation
distributionswasusedforthecalculations.Sinceyawedconditionswerebeingconsidered,
thecomputedinducedvelocityacrossthetunnelexitjetwasuneven.Theresultsareplotted
inFig.4.84.Itcanbenotedthatthepeakinducedvelocitiesareverysmallatboth300and
450 yaw (0.13 and 0.09m/s, respectively). This is much less than the pitotreading at the
tunnel exit (5.5m/s). The equivalent values for ke computed from Eqt. 4.64 are 0.024 and
0.016,respectively.Thiswaslessthantheacceptablelimitof0.05.Thusitmaybeconcluded
thatinyawedconditionsthetunnelblockageeffectswerealsominimal.
158
Chapter4 AerodynamicAnalysisoftheTUDelftModelTurbine
(deg) (deg)
=300=450
Figure 4.84 Induced velocities computed using HAWT_PVC at tunnel exit across area equal to
thatofrotor(n=21).
C.5AxialInductionFactorDistributions
Sincewindtunnelblockageeffectswerefoundtobesmall,thenitwasjustifiedtotakeUto
be equal to Ujet (5.5m/s). The axial induction distributions for ua,c at each measuring plane
(Ya=6,3.5,6and9cm)couldbefoundusingtheexperimentalvaluesofwa,cinFigs.4.19and
4.20(takingthesmoothenedvalues)andEqt.4.57.Theresultsforua,cat300and450yaware
showninFigs.4.85and4.86.
Thecorrespondingazimuthallyaveragedinductiondistributionscouldbefoundusingthe
samemethod asfor =00(i.e.usingEqts.4.81and4.82).Theresultsfor uaatthedifferent
measuringplanesareshowninFigs.4.87and4.88.Linearinterpolationwasthenusedto
estimate ua at the rotorplaneusing an equation similar toEqt. 4.7. Thedistributions for ua
weresmoothedusingtheGaussiankerneltechniquedescribedinEqt.4.8.Thesesmoothed
distributions are also included in Figs. 4.87 and 4.88. It may be easily observed that for
yawed conditions, ua is not always constant with the rotor azimuth angle (). A cyclic
variationinuaisobservedatthebladerootandtipregionsandisaconsequenceoftheroot
andtipvorticesinthewake.Thisvariationhasafrequencyoftwicetherotorangularspeed
forthesimplereasonthattherotorhastwoblades.
The axial induction factors at the rotorplane (a1,c and a1) were found by dividing the
interpolatedinducedvelocitiesattherotorplaneby5.5m/s.TheresultsareshowninFigs.
4.89and4.90.
159
Chapter4 AerodynamicAnalysisoftheTUDelftModelTurbine
0
0 45 90 135 180 225 270 315 360
-0.5
-1
ua,c (m/s)
-1.5
6cm UpStream
3.5cm DwnStream
r/R=0.4 -2
6cm DwnStream
9cm DwnStream
-2.5
RotorPlane (Interp)
(deg)
0
0 45 90 135 180 225 270 315 360
-0.5
-1
ua,c (m/s)
-1.5
6cm UpStream
r/R=0.5 -2
3.5cm DwnStream
6cm DwnStream
9cm DwnStream
RotorPlane (Interp)
-2.5
(deg)
0
0 45 90 135 180 225 270 315 360
-0.5
yaw30
-1
ua,c (m/s)
-1.5
6cm UpStream
r/R=0.6 3.5cm DwnStream
-2
6cm DwnStream
9cm DwnStream
-2.5
RotorPlane (Interp)
(deg)
Figure 4.85 Variation of axial induced velocity at blades with blade azimuth angle at different
radiallocationsat=300.
160
Chapter4 AerodynamicAnalysisoftheTUDelftModelTurbine
0
0 45 90 135 180 225 270 315 360
-0.5
-1
ua,c (m/s)
-1.5
6cm UpStream
3.5cm DwnStream
r/R=0.7 -2
6cm DwnStream
9cm DwnStream
RotorPlane (Interp)
-2.5
(deg)
0
0 45 90 135 180 225 270 315 360
-0.5
-1
ua,c (m/s)
-1.5
-2
6cm UpStream
r/R=0.8 3.5cm DwnStream
9cm DwnStream
RotorPlane (Interp)
-3
(deg)
0
0 45 90 135 180 225 270 315 360
-0.5
-1
ua,c (m/s)
-1.5
-2
6cm UpStream
6cm DwnStream
-2.5
9cm DwnStream
RotorPlane
(Interp)
-3
(deg)
Figure4.85contd.frompreviouspage.
161
Chapter4 AerodynamicAnalysisoftheTUDelftModelTurbine
0
0 45 90 135 180 225 270 315 360
-0.5
-1
ua,c (m/s)
-1.5
6cm UpStream
3.5cm DwnStream
r/R=0.4 -2 6cm DwnStream
9cm DwnStream
RotorPlane
-2.5
(Interp)
(deg)
0
0 45 90 135 180 225 270 315 360
-0.5
-1
ua,c (m/s)
-1.5
6cm UpStream
r/R=0.5 3.5cm DwnStream
-2
6cm DwnStream
9cm DwnStream
RotorPlane
(Interp)
-2.5
(deg)
0
0 45 90 135 180 225 270 315 360
-0.5
-1
ua,c (m/s)
-1.5
3.5cm DwnStream
-2 6cm DwnStream
9cm DwnStream
RotorPlane
(Interp)
-2.5
(deg)
Figure 4.86 Variation of axial induced velocity at blades with blade azimuth angle at different
radiallocationsat=450.
162
Chapter4 AerodynamicAnalysisoftheTUDelftModelTurbine
0
0 45 90 135 180 225 270 315 360
-0.5
-1
ua,c (m/s)
-1.5
6cm UpStream
9cm DwnStream
RotorPlane
(Interp)
-2.5
(deg)
0
0 45 90 135 180 225 270 315 360
-0.5
-1
ua,c (m/s)
-1.5
6cm UpStream
RotorPlane
(Interp)
-2.5
(deg)
0
0 45 90 135 180 225 270 315 360
-0.5
-1
ua,c (m/s)
-1.5
9cm DwnStream
RotorPlane
(Interp)
-2.5
(deg)
Figure4.86contd.frompreviouspage.
163
Chapter4 AerodynamicAnalysisoftheTUDelftModelTurbine
-0.6
0 45 90 135 180 225 270 315 360
-0.8
-1
-1.2
ua (m/s)
6cm UpStream
6cm UpStream_Smooth
-1.4
3.5cm DwnStream
3.5cm DwnStream_Smooth
-1.6
6cm DwnStream
6cm DwnStream_Smooth
r/R=0.4 9cm DwnStream
-1.8
9cm DwnStream_Smooth
Rotorplane (Interp)
-2
(deg)
-0.6
0 45 90 135 180 225 270 315 360
-0.8
-1
ua (m/s)
-1.2
6cm UpStream
6cm UpStream_Smooth
3.5cm DwnStream
-1.4 3.5cm DwnStream_Smooth
6cm DwnStream_Smooth
-1.6 9cm DwnStream
9cm DwnStream_Smooth
Rotorplane (Interp)
-1.8
(deg)
-0.6
0 45 90 135 180 225 270 315 360
-0.8
-1
-1.2
ua (m/s)
6cm UpStream
-1.4
6cm UpStream_Smooth
3.5cm DwnStream
3.5cm DwnStream_Smooth
6cm DwnStream_Smooth
9cm DwnStream
-1.8
9cm DwnStream_Smooth
-2
Rotorplane (Interp)
(deg)
Figure4.87Variationofazimuthallyaveragedaxialinducedvelocitywithbladeazimuthangleat
differentradiallocationsat=300.
164
Chapter4 AerodynamicAnalysisoftheTUDelftModelTurbine
-0.6
0 45 90 135 180 225 270 315 360
-0.8
-1
-1.2
ua (m/s)
6cm UpStream
-1.4
6cm UpStream_Smooth
3.5cm DwnStream
3.5cm DwnStream_Smooth
6cm DwnStream_Smooth
r/R=0.7 9cm DwnStream
-1.8
9cm DwnStream_Smooth
Rotorplane (Interp)
-2
(deg)
-0.6
0 45 90 135 180 225 270 315 360
-0.8
-1
-1.2
ua (m/s)
6cm UpStream
-1.4
6cm UpStream_Smooth
3.5cm DwnStream
-1.6
3.5cm DwnStream_Smooth
r/R=0.8 -1.8
6cm DwnStream
6cm DwnStream_Smooth
9cm DwnStream
-2
9cm DwnStream_Smooth
Rotorplane (Interp)
-2.2
(deg)
-0.4
0 45 90 135 180 225 270 315 360
-0.6
-0.8
-1
-1.2
ua (m/s)
6cm UpStream
-1.4
6cm UpStream_Smooth
3.5cm DwnStream
9cm DwnStream
-2 9cm DwnStream_Smooth
-2.2
Rotorplane (Interp)
(deg)
Figure4.87contd.frompreviouspage.
165
Chapter4 AerodynamicAnalysisoftheTUDelftModelTurbine
-0.6
0 45 90 135 180 225 270 315 360
-0.8
-1
ua (m/s)
6cm UpStream
-1.2
6cm UpStream_Smooth
3.5cm DwnStream
3.5cm DwnStream_Smooth
-1.4
6cm DwnStream
r/R=0.4 -1.6
6cm DwnStream_Smooth
9cm DwnStream
9cm DwnStream_Smooth
Rotorplane (Interp)
-1.8
(deg)
-0.4
0 45 90 135 180 225 270 315 360
-0.6
-0.8
ua (m/s)
6cm UpStream
-1
6cm UpStream_Smooth
3.5cm DwnStream
Rotorplane (Interp)
-1.6
(deg)
-0.4
0 45 90 135 180 225 270 315 360
-0.6
-0.8
ua (m/s)
6cm UpStream
-1
6cm UpStream_Smooth
3.5cm DwnStream
-1.6
Rotorplane (Interp)
(deg)
Figure4.88Variationofazimuthallyaveragedaxialinducedvelocitywithbladeazimuthangleat
differentradiallocationsat=450.
166
Chapter4 AerodynamicAnalysisoftheTUDelftModelTurbine
-0.4
0 45 90 135 180 225 270 315 360
-0.6
-0.8
-1
ua (m/s)
6cm UpStream
6cm UpStream_Smooth
-1.2
3.5cm DwnStream
3.5cm DwnStream_Smooth
-1.4 6cm DwnStream
6cm DwnStream_Smooth
r/R=0.7
-1.6 9cm DwnStream
9cm DwnStream_Smooth
Rotorplane (Interp)
-1.8
(deg)
-0.4
0 45 90 135 180 225 270 315 360
-0.6
-0.8
-1
ua (m/s)
6cm UpStream
-1.2
6cm UpStream_Smooth
3.5cm DwnStream
-1.4
3.5cm DwnStream_Smooth
r/R=0.8 -1.6
6cm DwnStream
6cm DwnStream_Smooth
-1.8
9cm DwnStream
9cm DwnStream_Smooth
Rotorplane (Interp)
-2
(deg)
-0.2
0 45 90 135 180 225 270 315 360
-0.4
-0.6
-0.8
ua (m/s)
6cm UpStream
-1
6cm UpStream_Smooth
3.5cm DwnStream
-1.2
r/R=0.9 3.5cm DwnStream_Smooth
6cm DwnStream
-1.4
6cm DwnStream_Smooth
9cm DwnStream
-1.6
9cm DwnStream_Smooth
Rotorplane (Interp)
-1.8
(deg)
Figure4.88contd.frompreviouspage.
167
Chapter4 AerodynamicAnalysisoftheTUDelftModelTurbine
-0.12
0 60 120 180 240 300 360
-0.14
-0.16
-0.18
-0.2
a1,c (m/s)
-0.22
-0.24
r/R = 0.5
-0.28 r/R = 0.6
r/R = 0.7
-0.3 r/R = 0.8
r/R = 0.9
-0.32
(deg)
-0.18
0 60 120 180 240 300 360
-0.2
-0.22
a1 (m/s)
-0.24
-0.26
r/R = 0.4
r/R = 0.5
r/R = 0.6
-0.28
r/R = 0.7
r/R = 0.8
r/R = 0.9
-0.3
(deg)
Figure4.89Variationoftheaxialinductionfactor(atbladeliftinglineandazimuthallyaveraged)
withblade/rotorazimuthangleat=300.
168
Chapter4 AerodynamicAnalysisoftheTUDelftModelTurbine
-0.04
-0.08
-0.12
-0.16
a1,c (m/s)
-0.2
r/R = 0.4
-0.24
r/R = 0.5
r/R = 0.6
-0.28 r/R = 0.7
r/R = 0.8
r/R = 0.9
-0.32
(deg)
-0.12
0 60 120 180 240 300 360
-0.14
-0.16
a1 (m/s)
-0.18
r/R = 0.4
r/R = 0.5
r/R = 0.7
r/R = 0.8
r/R = 0.9
-0.22
(deg)
Figure4.90Variationoftheaxialinductionfactor(atbladeliftinglineandazimuthallyaveraged)
withblade/rotorazimuthangleat=450.
169
Chapter4 AerodynamicAnalysisoftheTUDelftModelTurbine
C.6ComparisonofExperimentalAxialInducedVelocitieswiththosefromHAWT_PVC
Oncetheaxialinducedvelocitieswerederived(sectionC.5),anattemptwasmadetoassess
thecapabilityoftheprescribedvortexmodelHAWT_PVCintreatingyawedconditionsat
both 300 and 450 yaw. To do so, the parameters of table 4.6 were used together with the
unsteady bound circulation distributions of Figs. 4.78 and 4.79. These distributions were
prescribedtothemodeltocomputetheaxialinducedvelocities(ua,c)at3.5cmdownstream
oftherotorplane.Thesewherethencomparedwiththoseobtainedexperimentally(ofFigs.
4.85and4.86).ThecomparisonisshowninFigs.4.91and4.92.Notethatcomparisonofthe
valuesat3.5cmdownstreamisbeingmadeinsteadofthatattherotorplanesoastoavoid
anyambiguityresultingfromtheadditionaluncertaintiesinthelinearinterpolation(using
Eqt. 4.7). In Figs. 4.91 and 4.92, the error bars in the experimental values of ua,c are
displayed. These are equal to 0.26m/s at 300 yaw and 0.21m/s at 450 yaw. The vortex
modelresultsalsoincludeerrorsbars.Theserepresenttheuncertaintyinthevortexmodel
predictionsforua,cresultingfromtheuncertaintiesintheprescribedboundcirculations.Re
call from section C.1 that the prescribed bound circulations were derived from the inflow
measurements and therefore these circulations are also subject to the uncertainties in the
inflowmeasurements.SuchuncertaintiesweredisplayedearlierinFigs.4.78and4.79.
Despite the fact that in HAWT_PVC the wake was modelled using experimental data
collected from the smoke visualization measurements, the correlation ofthe vortex model
predictions with the experimental results was not as good as that achieved in axial
conditions (Fig. 4.56). One reason for this is the fact that wake rollup is not modelled in
HAWT_PVC.Anotherreasonisthatshedcirculationisalsonotincludedinthewakemodel
embedded in this vortex code. However, in the validation work of the newly developed
freewakevortexmodel,HAWT_FWC(whichaccountsforbothrollupandshedcirculation
in the wake), it was discovered that there are other reasons for not obtaining a good
correlationwiththeexperimentalmeasurements.Thelevelsofshedcirculationwerefound
tobeverysmallcomparedtotrailingcirculation.Butitwasdiscoveredthatitisverylikely
thattheflowinterferencecausedbythecentrebodyofthetunnelturbineshouldbeblamed.
Looking closely at Figs. 4.91 and 4.92 shows that for blade positions 0<<1800, the
correlation with the experimental values is relatively good, taking the uncertainty limits
intoaccount.Thelargedisagreementonlyoccursforbladepositions180<<3600.Thesmoke
visualizationphotosrevealedthatinyaw,thetipvortexpathwasbeingobstructedbythe
centrebody on the upwind side of the rotor, i.e. for the same blade positions 180<<3600.
ThisinterferencewasnotmodelledinHAWT_PVC.Furtherdetailsofthisinvestigationare
givenlateroninChapter5,page262.
170
Chapter4 AerodynamicAnalysisoftheTUDelftModelTurbine
1 1
Exp, r/R = 0.4 Exp, r/R = 0.7
0.5 HAWT_PVC, r/R = 0.4
0.5 HAWT_PVC, r/R = 0.7
0 0
0 45 90 135 180 225 270 315 360 0 45 90 135 180 225 270 315 360
-0.5
ua,c (m/s)
ua,c (m/s)
-0.5
-1 -1
-1.5 -1.5
-2 -2
-2.5 -2.5
(deg) ( deg )
1 1
Exp, r/R = 0.5 Exp, r/R = 0.8
0.5 HAWT_PVC, r/R = 0.5 0.5
HAWT_PVC, r/R = 0.8
0 0
0 45 90 135 180 225 270 315 360 0 45 90 135 180 225 270 315 360
-0.5
ua,c (m/s)
ua,c (m/s)
-0.5
-1 -1
-1.5 -1.5
-2 -2
-2.5 -2.5
(deg) ( deg )
1 1
Exp, r/R = 0.6
Exp, r/R = 0.9
0.5
HAWT_PVC, r/R = 0.6 0.5
HAWT_PVC, r/R = 0.9
0 0
0 45 90 135 180 225 270 315 360 0 45 90 135 180 225 270 315 360
-0.5
ua,c (m/s)
ua,c (m/s)
-0.5
-1 -1
-1.5
-1.5
-2 -2
-2.5 -2.5
(deg) (deg)
Figure 4.91 Comparison of axial induced velocity distributions computed by HAWT_PVC with
thoseobtainedfromthehotfilmmeasurementsat=300atYa=3.5cmdownstream.
171
Chapter4 AerodynamicAnalysisoftheTUDelftModelTurbine
1 1
0
0
0 45 90 135 180 225 270 315 360
0 45 90 135 180 225 270 315 360
ua,c (m/s)
ua,c (m/s)
-0.5
-0.5
-1
-1
-1.5
-2 -1.5
-2.5 -2
(deg) ( deg )
1 1
Exp, r/R = 0.5 Exp, r/R = 0.8
0.5 HAWT_PVC, r/R = 0.5 0.5 HAWT_PVC, r/R = 0.8
0
0 45 90 135 180 225 270 315 360 0
-0.5 0 45 90 135 180 225 270 315 360
ua,c (m/s)
ua,c (m/s)
-0.5
-1
-1
-1.5
-1.5
-2
-2.5 -2
(deg) (deg)
1 1
Exp, r/R = 0.6 Exp, r/R = 0.9
0.5 HAWT_PVC, r/R = 0.6 0.5 HAWT_PVC, r/R = 0.9
0
0 45 90 135 180 225 270 315 360 0
-0.5 0 45 90 135 180 225 270 315 360
ua,c (m/s)
ua,c (m/s)
-0.5
-1
-1
-1.5
-1.5
-2
-2.5 -2
(deg) (deg)
Figure 4.92 Comparison of axial induced velocity distributions computed by HAWT_PVC with
thoseobtainedfromthehotfilmmeasurementsat=450atYa=3.5cmdownstream.
172
Chapter4 AerodynamicAnalysisoftheTUDelftModelTurbine
C.7BladeLoadDistributions
Asforaxialconditions,thevaluesof B,3DofFigs.4.78and4.79wereusedtodeterminethe
unsteady lift coefficient corrected for tip/root loss (Cl,3D) in accordance with Eqt. 4.14 at
radiallocations40,50,60,70,80and90%Randatdifferentbladeazimuthanglesoverone
whole revolution of the rotor. Recall that the calculations with HAWT_PVC (section C.2)
haveshownthatradiallocations50,60,70,80%Rtendtobenotaffectedbythetip/rootloss
correction and therefore Cl,3D is equal to Cl,2D as estimated by the unsteady aerofoil model
described in section 4.3.2 and the angle of attack values of Figs. 4.68 and 4.69. The drag
coefficientwasdeterminedfromthe2DstaticdatafortheNACA0012aerofoilandwasnot
corrected for tip/root loss and unsteadiness. The error incurred was small since at low
anglesofattack,theCdareverysmallcomparedwithCl(<5%oftheCl).Figs.4.93and4.94
show the variations of the unsteady lift coefficient (Cl,3D) with angle of attack at different
radiallocationsat300and450yaw.Forthesakeofclarity,theerrorbarsarenotdisplayed
in these plots. Due to the unsteadiness at each radial location (i.e. each aerofoil section
experiencestimedependentvariationsofVrand ),hysterisisloopsareformed.Sincethe
angles of attack are small and the flow over the blades is attached, then no dynamic stall
takes place. The loops are modelled by the unsteady aerofoil model and if static aerofoil
datawasusedinthecomputations,noloopswouldbeobserved,butasimplestraightline
alongwhichCland willvary.Asthelevelofunsteadinessincreaseswithyawangle,the
loopsareingeneralwiderat =450thanat =300,eventhoughtheoverallvaluesofCland
aresmaller.
Figs.4.95and4.96showtheresultsobtainedfortheaxialthrustloadingasafunctionofthe
bladeazimuthangle().Theseresultswerecomputedusingthesamemethodasforaxial
conditions.At=300,theerrorintervalsbyassuminga0.26m/sdeviationinwa,crangefrom
1.2N/matr/R=0.4to2.8N/matr/R=0.9.Thisisapproximatelyequivalenttoanerror T in
therangeof1425%.At =450,theerrorintervalsbyassuminga0.21m/sdeviationinwa,c
rangefrom1.0N/matr/R=0.4to2.0N/matr/R=0.9.Thisisapproximatelyequivalentto
the percentage error T being in the range of 16 35%. These error intervals are large,
considerablylargerthanthoseoftheinflowmeasurementsfromwhichtheywerederived.
Thisconsolidateswhatwasprovedanalyticallyinsection4.3.5,i.e.thepercentageerrorin
thederivedaxialthrustloading( T )isverysensitivetotheerrorintheinflowattheblade
liftinglinesobtainedfromtheinflowmeasurements,( wa,c ).Itcanalsobeconfirmedthatthe
sensitivity increases at larger yaw angles. For the torque loading, the error intervals were
found to be much larger, sometimes exceeding 100% of the azimuthally averaged thrust
loadingatsomeradiallocations.Thisalsoconsolidateswhatwasshowninsection4.3.5,i.e.
the error in the derived torque loading ( Q ) resulting from the error in the inflow
measurementsisusuallylargerthanthatintheaxialthrustloading( T ).Thisexplainsthe
limitation for this analysis, i.e. to derive the unsteady aerodynamic loads from hotfilm
measurements in the near wake. The accuracy of the results could only be improved by
beingabletomeasuretheflowvelocitiesmoreaccurately.
173
Chapter4 AerodynamicAnalysisoftheTUDelftModelTurbine
0.8 0.45
0.7
0.4
0.6
0.35
0.5
Cl,3D
Cl,3D
0.4 0.3
0.3
0.25
0.2
r/R = 0.4 r/R = 0.7
0.1 0.2
0 0.15
0 1 2 3 4 5 6 7 8 9 10
( deg )
1.5 2 2.5
( deg )
3 3.5 4
0.7 0.38
0.6 0.36
0.34
0.5
0.32
0.4
0.3
Cl,3D
Cl,3D
0.3 0.28
0.26
0.2
0.24
r/R = 0.5
r/R = 0.8
0.1 0.22
0 0.2
0 1 2 3 4 5 6 7 1.5 2 2.5 3 3.5
( deg ) ( deg )
0.55 0.28
0.5 0.26
0.45 0.24
0.4
0.22
0.35
Cl,3D
0.2
Cl,3D
0.3
0.18
0.25
0.16
0.2
r/R = 0.6 r/R = 0.9
0.14
0.15
0.12
0.1
1.5 2 2.5 3 3.5 4 4.5 5 0.1
( deg ) 1.5 2 2.5 3 3.5 4
( deg )
Figure4.93Variationofunsteadyliftcoefficientwithangleofattackatvariousradiallocationsat
=300.
174
Chapter4 AerodynamicAnalysisoftheTUDelftModelTurbine
0.7 0.4
0.6 0.35
0.5 0.3
0.25
0.4
Cl,3D
Cl,3D
0.2
0.3
0.15
0.2 0.1
r/R = 0.4
r/R = 0.7
0.1 0.05
0 0
0 1 2 3 4 5
( deg )
6 7 8 9 0.5 1 1.5 2
( deg )
2.5 3 3.5
0.6 0.35
0.5 0.3
0.25
0.4
0.2
Cl,3D
Cl,3D
0.3
0.15
0.2
0.1
r/R = 0.5 r/R = 0.8
0.1
0.05
0 0
0 1 2 3 4 5 6 1 1.5 2 2.5 3
( deg ) ( deg )
0.45 0.24
0.4 0.22
0.35
0.2
0.3
0.18
Cl,3D
0.25
Cl,3D
0.2
0.16
0.15
0.14
0.1
r/R = 0.9
r/R = 0.6
0.12
0.05
0 0.1
0 0.5 1 1.5 2 2.5 3 3.5 4 1 1.5 2 2.5 3 3.5
( deg ) ( deg )
Figure4.94Variationofunsteadyliftcoefficientwithangleofattackatvariousradiallocationsat
=450.
175
Chapter4 AerodynamicAnalysisoftheTUDelftModelTurbine
10 20
9 18
8 16
7 14
dT3D (N/m)
dT3D (N/m)
6 12
5 10
4 8
3 6
r/R = 0.4 r/R = 0.7
2 4
1 2
16 25
14
20
12
10
dT3D (N/m)
dT3D (N/m)
15
8
6 10
r/R = 0.8
4 r/R = 0.5
5
2
0 0
0 30 60 90 120 150 180 210 240 270 300 330 360 0 30 60 90 120 150 180 210 240 270 300 330 360
( deg ) ( deg )
18 25
16
14 20
12
dT3D (N/m)
dT3D (N/m)
15
10
8 10
6
r/R = 0.9
4 r/R = 0.6
5
2
0 0
0 30 60 90 120 150 180 210 240 270 300 330 360 0 30 60 90 120 150 180 210 240 270 300 330 360
( deg ) ( deg )
Figure4.95Variationofunsteadyaxialthrustloadingwithbladeazimuthangleatvariousradial
locationsat=300.
176
Chapter4 AerodynamicAnalysisoftheTUDelftModelTurbine
16
8
14
7
12
6
10
5
dT3D (N/m)
dT3D (N/m)
8
4
6
3
4 r/R = 0.7
2
r/R = 0.4
2
1
0
0 0 30 60 90 120 150 180 210 240 270 300 330 360 0 30 60 90 120 150 180 210 240 270 300 330 360
( deg ) ( deg )
12 20
10 18
16
14
8
dT3D (N/m)
12
dT3D (N/m)
6 10
4 8
6
r/R = 0.8
r/R = 0.5 4
2
2
0 0
0 30 60 90 120 150 180 210 240 270 300 330 360 0 30 60 90 120 150 180 210 240 270 300 330 360
( deg ) ( deg )
14 20
12 18
16
10
14
dT3D (N/m)
12
dT3D (N/m)
8
10
6 8
4 6
r/R = 0.9
r/R = 0.6 4
2
2
0 0
177
Chapter4 AerodynamicAnalysisoftheTUDelftModelTurbine
0.9
0.5
0.45 0.8 r/R = 0.7
0.4
0.7
0.35
0.6
dQ3D (Nm/m)
dQ3D (Nm/m)
0.3 0.5
0.25 0.4
0.2
0.15 0.3
0.1 0.2
r/R = 0.4
0.05 0.1
0 0
0 30 60 90 120 150 180 210 240 270 300 330 360 0 30 60 90 120 150 180 210 240 270 300 330 360
( deg ) ( deg )
0.8 1
0.9
0.7
0.8
0.6 0.7
0.5
dQ3D (Nm/m)
dQ3D (Nm/m)
0.6
0.4 0.5
0.4
0.3
0.3
0.2 0.2
r/R = 0.5 r/R = 0.8
0.1 0.1
0 0
0 30 60 90 120 150 180 210 240 270 300 330 360 0 30 60 90 120 150 180 210 240 270 300 330 360
( deg ) ( deg )
0.9
0.8
0.8 0.7
r/R = 0.6
0.7 0.6
0.6 0.5
dQ3D (Nm/m)
dQ3D (Nm/m)
0.4
0.5
0.3
0.4 0.2
0.3 r/R = 0.9
0.1
0.2
0
0 30 60 90 120 150 180 210 240 270 300 330 360
0.1 -0.1
0 -0.2
0 30 60 90 120 150 180 210 240 270 300 330 360 ( deg )
( deg )
Figure 4.97 Variation of unsteady torque loading with blade azimuth angle at various radial
locationsat=300.
178
Chapter4 AerodynamicAnalysisoftheTUDelftModelTurbine
0.35 0.6
0.5
0.3
dQ3D (N/m)
0.2
0.15
0.1
0.1
0
0.05 0 30 60 90 120 150 180 210 240 270 300 330 360
-0.1
0 -0.2
0 30 60 90 120 150 180 210 240 270 300 330 360
-0.05 -0.3
( deg ) ( deg )
0.6 0.7
0.5 0.6
r/R = 0.8
0.5
r/R = 0.5
0.4 0.4
dQ3D (N/m)
dQ3D (N/m)
0.3
0.3
0.2
0.2 0.1
0.1 0
0 30 60 90 120 150 180 210 240 270 300 330 360
0
-0.1
0 30 60 90 120 150 180 210 240 270 300 330 360 -0.2
-0.1 -0.3
( deg ) ( deg )
0.6 0.6
dQ3D (N/m)
dQ3D (N/m)
0.2
0.2 0.1
0.1 0
0 30 60 90 120 150 180 210 240 270 300 330 360
0
-0.1
0 30 60 90 120 150 180 210 240 270 300 330 360
-0.1 -0.2
-0.2 -0.3
( deg ) ( deg )
Figure 4.98 Variation of unsteady torque loading with blade azimuth angle at various radial
locationsat=450.
179
Chapter4 AerodynamicAnalysisoftheTUDelftModelTurbine
C.8ComparisonofAxialThrustCoefficients
The axial thrust coefficient at each yaw angle for =8 and tip=20 was computed by
HAWT_LFIMbyintegratingnumericallytheaxialthrustloadingdistributionsofFigs.4.95
and4.96.Thesewerecomparedwiththosemeasuredusingstraingaugesinthewindtunnel
experiments(refersection4.2.3,Fig.4.26).ThecomparisonforCTisgivenintable4.7below.
NotethattheazimuthalaveragedvalueofCTisbeingcompared.
Table4.7:ComparisonofaxialthrustcoefficientsderivedbyHAWT_LFIMusinghotfilmnearwake
measurementswiththosemeasuredusingstraingaugetechniques.(Tip/rootlosscorrectionapplied)
In table 4.7, the percentage discrepancies relative to the measured values are given. The
uncertaintyintervalsinCTduetotheassumederrorsinwa,carealsoincludedtogetherwith
the corresponding uncertainties in the percentage discrepancy. The mean percentage
discrepancy is found to be quite high (about 32%) which is considerably larger than that
obtained for =00 (about 16%, refer to table 4.5). Even the range in the uncertainity
resulting from the uncertainty in wa,c is verylarge (about 12%) and thisshows that the CT
values derived using HAWT_PVC are very sensitive to the errors in wa,c. Apart from the
errors in wa,c, another source of uncertainty is due to the tip/root loss correction used
(describedinsectionC.2).
TobeabletoassessthesensitivityofCTtotheappliedtip/rootlosscorrection,HAWT_LFIM
computedtheCTvalueswithnotip/rootlosscorrection.Theaxialthrustloading(dT)values
atradiallocations40,50,60,70,80and90%Rwerecomputedusingthe2DvaluesofCland
Cd(i.e.notip/rootlosscorrection).Foreachbladeazimuthangle,thespanwisedistributions
werethenextrapolatedbyapplyingtheboundaryconditionthatat30and100%RdTiszero
andapplyingasplineinterpolation(usingmethodofAppendixD).Suchdistributionswere
integratedtoyieldCT.Theseresultswerecomparedwiththemeasuredvalues,asshownin
table4.8below.
In table 4.8, it is noted that the mean percentage discrepancy is lower than that with no
tip/loss correction (20% instead of 32%) indicating that the uncertainty in this correction
hasaconsiderableinfluenceonthederivedvaluesforCT.Yettheuncertaintyresultingfrom
errorsinwa,carisingfromthehotfilmmeasurementsislarger.Anotherimportantpointis
that the uncertainty in the tip/root losscorrection is not expected toinfluence the loading
distributionsoverthemiddlebladesections(between50%and80%R).
180
Chapter4 AerodynamicAnalysisoftheTUDelftModelTurbine
Table4.8:ComparisonofaxialthrustcoefficientsderivedbyHAWT_LFIMusinghotfilmnearwake
measurements with those measured using strain gauge techniques. (Tip/root loss correction not
applied)
181
Chapter4 AerodynamicAnalysisoftheTUDelftModelTurbine
4.4InvestigatingtheLimitationsoftheBEMTheoryfor
Axial/YawedWindTurbines
This section describes the work in which the hotfilm inflow measurements and the
derivedaerodynamicloadingdistributionsonthebladeswereusedtoassessthelimitations
oftheBEMtheorywhenmodellingwindturbinesinbothaxialandyawedconditions.This
study demonstrated the importance of knowing the experimental blade aerodynamic
loadingdistributions(apartfromtheinflowmeasurements)tobeabletocarryoutamore
detailedassessmentofthelimitationsofBEMbasedmodels.However,oneshouldbearin
mind the fact that, for the TUDelft rotor, the blade loads were derived from the inflow
measurements (as documented in section 4.3). It was found that the uncertainty in such
loads was found to be considerably large, especially when the rotor was yawed. This
uncertaintywasamajorstumblingblocktoperformanaccuratequantitativeassessmentof
the actual deficiencies of the BEM theory. The uncertainty in the tip/root correction also
presented difficulties, but this tended to influence only radial locations r/R<50% and
r/R>80%.
This assessment was only restricted to the operating conditions at which the inflow
measurementsontheTUDelftrotorwerecarriedout(=8, tip=20and =0,30and450),and
for which the flow behaviour over the blades is known to be fully attached (i.e. no blade
stall).Twoseparateapproacheswereused:ApproachAandApproachB.
(i)ApproachAisanoriginalapproachthatmadeuseofinflowmeasurementstocheckthe
BEMEqt.3.21whenmodellingbothaxialandyawedconditions.Theinflowmeasurements
andtheunsteadyaerofoilmodeldescribedinsection4.3.2wereusedtocalculateseparately
thebladeelementtheoryandmomentumpartsdenotedbyFA1andFA2where
c B 1 B 1
FA1 =
Vr CL ( , Re ) V Cos V Sin + Vr CD ( , Re) V Cos + V Sin
2 r b =0
b =0
(4.83a)
(4.83b)
The capability of the BEM theory to simulate yawed conditions depends on the
discrepanciesbetweenthevaluesofFA1andthecorrespondingvaluesofFA2.Thesetwoso
called BEM parameters were computed at different radial positions, rotor azimuth angles
and yaw angles. A large discrepancy implied a deficiency of the BEM theory to simulate
such conditions for the model wind turbine. Actually, this approach investigated the
validityofthemomentumequationforaxialthrust(Eqt.3.6a).
182
Chapter4 AerodynamicAnalysisoftheTUDelftModelTurbine
ForthiscasestudywiththeTUDelftrotor,thelinearlyinterpolatedvaluesofwa,cwithinthe
rotorplane (Figs. 4.18(a), 4.19 and 4.20) were used to compute FA1. The azimuthally
averagedaxialinducedvelocitiesobtainedfromtheinflowmeasurements(Fig.4.59(b),4.87
and4.88)wereusedtocomputeFA2.Theliftcoefficientvaluescorrectedfortip/rootlossat
r/R=40% and 90% (Cl,3D) were used for computing FA1 (see Figs. 4.62, 4.93 and 4.94). 2D
NACA0012staticaerofoildataforthedragcoefficientwasusedbutthiswasnotcorrected
fortip/rootloss.
(ii) In Approach B, the aerodynamic parameters predicted by the developed BEM code
HAWT_BEMwerecomparedwiththosederiveddirectlyfromtheexperimentaldata(using
thecomputercodeHAWT_LFIMasdescribedintheprevioussection4.3).Foradescription
ofHAWT_BEM,refertosection3.6.ThisBEMmodelimplementsboththetipandroottip
lossesusingthePrandtlequations(Eqts.3.10).Theliftanddragcoefficientsweremodelled
using the same method as for the calculations with HAWT_LFIM, i.e. using the unsteady
aerofoilmodelofsection4.3.2fortheliftcoefficient(Cl,2D)and2DNACA0012staticaerofoil
data for the drag coefficient. In the modelling, the blades were descretized using 22
elements.SkewedwakeeffectsinyawedrotorconditionsweremodelledbyusingGlauerts
model(Eqt.3.28)andColemansequation(givenintable3.1)fortheKfactor.
4.4.1ResultsfromApproachA
AxialConditions
Fig.4.99comparesparametersFA1withFA2for =00.Thetwoparametersareverycloseto
each other, proving the reliability of theBEM equation (Eqt. 3.21) in axial conditions.The
mean values obtained over one whole rotor revolution are plotted and the error bars
representonestandarddeviationsinthedataoveronewholerevolution.
YawedConditions
Figs.4.100and4.101compareFA1andFA2at =300and450.Sinceunsteadyconditionsare
considered due to rotor yaw, the results are being plotted as a function of rotor azimuth
angle (). The error bars represent the uncertainties in both FA1 and FA2 due the 8%
uncertaintyinthehotfilmmeasurementsforwa.Itiseasilynotedthattheseerrorbarsare
considerablywideandthisindicatesthat,likethederivedaerodynamicloads,parameters
FA1 and FA2 are also very sensitive to the errors in the hotfilm measurements. Thiscauses
ambiguity in comparing FA1 and FA2, as may be easily observed in Figs. 4.100 and 4.101
unless the errors in the inflow measurements are kept very small. Yet, a considerable
discrepancy which increases with yaw angle may still be observed between these two
parameters, demonstrating the limitation of the momentum equation (Eqt. 3.6a) when
modellingyawedconditions.Actuallythislimitationyieldsanincorrectpredictionforthe
183
Chapter4 AerodynamicAnalysisoftheTUDelftModelTurbine
30
25
20
FA1 or FA2
15
10
FA1 mean
5 FA2 mean
0
0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
r/R
Figure4.99ComparisonofBEMparametersFA1andFA2at=00.
azimuthallyaveragedaxialinducedvelocity(ua)inBEMcodeswhenEqt.3.6aisused.The
discrepancy between FA1 and FA2 should be corrected if BEM predictions in yaw are to be
improved. One way of doing so is to include another correction parameter ka to the BEM
equationsuchthat
c B 1 B 1
Vrel CL ( , Re ) V Cos V Sin + Vrel CD ( , Re) V Cos + V Sin
2 r b =0 b=0
(4.84)
kaisafunctionofseveralparameters,includingrotorgeometry,operatingconditions,radial
location(r/R)androtorazimuthposition().
In stateofart BEMbased design codes, the implemented correction models for skewed
wake effects in yaw (Type I engineering models described in Chapter 3, page 35) only
correct the ratio of local blade element induced velocity to the annular average induced
velocity. No correction is applied to the annular averaged induced velocity. The fact that
Figs. 4.100 and 4.101 indicate a considerable discrepancy between FA1 and FA2 shows that
correctingonlythisratiomaybeinsufficient,evenwhentreatingattachedflowconditions
(lowanglesofattack).Theinclusionofparameterkaasin Eqt.4.84shouldresultinbetter
estimates for ua and thus improve BEM predictions. An engineering model for this
parameter may be derived from experiments that include both unsteady inflow
measurements (to obtain the inflow at the rotorplane) and unsteady aerodynamic load
measurements.FA1andFA2maythenbefoundfromthemeasurementstoestimatekausing
184
Chapter4 AerodynamicAnalysisoftheTUDelftModelTurbine
30 35
25 30
25
20
FA1 and FA2
0 0
0 30 60 90 120 150 180 210 240 270 300 330 360 0 30 60 90 120 150 180 210 240 270 300 330 360
( deg ) ( deg )
30 35
25 30
20 25
FA1 and FA2
20
15
15
10
FA1 r/R = 0.6 10
FA1 r/R = 0.9
FA2 r/R = 0.6
5
5 FA2 r/R = 0.9
0 0
0 30 60 90 120 150 (180 deg )
210 240 270 300 330 360 0 30 60 90 120 150 180 210 240 270 300 330 360
( deg )
Figure4.100ComparisonofBEMparametersFA1andFA2at=300.
185
Chapter4 AerodynamicAnalysisoftheTUDelftModelTurbine
25 30
25
20
20
15
FA1 and FA2
0 0
0 30 60 90 120 150 180 210 240 270 300 330 360 0 30 60 90 120 150 180 210 240 270 300 330 360
( deg ) ( deg )
25 30
20 25
20
15
15
10
10
FA1 r/R = 0.8
FA1 r/R = 0.5
5
FA2 r/R = 0.5
5 FA2 r/R = 0.8
0 0
0 30 60 90 120 150 180 210 240 270 300 330 360 0 30 60 90 120 150 180 210 240 270 300 330 360
( deg ) ( deg )
25 25
20 20
FA1 and FA2
FA1 and FA2
15 15
10 10
0
0
0 30 60 90 120 150 180 210 240 270 300 330 360 0 30 60 90 120 150 180 210 240 270 300 330 360
( deg ) ( deg )
Figure4.101ComparisonofBEMparametersFA1andFA2at=450.
186
Chapter4 AerodynamicAnalysisoftheTUDelftModelTurbine
theformula
(FA1 ) ,i
(ka ) ,i =
(FA2 ) ,i
(4.85)
Alternatively, an engineering model may be derived with the help of more advanced
aerodynamicmodelssuchasvortexmodelsandCFD.
InthisstudyontheTUDelftrotor,anattemptwasmadetoderivereasonablevaluesforka
at=300and450fromtheresultsforFA1andFA2plottedinFigs.4.100and4.101(i.e.fromthe
resultsderivedfromtheinflowmeasurements).Eqt.4.85wasappliedandkawasfoundto
be relatively constant with the rotor azimuth angle (). At each yaw angle, an averaged
value for ka was obtained by averaging both azimuthally and radially. The values are
plotted in Fig. 4.102. The uncertainties in ka due to the errors in the inflow measurements
arealsoincluded.Itisnotedthattheseuncertaintiesareverylargeandthusmakeitvery
difficulttoestablishrealisticvaluesforka.At=00,FA1andFA2arenearlyequal(seeFig.4.99)
and thus ka is very close to unity in accordance with Eqt. 4.85. One can still observe that
largeryawanglescausethevalueofkatodecreaseandthisreflectstheincreaseddeficiency
ofthemomentumequation(Eqt.3.6a)whenincreasingtherotoryawangle.
1.2
0.8
ka
0.6
0.4
0.2
0
0 10 20 30 40 50
(deg)
Figure 4.102 Variation of ka with yaw angle for the TUDelft rotor with and tip maintained
constantat8and20,respectively.
187
Chapter4 AerodynamicAnalysisoftheTUDelftModelTurbine
4.4.2ResultsfromApproachB
TheresultsfromBEMcodeHAWT_BEMarenowcomparedwiththoseobtainedfromthe
hotfilminflowmeasurementsandunsteadyaerofoilmodelusingHAWT_LFIM.
AxialConditions
Thissectionpresentsthecomparisonfor =00.InthecaseoftheresultsfromHAWT_LFIM
(denoted in graphs by Exp), the mean values are plotted (mean over one whole rotor
revolution),andtheerrorbarsrepresentthecorrespondingonestandarddeviations.The
resultsfromHAWT_BEMaredenotedinthegraphsbyBEM.
Figs. 4.103 and 4.104 compare the distributions of the axial induction factor at the blade
liftingline(a1,c=ua,c/U)andtheazimuthallyaveragedvalues(annularaveraged,i.e.a1=ua/U),
respectively.Thecorrelationfora1isverygoodevenatthesectionsclosetothebladeroot
and tip (40 and 90%R), see Fig. 4.104. However the correlation for a1,c is not good and the
discrepancybetweentheexperimentalandBEMresultsincreasestowardsthebladetipand
root,seeFig.4.103.IntheBEMmodelfor =00,a1,ciscomputedfroma1usingtheequation
a1,c = a1/f where f is the Prandtl tip/root loss factor in accordance with Eqts. 3.10 and 3.27
(takingparameterFsaequaltounitysincewearedealingwithaxialflow).Thisindicatesthat
the discrepancy in Fig. 4.103 is due to f. Also good agreement is achieved between the
distribution of f predicted by the BEM model and that derived from the inflow
measurements (i.e. using Eqts. 3.10 with the inflow angle () obtained from the inflow
measurements), see Fig. 4.105. It may therefore be concluded that the discrepancy in Fig.
4.103isduetotheintrinsicdeficiencyofthePrandtltip/rootlossmodelitself.Thismodel
corrects for the decreased aerodynamic loading at the blade tip and root regions by
artificially increasing a1,c here to reduce the local angle of attack and thus also reduce the
local 2D lift and drag coefficients. This is physically not accurate since it results in an
incorrectpredictionfortheaxialinductionfactoratthebladetipandroot.Itisimportantto
emphasizethefactthatinmanystateofartBEMdesigncodes,2Daerofoildataisstillused
atthebladetipandrootregions.Howeverrecallthefactthat,inthisstudywiththeTUDelft
rotor, an unrealistically high bound circulation resulted at the blade tip and root region
when using an angle of attack derived directly from the near wake inflow measurements
andapplyinga2Dliftcoefficient(refertosection4.3.6,B.1andFig.4.53).Thissuggeststhat
forBEMcodestopredictmoreaccuratelytheinductionatthebladetipandrootwhileat
thesametimemodellingtheloadingdistributionatthetipandrootcorrectly,modified3D
aerofoildatashouldbeusedinstead.Thus,ratherthanartificiallyincreasingtheinduction
locallytoreducetheangleofattack,abettertip/rootlosscorrectionmodelshouldmodify
the2Daerofoildatato3Dvaluesinawaytoreducetheloadingatthetipandroot.
Figs.4.106,4.107and4.108comparethedistributionsofVr, ,andCl.Excellentagreement
wasobtainedinVratallradiallocationsalongtheblades,butthisisonlyduetothesimple
188
Chapter4 AerodynamicAnalysisoftheTUDelftModelTurbine
reason that r>>U and therefore the error in the computed induction has negligible
influenceonVr.Stillitisobservedthatthereisanotabledisagreementatthebladetipand
root in both and Cl. Again this results from the inadequacy of the Prandtl tip/root loss
correction. It is noted that this correction overpredicts the lift coefficient at the tips. Fig.
4.109 and 4.110 compare the axial thrust and torque loading distributions. This over
predictionoftheliftcoefficientatthetipsresultsinahightiploading.
It can be concluded from this analysis that for axial conditions in attached flow over the
blades,thepredictionsbyHAWT_BEMcorrelatedverywellwiththosederivedfromthe
0
0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
-0.05
-0.15
-0.2
a1,c
-0.25
-0.3
-0.35
-0.4
-0.45
r/R
Figure4.103Comparisonofa1,cat=00.
0
0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
-0.05
Exp BEM
-0.1
-0.15
a1
-0.2
-0.25
-0.3
-0.35
r/R
Figure4.104Comparisonofa1at=00.
189
Chapter4 AerodynamicAnalysisoftheTUDelftModelTurbine
experiments at the middle blade sections. However this cannot be stated for the tip/root
regions due to the deficiency of the Prandtl correction. The significance of this deficiency
depends on the aspect ratio of the blades. The higher the blade aspect ratio, the less
influentialthisdeficiencywillbe.
1.2
0.8
f
0.6
0.4
Exp BEM
0.2
0
0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
r/R
Figure4.105ComparisonofPrandtltiplossfactorat=00.
50
45
40
35
30
Vr (m/s)
25
20
15
10
Exp BEM
5
0
0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
r/R
Figure4.106ComparisonofVrat=00.
190
Chapter4 AerodynamicAnalysisoftheTUDelftModelTurbine
8
Exp BEM
7
(deg)
5
0
0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
r/R
Figure4.107Comparisonofat=00.
0.9
0.8
Exp BEM
0.7
0.6
0.5
Cl
0.4
0.3
0.2
0.1
0
0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
r/R
Figure4.108ComparisonofClat=00.
191
Chapter4 AerodynamicAnalysisoftheTUDelftModelTurbine
30
Exp BEM
25
20
dT (N/m)
15
10
0
0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
r/R
Figure4.109ComparisonofdTat=00.
1.2
Exp BEM
1
0.8
dQ (N/m)
0.6
0.4
0.2
0
0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
r/R
Figure4.110ComparisonofdQat=00.
192
Chapter4 AerodynamicAnalysisoftheTUDelftModelTurbine
YawedConditions
Thissectionpresentsthecomparisonfor=300and450.Theresultsareplottedasafunction
oftheblade/rotorazimuthangle().TheresultsfromHAWT_LFIMaredenotedingraphs
by Exp. The error bars in these plots denote the uncertainty resulting from the assumed
errors in the inflow measurements (i.e. 0.26m/s at =300 and 0.21m/s at =450). The
resultsfromHAWT_BEMaredenotedbyBEM.
Figs.4.111and4.112comparetheresultsfora1,cat =300and450,respectively.Remember
that in the BEM calculations, skewed wake effects were modelled using Glauerts model
(Eqt.3.23)withtheColemanmodelforthefactorK.Thismodelyieldsasinusoidalvariation
ofa1,cwith withtheaxialinductionreachingthemaximumandminimumnegativevalues
at900and2700respectively.AqualitativecomparisonwiththemeasurementsinFigs.4.111
and 4.112 clearly demonstrates large differences between Glauerts predictions and the
experimentalresults.Thesedifferencesarelargesttowardstheinboardregionsoftheblades
and are mainly due to root vorticity which in reality causes the induction at the inboard
bladesectionstobehigheratbladepositions180<<3600.Asalreadyoutlinedinsection3.5,
Glauerts model only accounts for the tip vorticity in skewed flows and excludes the
presenceofrootvorticity.Howeveronthesubjectrotor,itwasfoundthatthehighnegative
induction measured at the inboard blade sections for blade positions 180<<3600 was not
onlyduetothepresenceofconsiderablerootvorticitybutalsobecauseofinfluencesofflow
obstructionformthecentrebodyoftheturbinemodel.Duetothecomplexgeometryofthis
centrebodystructure,itwasverydifficulttomodelsuchinfluencesinHAWT_BEM.
Figs.4.113and4.114comparetheresultsfora1.ItisobservedthattheBEMcodetendsto
under predict the value of a1 at almost all radial locations both at =300 and 450. The
disagreement with the experimental values is not only a consequence of the deficiency of
Glauerts equation but also due to the limitation of the axial momentum Eqt. 3.6a when
treating yawed conditions. The issue that the deficiency in this equation brings about an
inaccuratepredictionforua(andhencealsoina1)hasalreadybeenbroughtforwardearlier
in section 4.4.1. Better predictions for a1 could have been achieved but modifying the
momentum equation for thrust as Eqt. 4.84 with a suitable engineering model for ka. This
wasalreadyshowninanearlierstudydocumentedin[67].
An incorrect prediction for the induction factors obviously results also from the
inappropriate Prandtl tip/root correction. This issue was brought forward earlier in this
sectionwhencomparingthe=00.
Incorrect predictions for a1,c and a1 yield incorrect predictions for the angles of attack,
aerofoil coefficients and also for the blade loading distributions. Figs. 4.115 Figs. 4.122
comparetheresultsfortheunsteadyparameters ,Cl,dTanddQfromHAWT_BEMwith
those derived from the measurements using HAWT_LFIM. Unfortunately the large
uncertaintiesintheresultsderivedfromthemeasurements(resultingfromuncertaintiesin
193
Chapter4 AerodynamicAnalysisoftheTUDelftModelTurbine
thehotfilmmeasurements)makeitverydifficulttoquantifythediscrepanciesbetweenthe
results of HAWT_BEM and HAWT_LFIM. However it is clear that, given that BEM codes
incorporate accurate engineering models for skewed wake effects (i.e. better models for
parameterFsainEqt.3.27)togetherwithengineeringmodelstocorrectforthedeficiencyof
the momentum equation inyaw (i.e. the inclusion of model for parameter kain Eqt. 4.84),
theirreliabilitywillbeimprovedwhenmodellingyawedconditions.Yetweshouldkeepin
mindthefactthatinthisstudyontheTUDelftrotor,wearelimitingourselvestoattached
flowconditionsonlyforwhichthetimedependentanglesofattackatthebladesareknown
to be small. For such conditions, the aerofoil data used in the calculations is known to be
reasonablyaccurate.Whendealingwithturbineoperatingconditionsinwhichtheanglesof
attack are large (i.e. where stalldelay and dynamic stall take place), the reliability of the
aerofoildatausedinBEMcodesalsobecomesanimportantissue.
194
Chapter4 AerodynamicAnalysisoftheTUDelftModelTurbine
0 0
0 30 60 90 120 150 180 210 240 270 300 330 360 0 30 60 90 120 150 180 210 240 270 300 330 360
-0.05 -0.05
Exp r/R = 0.4 BEM r/R = 0.4 Exp r/R = 0.7 BEM r/R = 0.7
-0.1
-0.1
-0.15
a1,c (deg)
a1,c (deg)
-0.15
-0.2
-0.2
-0.25
-0.25
-0.3
-0.35 -0.3
-0.4 -0.35
( deg ) ( deg )
0 0
0 30 60 90 120 150 180 210 240 270 300 330 360 0 30 60 90 120 150 180 210 240 270 300 330 360
-0.05 -0.05
Exp r/R = 0.5 BEM r/R = 0.5 Exp r/R = 0.8 BEM r/R = 0.8
-0.1
-0.1
-0.15
a1,c (deg)
a1,c (deg)
-0.15
-0.2
-0.2
-0.25
-0.25
-0.3
-0.35 -0.3
-0.4 -0.35
( deg ) ( deg )
0 0
0 30 60 90 120 150 180 210 240 270 300 330 360 0 30 60 90 120 150 180 210 240 270 300 330 360
-0.05 -0.05
Exp r/R = 0.6 BEM r/R = 0.6 Exp r/R = 0.9 BEM r/R = 0.9
-0.1 -0.1
-0.15 -0.15
a1,c (deg)
a1,c (deg)
-0.2 -0.2
-0.25 -0.25
-0.3 -0.3
-0.35 -0.35
-0.4 -0.4
( deg ) ( deg )
Figure4.111Comparisonofa1,cat=300.
195
Chapter4 AerodynamicAnalysisoftheTUDelftModelTurbine
0 0
0 30 60 90 120 150 180 210 240 270 300 330 360 0 30 60 90 120 150 180 210 240 270 300 330 360
-0.05 Exp r/R = 0.4 BEM r/R = 0.4 -0.05
Exp r/R = 0.7 BEM r/R = 0.7
-0.1
-0.1
-0.15
a1,c
a1,c
-0.15
-0.2
-0.2
-0.25
-0.25
-0.3
-0.35 -0.3
( deg ) ( deg )
0 0
0 30 60 90 120 150 180 210 240 270 300 330 360 0 30 60 90 120 150 180 210 240 270 300 330 360
-0.05 Exp r/R = 0.5 BEM r/R = 0.5
-0.05 Exp r/R = 0.8 BEM r/R = 0.8
-0.1
-0.1
-0.15
a1,c
a1,c
-0.2
-0.15
-0.25
-0.2
-0.3
-0.35 -0.25
( deg ) ( deg )
0 0
0 30 60 90 120 150 180 210 240 270 300 330 360 0 30 60 90 120 150 180 210 240 270 300 330 360
-0.05 Exp r/R = 0.6 BEM r/R = 0.6 -0.05
Exp r/R = 0.9 BEM r/R = 0.9
-0.1 -0.1
a1,c
-0.15
a1,c
-0.15
-0.2 -0.2
-0.25 -0.25
-0.3 -0.3
( deg ) ( deg )
Figure4.112Comparisonofa1,cat=450.
196
Chapter4 AerodynamicAnalysisoftheTUDelftModelTurbine
0 0
0 30 60 90 120 150 180 210 240 270 300 330 360 0 30 60 90 120 150 180 210 240 270 300 330 360
-0.05 -0.05
Exp r/R = 0.4 BEM r/R = 0.4 Exp r/R = 0.7 BEM r/R = 0.7
-0.1
-0.1
-0.15
-0.15
a1
a1
-0.2
-0.2
-0.25
-0.25
-0.3
-0.3 -0.35
( deg ) ( deg )
0 0
0 30 60 90 120 150 180 210 240 270 300 330 360 0 30 60 90 120 150 180 210 240 270 300 330 360
-0.05 -0.05
Exp r/R = 0.5 BEM r/R = 0.5 Exp r/R = 0.8 BEM r/R = 0.8
-0.1 -0.1
-0.15 -0.15
a1
a1
-0.2 -0.2
-0.25 -0.25
-0.3 -0.3
-0.35 -0.35
( deg ) ( deg )
0 0
0 30 60 90 120 150 180 210 240 270 300 330 360 0 30 60 90 120 150 180 210 240 270 300 330 360
-0.05 -0.05
Exp r/R = 0.6 BEM r/R = 0.6 Exp r/R = 0.9 BEM r/R = 0.9
-0.1 -0.1
-0.15 -0.15
a1
a1
-0.2 -0.2
-0.25 -0.25
-0.3 -0.3
-0.35 -0.35
( deg ) ( deg )
Figure4.113Comparisonofa1at=300.
197
Chapter4 AerodynamicAnalysisoftheTUDelftModelTurbine
0 0
0 30 60 90 120 150 180 210 240 270 300 330 360 0 30 60 90 120 150 180 210 240 270 300 330 360
-0.05
-0.05
Exp r/R = 0.4 BEM r/R = 0.4 Exp r/R = 0.7 BEM r/R = 0.7
-0.1
-0.1
-0.15
a1
a1
-0.15
-0.2
-0.2
-0.25
-0.25 -0.3
( deg ) ( deg )
0 0
0 30 60 90 120 150 180 210 240 270 300 330 360 0 30 60 90 120 150 180 210 240 270 300 330 360
-0.05
-0.05
Exp r/R = 0.5 BEM r/R = 0.5 Exp r/R = 0.8 BEM r/R = 0.8
-0.1
-0.1
-0.15
a1
a1
-0.15
-0.2
-0.2
-0.25
-0.25 -0.3
( deg ) ( deg )
0 0
0 30 60 90 120 150 180 210 240 270 300 330 360 0 30 60 90 120 150 180 210 240 270 300 330 360
-0.1 -0.1
a1
a1
-0.15 -0.15
-0.2 -0.2
-0.25 -0.25
( deg ) ( deg )
Figure4.114Comparisonofa1at=450.
198
Chapter4 AerodynamicAnalysisoftheTUDelftModelTurbine
12 5
Exp r/R = 0.7 BEM r/R = 0.7
10 Exp r/R = 0.4 BEM r/R = 0.4
4
8
3
(deg)
(deg)
6
2
4
2 1
0 0
0 30 60 90 120 150 180 210 240 270 300 330 360 0 30 60 90 120 150 180 210 240 270 300 330 360
( deg ) ( deg )
8 5
7 Exp r/R = 0.5 BEM r/R = 0.5 Exp r/R = 0.8 BEM r/R = 0.8
4
6
5
3
(deg)
(deg)
4
3 2
2
1
1
0 0
0 30 60 90 120 150 180 210 240 270 300 330 360 0 30 60 90 120 150 180 210 240 270 300 330 360
( deg ) ( deg )
6 5
4
3
(deg)
(deg)
2
2
1
1 Exp r/R = 0.9 BEM r/R = 0.9
0 0
0 30 60 90 120 150 180 210 240 270 300 330 360 0 30 60 90 120 150 180 210 240 270 300 330 360
( deg ) ( deg )
Figure4.115Comparisonofat=300.
199
Chapter4 AerodynamicAnalysisoftheTUDelftModelTurbine
9 4.5
Exp r/R = 0.4 BEM r/R = 0.4 Exp r/R = 0.7 BEM r/R = 0.7
8 4
7 3.5
6 3
(deg)
(deg)
5 2.5
4 2
3 1.5
2 1
1 0.5
0 0
0 30 60 90 120 150 180 210 240 270 300 330 360 0 30 60 90 120 150 180 210 240 270 300 330 360
( deg ) ( deg )
7 4
Exp r/R = 0.5 BEM r/R = 0.5
6 3.5
3
5
2.5
4
(deg)
(deg)
2
3
1.5
2
1
0 0
0 30 60 90 120 150 180 210 240 270 300 330 360 0 30 60 90 120 150 180 210 240 270 300 330 360
( deg ) ( deg )
5.5 4
5 Exp r/R = 0.6 BEM r/R = 0.6
3.5
4.5
4 3
3.5
2.5
(deg)
3
(deg)
2
2.5
2 1.5
1.5
1
1 Exp r/R = 0.9 BEM r/R = 0.9
0.5 0.5
0 0
0 30 60 90 120 150 180 210 240 270 300 330 360 0 30 60 90 120 150 180 210 240 270 300 330 360
( deg ) ( deg )
Figure4.116Comparisonof at=450.
200
Chapter4 AerodynamicAnalysisoftheTUDelftModelTurbine
0.9 0.5
0.8 0.45
0.7 0.4
0.35
0.6
0.3
Cl 0.5
Cl
0.25
0.4
0.2
0.3
0.15
0.2
0.1 Exp r/R = 0.7 BEM r/R = 0.7
0.1 Exp r/R = 0.4 BEM r/R = 0.4 0.05
0 0
0 30 60 90 120 150 180 210 240 270 300 330 360 0 30 60 90 120 150 180 210 240 270 300 330 360
( deg ) ( deg )
0.8 0.45
0.7 0.4
0.35
0.6
0.3
0.5
0.25
Cl
Cl
0.4
0.2
0.3
0.15
0.2 Exp r/R = 0.8 BEM r/R = 0.8
0.1
Exp r/R = 0.5 BEM r/R = 0.5
0.1 0.05
0 0
0 30 60 90 120 150 180 210 240 270 300 330 360 0 30 60 90 120 150 180 210 240 270 300 330 360
( deg ) ( deg )
0.6 0.4
0.5 0.35
0.3
0.4
0.25
Cl
0.3
Cl
0.2
0.2 0.15
0 0
0 30 60 90 120 150 180 210 240 270 300 330 360 0 30 60 90 120 150 180 210 240 270 300 330 360
( deg ) ( deg )
Figure4.117ComparisonofClat=300.
201
Chapter4 AerodynamicAnalysisoftheTUDelftModelTurbine
0.8 0.45
0.7 Exp r/R = 0.4 BEM r/R = 0.4 0.4 Exp r/R = 0.7 BEM r/R = 0.7
0.35
0.6
0.3
0.5
Cl 0.25
Cl
0.4
0.2
0.3
0.15
0.2
0.1
0.1 0.05
0 0
0 30 60 90 120 150 180 210 240 270 300 330 360 0 30 60 90 120 150 180 210 240 270 300 330 360
( deg ) ( deg )
0.7 0.4
0.3
0.5
0.25
0.4
Cl
Cl
0.2
0.3
0.15
0.2
0.1
0 0
0 30 60 90 120 150 180 210 240 270 300 330 360 0 30 60 90 120 150 180 210 240 270 300 330 360
( deg ) ( deg )
0.5 0.35
0.45 Exp r/R = 0.6 BEM r/R = 0.6
0.3
0.4
0.35 0.25
0.3
0.2
Cl
0.25
Cl
0.2 0.15
0.15
0.1
0.1
0.05 Exp r/R = 0.9 BEM r/R = 0.9
0.05
0 0
0 30 60 90 120 150 180 210 240 270 300 330 360 0 30 60 90 120 150 180 210 240 270 300 330 360
( deg ) ( deg )
Figure4.118ComparisonofClat=450.
202
Chapter4 AerodynamicAnalysisoftheTUDelftModelTurbine
12 20
18
10
16
14
8
12
dT (N/m)
dT (N/m)
6 10
8
4
6
Exp r/R = 0.4 BEM r/R = 0.4 4 Exp r/R = 0.7 BEM r/R = 0.7
2
2
0 0
0 30 60 90 120 150 180 210 240 270 300 330 360 0 30 60 90 120 150 180 210 240 270 300 330 360
( deg ) ( deg )
16 25
14
20
12
10
15
dT (N/m)
dT (N/m)
8
6 10
4
Exp r/R = 0.5 BEM r/R = 0.5 5 Exp r/R = 0.8 BEM r/R = 0.8
2
0 0
0 30 60 90 120 150 180 210 240 270 300 330 360 0 30 60 90 120 150 180 210 240 270 300 330 360
( deg ) ( deg )
18 30
16
25
14
12 20
dT (N/m)
dT (N/m)
10
15
8
6 10
0 0
0 30 60 90 120 150 180 210 240 270 300 330 360 0 30 60 90 120 150 180 210 240 270 300 330 360
( deg ) ( deg )
Figure4.119ComparisonofdTat=300.
203
Chapter4 AerodynamicAnalysisoftheTUDelftModelTurbine
9 16
8 14
7
12
6
10
dT (N/m)
dT (N/m)
5
8
4
6
3
4
2
Exp r/R = 0.7 BEM r/R = 0.7
1 2
Exp r/R = 0.4 BEM r/R = 0.4
0 0
0 30 60 90 120 150 180 210 240 270 300 330 360 0 30 60 90 120 150 180 210 240 270 300 330 360
( deg ) ( deg )
12 20
18
10
16
14
8
12
dT (N/m)
dT (N/m)
6 10
8
4
6
4
2
Exp r/R = 0.5 BEM r/R = 0.5 Exp r/R = 0.8 BEM r/R = 0.8
2
0 0
0 30 60 90 120 150 180 210 240 270 300 330 360 0 30 60 90 120 150 180 210 240 270 300 330 360
( deg ) ( deg )
14 25
12
20
10
15
dT (N/m)
dT (N/m)
6
10
4
5
2 Exp r/R = 0.6 BEM r/R = 0.6 Exp r/R = 0.9 BEM r/R = 0.9
0 0
0 30 60 90 120 150 180 210 240 270 300 330 360 0 30 60 90 120 150 180 210 240 270 300 330 360
( deg ) ( deg )
Figure4.120ComparisonofdTat=450.
204
Chapter4 AerodynamicAnalysisoftheTUDelftModelTurbine
0.6 0.9
0.8
0.5
0.7
0.4 0.6
dQ (Nm/m)
dQ (Nm/m)
0.5
0.3
0.4
0.2 0.3
0.2
0.1
Exp r/R = 0.4 BEM r/R = 0.4 0.1
Exp r/R = 0.7 BEM r/R = 0.7
0 0
0 30 60 90 120 150 180 210 240 270 300 330 360 0 30 60 90 120 150 180 210 240 270 300 330 360
( deg ) ( deg )
0.8 1
0.9
0.7
0.8
0.6
0.7
0.5
dQ (Nm/m)
dQ (Nm/m)
0.6
0.4 0.5
0.4
0.3
0.3
0.2
0.2
0.1
Exp r/R = 0.5 BEM r/R = 0.5 0.1
Exp r/R = 0.8 BEM r/R = 0.8
0 0
0 30 60 90 120 150 180 210 240 270 300 330 360 0 30 60 90 120 150 180 210 240 270 300 330 360
( deg ) ( deg )
0.9 1.2
dQ (Nm/m)
0.5 0.6
0.4
0.4
0.3
0.2 0.2
0.1
Exp r/R = 0.6 BEM r/R = 0.6 0
0 0 30 60 90 120 150 180 210 240 270 300 330 360
0 30 60 90 120 150 180 210 240 270 300 330 360 -0.2
( deg ) ( deg )
Figure4.121ComparisonofdQat=300.
205
Chapter4 AerodynamicAnalysisoftheTUDelftModelTurbine
0.4 0.6
Exp r/R = 0.4 BEM r/R = 0.4
0.35 0.5 Exp r/R = 0.7 BEM r/R = 0.7
0.3 0.4
0.25 0.3
dQ (N/m)
dQ (N/m)
0.2 0.2
0.15 0.1
0.1 0
0 30 60 90 120 150 180 210 240 270 300 330 360
0.05 -0.1
0 -0.2
0 30 60 90 120 150 180 210 240 270 300 330 360
-0.05 -0.3
( deg ) ( deg )
0.6 0.7
Exp r/R = 0.5 BEM r/R = 0.5 0.6 Exp r/R = 0.8 BEM r/R = 0.8
0.5
0.5
0.4
0.4
0.3
dQ (N/m)
dQ (N/m)
0.3
0.2
0.2
0.1
0.1 0
0 30 60 90 120 150 180 210 240 270 300 330 360
-0.1
0
0 30 60 90 120 150 180 210 240 270 300 330 360 -0.2
-0.1 -0.3
( deg ) ( deg )
0.6 0.7
Exp r/R = 0.6 BEM r/R = 0.6 0.6
0.5
Exp r/R = 0.9 BEM r/R = 0.9
0.5
0.4
0.4
0.3
dQ (N/m)
0.3
dQ (N/m)
0.2 0.2
0.1 0.1
0
0
0 30 60 90 120 150 180 210 240 270 300 330 360
0 30 60 90 120 150 180 210 240 270 300 330 360 -0.1
-0.1
-0.2
-0.2
-0.3
( deg ) ( deg )
Figure4.122ComparisonofdQat=450.
206
Chapter4 AerodynamicAnalysisoftheTUDelftModelTurbine
4.5Conclusions
Wind tunnel measurements were taken on the TUDelftmodel turbine carried out inan
openjet facility with close collaboration with Wouter Haans, a Phd collegue carrying out
research in the same field. Two different measurement campaigns were carried out: (1)
detailedhotfilmmeasurementsinthenearwakeoftherotor,atvariousplanesparallelto
the rotorplane; (2) smoke visualization measurements to track the tipvortex paths of the
rotor wake together with measurements of the rotor axial thrust. A methodology was
appliedforderivingtheunsteadyaerodynamicloadingdistributionsatthebladesfromthe
hotfilm measurements using an unsteady aerofoil model, developed by Leishman [49].
Despite the fact that the analysis was only restricted to attached flow conditions at the
blades(forwhichtheunsteadyaerofoilmodelisconsiderablyaccurate),thismethodology
wasfoundtobequitechallengingtoapplyduetothefollowingdrawbacks:
(1) The derived aerodynamic loads are very sensitive to the errors in the hotfilm
measurements.Inotherwords,asmallpercentageerrorintheinflowmeasurements
mayresultintoasignificantlylargererrorinloading.Twomainconclusionscanbe
made:(a)theuncertaintyinthederivedloadingdistributionsresultingfrominflow
measurement errorsincreaseswithyawangleand(b)theuncertaintyinthetorque
loadingisnormallylargerthanthatfortheaxialthrustloading.Thiscouldbeshown
through a simple analytical analysis and was actually proved when applying an
uncertaintyanalysiswiththehotfilmmeasurementstakenontheTUDelftrotor.The
percentageerrorofthesemeasurementswasestimatedtobeintherange610%and
thiswasmainlyduetothenonuniformityinthetunnelexitjetofthewindtunnel,
positioningerrorsofthetraversingsystem,errorsincalibratingthehotfilmprobes
and errors in the datareduction of the velocity components. When treating axial
conditions,theerrorinthederivedaerodynamicloadingresultingfromerrorsinthe
hotfilm data did not exceed 20%, which is considered to be reasonable. In yawed
conditions the error was larger. The maximum errors in the axial thrust loading at
=300 and =450 were found to be 25% and 35%, respectively. The corresponding
errors in the torque loading were found to exceed 100% and this mainly occurred
whenthisloadingwasclosetozero.
It is important to point out again the fact that due to physical restrictions, it is
impossibletomeasuretheinflowvelocitiesdirectlyattherotorplanewithhotfilms.
In this study, linear interpolation was used to estimate such velocities from the
measurements taken upstream and downstream of the rotor. This introduced an
additional uncertainty in the derived loading which was not included in the
calculations.Thisuncertaintycouldnotbedeterminedasitwasnotpossibletotake
themeasurementsattherotorplanewiththeequipmentavailable.
207
Chapter4 AerodynamicAnalysisoftheTUDelftModelTurbine
(2) A tip/root correction is required. It was found that the bound circulation derived
usingtheemployedunsteadyaerofoilmodelwasunrealisticallyhighatthebladetip
and root regions, even though the angle of attack was estimated directly from the
inflowmeasurements.Thiswasduetothefactthattheunsteadyaerofoilmodelused
is a 2D model and thus does not cater for the highly 3D flow phenomena taking
placeatthebladetipandroot.Theimportanceofthetip/rootcorrectiondependson
the aspect ratio of the blade and also on the blade geometry. For blades having a
smallaspectratioandarectangulartipandrootasthatoftheTUDelftrotor,then3D
effectsaremoreprominentandthiscorrectioncouldnotbeignored.
Inthisstudy,aprescribedwakevortexmodelwasdevelopedtodeterminewhethertunnel
blockageinfluencesweresignificant.Luckily,theseinfluenceswerefoundtobeverysmall.
Recommendationsforfuturework
This work on the TUDelft rotor has revealed the various difficulties associated with
derivingthesteady/unsteadyaerodynamicloaddistributionsonawindturbinebladefrom
detailed hotfilm measurements in the near wake in axial/yawed conditions. In certain
cases,bladepressuremeasurementsarenotpossibleandalternativelytheloadshavetobe
derived from the near wake measurements. The detailed hotfilm measurements in this
study have proved to be very time consuming and the following recommendations from
lessonslearntarebeingmadeforfuturework:
The rotor should have a high aspect ratioto minimize theuncertainty in the tip/root
losscorrection.
The uncertainty in the measurements should be kept very low (<3% is being
recommended)sothaterrorsinthederivedloadingareminimized.Itisadvisableto
use higher windspeeds for the same tip speed ratio to avoid having small flow
velocitiesatthebladesthatcouldyieldlargepercentageerrors.
The hotfilm measurement equipment should be automated as much as possible. In
this way it would be possible to take more repetitive measurements for the same
periodoftimeavailableforthewindtunneltesting.
ItshouldbeconsideredtouseParticleImagingVelocimetry(PIV)insteadofhotfilm
anemometrysincetheformerhasthecapabilitytomeasuretheinflowdirectlyatthe
rotorplane.PIVisveryefficientsincemeasurementsatdifferentpointsmaybetaken
simultaneouslyandveryquickly.
InvestigationofthelimitationsofBEMCodes
TheresultsderivedfromtheinflowmeasurementsontheTUDelftrotorwereusedtocarry
outadetailedinvestigationofatypicalBEMcode(HAWT_BEM)whenmodellingaxialand
208
Chapter4 AerodynamicAnalysisoftheTUDelftModelTurbine
yawedturbines.Despitethatattachedflowconditionswereonlybeingtreatedandalsothe
factthattheresultsderivedfromtheinflowmeasurementshadahighlevelofuncertainty
in general, it was still possible to have a better understanding of the limitations of BEM
based design codes and obtain further intuition of how these may be improved. Various
conclusionscouldbedrawn,butthesewillbediscussedinChapter7,togetherwithother
conclusions drawn from a similar investigation on the NREL Phase VI rotor described in
Chapter6.
209
Chapter4 AerodynamicAnalysisoftheTUDelftModelTurbine
210
Chapter5DevelopmentofaFreeWakeVortexModel
5.DevelopmentofAFreewakeVortexModel
5.1Introduction
A special class of rotor aerodynamic models that are less computationally demanding
thanCFDtechniquesbutaremorereliablethanBEMmethods,arethesocalledfreewake
vortex methods. These methods are based on the principle that, for flows that may be
assumedtobeincompressibleandinviscid,vorticityformedatthebladesisconvectedinto
thewakeastrailingandshedvorticity(asshowninFig.1.4)withalocalvelocitythatisthe
vectorial sum of the free stream velocity and that induced by all vorticity sources in the
wakeandfromtheblades.Thesemethodsaretypicallyunsteadyinnature:vorticityinthe
wake is allowed to diffuse freely and the evolution of the wake is calculated in time.
Circulationinthewakeismodelledbyaseriesofvortexfilamentsthatmaytaketheformof
lines(Afjehetal.[1],Bareietal.[6],Leishmanetal.[51],Garrel[26])orparticles(Leeetal.
[46],Voutsinasetal.[101]).Circulationaroundthebladesismodelledwithaliftinglineor
lifting surface representation. The induced velocity at different points in the wake is
computedusingtheBiotSavartlaw.
In this project, a new freewake vortex model, named HAWT_FWC (FWC meaning Free
WakeCode),wasdeveloped.Themodelisapplicabletobothaxialandyawedconditions.It
was specifically designed to model HAWT rotor wakes from knowledge of the
aerodynamicloadsattheblades.Unlikeotherfreewakemodels,itdoesnotdirectlyrelyon
theavailabilityofaerofoildatatoiterativelydeterminethebladeloading.Theinputtothis
code is a prescribed spanwise distribution of bound circulation that may be time
dependent. From this prescription, the code generates a wake and then calculates the 3D
induced velocities at different points in the flow field of the rotor. As already outlined in
Chapter 2, the main reason for developing this code was to use it in the novel approach
beingproposedforderivingtheangleofattackdistributionsinHAWTsfrombladepressure
measurements.ThisnovelmethodwillbepresentedindetailedinChapter6.
Thischapterisorganizedintwosections:
A. Section5.2describesthenumericalmodelimplementedinthefreewakecode
HAWT_FWC,togetherwiththeprogramstructure.
B. Section5.3describestheverificationandvalidationworkcarriedouton
HAWT_FWCusingtheexperimentaldataobtainedfortheTUDelftrotor.
211
Chapter5DevelopmentofaFreeWakeVortexModel
5.2FreewakeNumericalModel
5.2.1BladeModel
Inthismodel,eachrotorbladeisrepresentedbyaliftinglineconsistingofafixednumber
of piecewise constant spanwise segments located at the quarterchord location (c/4). The
arrangementofthesegmentationisillustratedinFig.5.1.Thenumberofsegmentsisequal
tonwhileiisanindexrepresentingeachparticularsegment.Animportantrequirementof
this piecewise constant representation is that the spanwise segments be small enough so
that any variation in the prescribed bound circulation, blade chord and twist is
approximatelyrepresentedbythestraightlinesegmentsbetweeniandi=i+1.
rn1=Rt
ri+1
ri
ri1
r
r0=Rr
Rotoraxis
i=0
i=1
i=2
i=3
i = n-3
i = n-2
i = n-1
.
.
.
.
Prescribedbound
circulationdistribution
n-4
n-3
n-2
B2
B1
B0
.
.
.
.
Rotoraxis
Figure5.1DiscretizationofBladeandBoundCirculationDistribution.
212
Chapter5DevelopmentofaFreeWakeVortexModel
Acosineradialsegmentationisadoptedtousesmallersegmentstowardsthebladetipand
rootsuchthat
Rr + Rt Rt Rr
ri = cos ( i ) (5.1)
2 2
i +1 = i + where 0 i n 1; 0 = 0 rad (5.2)
n 1
This cosine segmentation increases the numerical accuracy in vortex codes [26, 90]
especiallytowardsthebladetipandroot,wheretheboundcirculationvariesrapidlywith
radialposition.
5.2.2NearWakeModel
Thenearwakeismodelledasvortexsheets,oneforeachblade.Figure5.2illustrateshow
the each vortex sheet is discretized by means of straightline vortex filaments, the latter
being interconnected by nodes to form vortex nodes. The time step is denoted by m. a
representsthevortexage.Filamentshavinga=0aretheoldestfilamentswhilethosehaving
a=m are the youngest ones that have just formed at the blades at time step m. At a given
timestep,eachnodeinthewakeisidentifiedbyadualnumberingsystem(i,a).Thenumber
of trailing vortex filaments emerging from each blade is equal to the number of blade
stations(n). T representsthetrailingcirculationinthewakeduetospanwisevariationin
bound circulation at the blades. Each trailing vortex filament is attached to two nodes
numbered(i,a)and(i,a+1),(seeFig.5.2).Whentheturbineisoperatingwithitsshaftparallel
to a uniform windspeed, the trailing vortex segments from the same blade radius have a
constant strength downstream. This is not the case when the turbine is operating in
unsteady environments. S represents the shed circulation resulting from the time
dependentvariationinboundcirculationexperiencedinunsteadyenvironments.Eachshed
vortexfilamentisattachedtotwonodesnumbered(i,a)and(i+1,a).
TosatisfytheKuttacondition,thewakemustbeshedatthetrailingedgeofablade.Thus
thewakenodesshouldleavethebladesfromthetrailingedges,fromwhichtheyadvance
aftereachtimestepwaydownstreamwithalocalvelocitythatisequaltothevectorialsum
ofthefreestreamvelocityandthevorticityinducedvelocity.Ateachtimestep,anewsetof
trailing and shed vortex filaments are modelled to be shed from the blade trailing edge.
Thesefilamentsarenumericallyrepresentedasfollows:
213
Chapter5DevelopmentofaFreeWakeVortexModel
The wake filaments are convected downstream depending on the absolute velocity of the
nodestowhichtheyareconnected.Thetotalinducedvelocityatagivennodeisduetothe
bound,trailingandshedvorticesfromallrotorblades.
Z,Za
i = n-2
Wake
i =n-1
i=1
i=4
i=0
i=2
i=3
nodes
a=0
T S
Winddirection Ya
a=1
O
Y
(i,a)
a=2
(i+1,a)
(0,0)
(i,a+1) a=3
B
S T
X
a = m-1
RotorAxis
Xa
a=m
BladeLiftingLine
Blade
nodes
BladeTip
Figure5.2Modellingofbladesandnearwakeinfreewakevortexmodel.
The induced velocity induced at each node due to a vortex is computed from the Biot
Savartequation:
r r
r ds r
du = r 3 (5.4)
4 r
Treatmentofviscouseffectsinthenearwakemodel
Viscous effects in rotor wakes usually occur at much smaller scales than potential flow
phenomena. However, these effects may be significant in yawed conditions due to the
formation of complex and unsteady bladetowake interactions and skewed wake effects.
Leishman et al. [51] modify the BiotSavart equation to account for viscous effects by
introducingaviscousparameter,KV,suchthat
r r
r ds r
du = K V r 3 (5.5)
4 r
214
Chapter5DevelopmentofaFreeWakeVortexModel
2
h
where KV = (5.6)
( )
1
rc + h
2z 2z z
Eqt.5.6isaviscouscoremodelinwhichtheinnerpartofthevortex,denotedasthecore
region (r < rc), almost rotates as a solid body. The outer region (outside the core, r > rc)
almostbehavesasapotentialflow.histheperpendiculardistanceoftheevaluationpoint
from the vortex element while rc is the viscous core radius of the vortex. In effect, KV
r
desingularizes the BiotSavart equation as r tends to zero. Fig. 5.3 illustrates how this
vortexcoremodelmodifiestheswirlvelocitydistributionpredictedbytheBiotSavartlaw
closetothefilament.
1.6
rc=0.2m
0.8
rc=0.1m
0.4
swirl velocity
0
-1 -0.8 -0.6 -0.4 -0.2 0 0.2 0.4 0.6 0.8 1
-0.4
-0.8
-1.2
-1.6
distance from vortex axis
Figure5.3ModificationoftheswirlvelocitydistributionatavortexfilamentpredictedbytheBiot
Savartlawusingacoremodel(Eqt.5.6)toaccountforviscouseffects.
In Eqt. 5.6, parameter z defines the form of velocity profile. If z = 1, the Scully model is
obtainedandifz=2thenanalgebraicapproximationtotheLambOseenvortexmodelis
obtained (refer to references [49], [51]). In Leishmans freewake model [51], the wake is
modelled by using a single concentrated helical tip vortex per blade. In HAWT_FWC, the
wake is modelled by vortex sheets from each blade each consisting of a mesh of straight
linefilamentsandEqt.5.6isappliedtoeachofthetrailingandshedfilaments.TheLamb
Oseenviscouscoremodelisused.
The viscous diffusion of vorticity in this freewake model is handled by applying a core
growthmodel.AcoregrowthmodelwasadoptedfromAnanthanetal.[2]andisappliedto
every trailing and shed vortex filament. The onedimensional laminar NavierStokes
equations reveal that the viscous vortex core radius grows with time in accordance with
([2],[10][49],[51]):
215
Chapter5DevelopmentofaFreeWakeVortexModel
rc (t ) = 4 v t (5.7)
rc (t ) = 4 v (t + S c ) (5.8)
Scisatimeoffsetparameterthatsetsanonzeroviscouscoreradiusforvorticesthatarejust
released from the trailing edge of each blade. Fig. 5.4 shows a typical variation of the
viscous core radius with time as modelled by Eqt. 5.8. In a freewake solution the wake
nodes are allowed to convect freely, thereby causing the vortex filaments to be strained.
This results in a change in the vorticity content of the individual filaments which in turn
modifies the induced velocity field around the vortex core. To account for filament
straining,thecoreradiusestimatedbyEqt.5.8iscorrectedusing:
1
rc = rc (5.9)
eff
1+
where isthevortexfilamentstrain.ThederivationofEqt.5.9ispresentedinAppendixE
andisbasedonHelmhotzsthirdlawstatingthatthenetstrengthofavortexshouldremain
constant.
Coreradius(rc)
Time(t)
Sc
Figure5.4TypicalvariationinthevortexcoreradiusasmodelledbycoregrowthmodelEqt.5.8.
216
Chapter5DevelopmentofaFreeWakeVortexModel
Computationof3Dinducedvelocitiesatapoint
Thewakefilamentstraveldownstreamdependingoftheabsolutevelocityofthenodesto
whichtheyareconnected.Thistotalinducedvelocityatagivennodeisduetothebound,
trailing and shed vortices of all the rotor blades. This velocity has three components and
maybefoundwiththeapplicationoftheBiotSavartequations(seeAppendixC,Eqts.C.24
C.32).Forthisnumericalmodel,thetotalinducedvelocityateachparticularwakenode
whentherotorisattimestepismathematicallyexpressedintheXYZreferenceframein
termsofthefollowingequations:
B 1 n 2
( u )
X bp , ip , ap
= KV * ( Bb )i , * IBX b ,i , ,bp ,ip , ap + ...
b =0 i =0
B 1 n 1 m B 1 n 2 m
...
+
b =0 i =0 a =0
KV ( Tb )i , a ITX b ,i , ,bp ,ip , ap +
b=0 i =0 a =0
KV ( S b )i , a ISX b ,i , ,bp ,ip , ap
B 1 n 2
B 1 n 1 m B 1 n 2 m
+
b =0 i =0 a =0 b =0 i =0 a =0
B 1 n 2
...
+
b =0 i =0 a =0
KV ( Tb )i , a ITZ b ,i , ,bp ,ip , ap + KV ( S b )i , a ISZ b ,i , ,bp ,ip , ap
b =0 i =0 a =0
(5.10)
whereindices(bp,ip,ap)denotetheparticularwakenodeatwhichthetotalinducedvelocity
isbeingevaluated.TheparametersIBX,IBY,IBZarethegeometricinfluencecoefficientsfor
the bound vortices, ITX, ITY, ITZ are the geometric influence coefficients for the trailing
vorticesandISX,ISY,ISZarethegeometricinfluencecoefficientsfortheshedvortices.The
numerical equations for these influence coefficients are given in tables 5.1, 5.2 and 5.3. In
table5.1,XB,YBandZBdenotethenodecoordinatesontheliftinglineofabladebatradial
locationiandrotorazimuthposition .
217
218
( r1 ) + r Z Z Y Y Y Y Z Z
bp , ap ,ip , ,b ,i ( 2 )bp , ap ,ip , ,b ,i ( )bp ,ip ,ap ( B ) ,b ,i +1 ( B ) ,b ,i +1 ( B ) ,b ,i ( )bp ,ip ,ap ( B ) ,b ,i +1 ( B ) ,b ,i +1 ( B ) ,b ,i
IBX b ,i , ,bp ,ip , ap = 2 2
2 ( r1 )bp , ap ,ip , ,b ,i ( r2 )bp ,ap ,ip , ,b ,i ( r1 )bp , ap ,ip , ,b ,i + ( r2 )bp , ap ,ip , ,b ,i ( L ) ,b ,i
r + r X X Z Z Z Z X X
( 1 )bp ,ap ,ip , ,b ,i ( 2 )bp ,ap ,ip , ,b ,i ( )bp ,ip ,ap ( B ) ,b ,i +1 ( B ) ,b ,i +1 ( B ) ,b ,i ( )bp ,ip , ap ( B ) ,b ,i +1 ( B ) ,b ,i +1 ( B ) ,b ,i
IBYb ,i , ,bp ,ip ,ap = 2 2
2 ( r1 )bp , ap ,ip , ,b ,i ( r2 )bp ,ap ,ip , ,b ,i ( r1 )bp , ap ,ip , ,b ,i + ( r2 )bp , ap ,ip , ,b ,i ( L ) ,b ,i
( r1 ) + r Y Y X X X X Y Y
bp ,ap ,ip , ,b ,i ( 2 )bp ,ap ,ip , ,b ,i ( )bp ,ip ,ap ( B ) ,b ,i +1 ( B ) ,b ,i +1 ( B ) ,b ,i ( )bp ,ip , ap ( B ) ,b ,i +1 ( B ) ,b ,i +1 ( B ) ,b ,i
IBZ b ,i , ,bp ,ip ,ap = 2 2
2 ( r1 )bp ,ap ,ip , ,b ,i ( r2 )bp , ap ,ip , ,b ,i ( r1 )bp ,ap ,ip , ,b ,i + ( r2 )bp ,ap ,ip , ,b ,i ( L ) ,b ,i
2 2 2
( r1 )bp ,ip ,ap , ,b,i = ( X )bp ,ip , ap ( X B ) ,b ,i + (Y )bp ,ip , ap (YB ) ,b ,i + ( Z )bp ,ip , ap ( Z B ) ,b ,i
2 2 2
( r2 )bp ,ip ,ap , ,b,i = ( X )bp ,ip , ap ( X B ) ,b ,i +1 + (Y )bp ,ip ,ap (YB ) ,b ,i +1 + ( Z )bp ,ip , ap ( Z B ) ,b ,i +1
2 2 2
( L ) ,b,i = ( X B ) ,b ,i +1 ( X B ) ,b ,i + (YB ) ,b ,i +1 (YB ) ,b ,i + ( Z B ) ,b ,i +1 ( Z B ) ,b ,i
Table5.1:Discretization equationsforgeometricinfluencecoefficientsforboundcirculation
Chapter5DevelopmentofaFreeWakeVortexModel
( r1 ) + r Z Z Y Y Y Y Z Z
bp ,ap ,ip ,b ,i ,a ( 2 )bp ,ap ,ip ,b ,i ,a ( )bp ,ip ,ap ( )b ,i ,a 1 ( )b,i ,a 1 ( )b ,i ,a ( )bp ,ip ,ap ( )b ,i ,a 1 ( )b ,i ,a 1 ( )b ,i ,a
ITX bp ,ip ,ap ,bia = 2 2
2 ( r1 )bp ,ap ,ip ,b ,i ,a ( r2 )bp ,ap ,ip ,b ,i ,a ( r1 )bp ,ap ,ip ,b ,i ,a + ( r2 )bp ,ap ,ip ,b,i ,a ( L )b ,i ,a
r + r X X Z Z Z Z X X
( 1 )bp ,ap ,ip ,b,i ,a ( 2 )bp ,ap ,ip ,b ,i ,a ( )bp ,ip ,ia ( )b,i ,a 1 ( )b ,i ,a 1 ( )b,i ,a ( )bp ,ip ,ap ( )b ,i ,a 1 ( )b ,i ,a 1 ( )b ,i ,a
ITYbp ,ip ,ap ,b ,i ,a = 2 2
2 ( r1 )bp ,ap ,ip ,b ,i ,a ( r2 )bp ,ap ,ip ,b,i ,a ( r1 )bp ,ap ,ip ,b ,i ,a + ( r2 )bp ,ap ,ip ,b ,i ,a ( L )b,i ,a
r + r Y Y X X X X Y Y
( 1 )bp ,ap ,ip ,b ,i ,a ( 2 )bp ,ap ,ip ,b ,i ,a ( )bp ,ip ,ap ( )b ,i ,a 1 ( )b ,i ,a 1 ( )b ,i ,a ( )bp ,ip ,ap ( )b ,i ,a 1 ( )b ,i ,a 1 ( )b ,i ,a
ITZ bp ,ip ,ap ,b,i ,a = 2 2
2 ( r1 )bp ,ap ,ip ,b ,i ,a ( r2 )bp ,ap ,ip ,b ,i ,a ( r1 )bp ,ap ,ip ,b ,i ,a + ( r2 )bp ,ap ,ip ,b,i ,a ( L )b ,i ,a
2 2 2
( r1 )bp ,ip,ap ,b,i ,a = ( X )bp ,ip ,ap ( X )b ,i ,a + (Y )bp ,ip ,ap (Y )b,i ,a + ( Z )bp ,ip ,ap ( Z )b ,i ,a
2 2 2
( r2 )bp ,ip ,ap ,b,i ,a = ( X )bp ,ip ,ap ( X )b ,i ,a 1 + (Y )bp ,ip ,ap (Y )b ,i ,a 1 + ( Z )bp ,ip ,ap ( Z )b ,i ,a 1
2 2 2
( L )b ,i , a = ( X )b ,i ,a 1 ( X )b ,i ,a + (Y )b,i ,a 1 (Y )b ,i ,a + ( Z )b ,i ,a 1 ( Z )b ,i ,a
Table5.2:Discretization equationsforgeometricinfluencecoefficientsfortrailingcirculation
219
Chapter5DevelopmentofaFreeWakeVortexModel
220
( r1 ) + r Z Z Y Y Y Y Z Z
bp ,ap ,ip ,b ,i ,a ( 2 )bp ,ap ,ip ,b,i ,a ( )bp ,ap ,ip ( )b,a ,i +1 ( )b ,a ,i +1 ( )b ,a ,i ( )bp ,ap ,ip ( )b,a ,i +1 ( )b ,a ,i +1 ( )b,a ,i
ISX bp ,ip ,ap ,b ,i ,a = 2 2
2 ( r1 )bp ,ap ,ip ,b ,i ,a ( r2 )bp ,ap ,ip ,b ,i ,a ( r1 )bp ,ap ,ip ,b ,i ,a + ( r2 )bp ,ap ,ip ,b ,i ,a ( L )b ,i ,a
r + r X X Z Z Z Z X X
( 1 )bp ,ap ,ip ,b ,i ,a ( 2 )bp ,ap ,ip ,b,i ,a ( )bp ,ap ,ip ( )b ,a ,i +1 ( )b ,a ,i +1 ( )b,a ,i ( )bp ,ap ,ip ( )b,a ,i +1 ( )b ,a ,i +1 ( )b,a ,i
ISYbp ,ip ,ap ,b ,i ,a = 2 2
2 ( r1 )bp ,ap ,ip ,b ,i ,a ( r2 )bp ,ap ,ip ,b ,i ,a ( r1 )bp ,ap ,ip ,b,i ,a + ( r2 )bp ,ap ,ip ,b ,i ,a ( L )b ,i ,a
( r1 ) + r Y Y X X X X Y Y
bp ,ap ,ip ,b ,i ,a ( 2 )bp ,ap ,ip ,b ,i ,a ( )bp ,ap ,ip ( )b,a ,i +1 ( )b ,a ,i +1 ( )b ,a ,i ( )bp ,ap ,ip ( )b,a ,i +1 ( )b ,a ,i +1 ( )b ,a ,i
ISZ bp ,ip ,ap ,b ,i ,a = 2 2
2 ( r1 )bp ,ap ,ip ,b,i ,a ( r2 )bp ,ap ,ip ,b,i ,a ( r1 )bp ,ap ,ip ,b,i ,a + ( r2 )bp ,ap ,ip ,b,i ,a ( L )b ,i ,a
2 2 2
( r1 )bp ,ip ,ap ,b,i ,a = ( X )bp ,ip ,ap ( X )b ,i ,a + (Y )bp ,ip ,ap (Y )b ,i ,a + ( Z )bp ,ip ,ap ( Z )b ,i ,a
2 2 2
( r2 )bp ,ip ,ap ,b,i ,a = ( X )bp ,ip ,ap ( X )b,i +1,a + (Y )bp ,ip ,ap (Y )b,i +1,a + ( Z )bp ,ip ,ap ( Z )b ,i +1,a
2 2 2
( L )b ,i , a = ( X )b ,i +1,a ( X )b ,i ,a + (Y )b ,i +1,a (Y )b ,i ,a + ( Z )b ,i +1,a ( Z )b,i ,a
Table5.3:Discretization equationsforgeometricinfluencecoefficientsforshedcirculation
Chapter5DevelopmentofaFreeWakeVortexModel
Chapter5DevelopmentofaFreeWakeVortexModel
5.2.3FarWakeModel
Inordertoaccountfortheinfluenceofthefarfieldonthenearwake,afarwakemodelis
incorporated. This vortex model is very similar to the prescribedwake vortex model
HAWT_PVC(seesection4.3.3),withthedifferencethatonlyasingletipvortexhelicalmodel
per blade is modelled. Also, each helix has a constant diameter and is attached to the
outboardendofthecorrespondingnearwake,i.e.tonearwakenodewithcoordinate(0,n
1), see Fig. 5.2. Thus at this node, the near wake vortex sheet is modelled to rollup
immediatelyintoaconcentratedtipvortex.Eachhelixissegmentedintostraightlinevortex
filaments and the discretized BiotSavart equations (Appendix C, Eqts. C.24 C.32) are
applied with a simple numerical cutoff method. As the rotor model is set to rotate, the
vortex sheet of the near wake starts growing with time and consequently the farwake
helicaltipvorticesareconvecteddownstream.
In this far wake model, the helical pitch (pfw) and the number of helical revolutions
(nfwRev)needtobeprescribedbythecodeuser.Whenmodellingyawedconditions,thefar
waketiphelicesareskewed,similartoHAWT_PVCandthewakeskewangle(s)alsoneeds
tobeprescribed.Fig.5.5depictsatypicalplotofthemodelledhelicaltipvortexfarwake
duetoasingleblade.
Fig.5.5Typicalhelicaltipvortexgeometryforfarwakemodelforoneblade.s=350andnfwRev=5.
221
Chapter5DevelopmentofaFreeWakeVortexModel
5.2.4NumericalSolution
The procedure for obtaining the unsteady free near wake solution for this model is now
described. A timemarching algorithm was implemented with the solution started by an
impulsivestartoftherotor.Fig.5.6presentsaflowchartforthissolution.
Aboundaryconditionusedspecifiesthateachtrailingandshedvortexfilamentshouldbe
attachedtothebladetrailingedgeatthepointoforigin.Initially,thefollowingdatahasto
beinputtedtothemodel:
i. Rotor geometry parameters: blade tip and root radii (Rt and Rr), rotor number of
blades(B)andradialdistributionsofbladechordandtwist(cand).
ii. Operating condition parameters: rotor parameters (angular speed (), blade pitch
angle(),andyawangle()andwindspeed(U)).
iii. Bound circulation distributions at different blade azimuth angles over one whole
revolutioncorrespondingtotheoperatingconditionsbeingconsidered.
iv. Viscousmodellingparametersfornearwake:Scand v.Thesearemaintainedconstant
throughoutthesolution.
v. Farwakemodelparameters:tipvortexpitch(pfw),numberofrevolutionsforhelical
tipvortices(nfwRev),wakeskewangle(s)andnumericalcutoffparameter.
Prior to commencing the freewake solution, one whole rotor revolution is divided into a
fixednumberofazimuthpositions(equalto tot)asinFig.3.7,Chapter3.Thetimestepis
equalto andiscalculatedbyEqt.4.31,Chapter4.Thenumberofrotorrotations(nRev)to
generate the free (near) wake is inputted. The total number of rotor time steps for the
solutionisthengivenbyMtotwhereMtot=nRevXtot.Initially,thefirstbladeissetatthezero
azimuthposition(bladeverticalpointingupwards)andthereisnonearwake,butonlythe
prescribed far wake consisting of a vortex helix emerging from each blade tip. In the far
wake, the number of filaments per helical revolution is set equal to tot and this remains
constant throughout the whole computation. The rotor then starts rotating at constant
speed().Ateachtimestep,trailingandshedvortexfilamentsareshedfromtheblades
trailing edges with a circulation calculated from the spanwise and time variations of the
bound circulation distributions at the blades. This will eventually form a near wake that
willextenddownstream.
Fig. 5.7 illustrates the growth (formation) of the near free wake after each time step m.
Initially, (m = 0, = 0), shed vortices are placed at the trailing edge of each blade with a
circulationstrengthequalandoppositetotheboundcirculationinordertosatisfyKelvins
condition.Ateachtimestepm,thefollowingstepsaremade:
i. newlyformedtrailingandshedvorticesareplacedatthetrailingedgeoftheblades.
Their circulation strength is computed from the local radial and time variations in
boundcirculationusingEqts.5.3.
222
Chapter5DevelopmentofaFreeWakeVortexModel
Start DefineRotorGeometry
InputOperatingConditions
InputBoundCirculation
DistributionatEachTimeStep
DefineInitialNodesat
TrailingEdgeofEachBlade
=0
m=0
CreateNewTailing/ShedVortexFilamentsatthe
TrailingEdgeofEachBlade(Eqts.5.3)
CalculatetheNewViscousCoreSizeatEachVortexFilament(Eqt.5.8)
CalculatetheNewStrainatEachVortexFilament(Eqt.E.3,AppendixE)
CorrecttheViscousCoreSizeatEach
VortexFilamentforStrainingEffects (Eqt.5.9)
UpdatethePosition
ofEachBlade
Calculatethe3DInducedVelocitiesattheWake
NodesusingTheBiotSavart Law(Eqt.5.10)
FindtheNewPosition
ofEachWakeNode
Alltimesteps
= +1 End
analysed?
m=m+1
Figure5.6Flowchartdescribingthenumericalsolutionforgeneratingnearfreewake.
223
Chapter5DevelopmentofaFreeWakeVortexModel
ii. a core size is determined for all wake shed and trailing filaments accounting for
viscousgrowthandfilamentstrainingusingEqts.5.8and5.9.
iii. the 3D induced velocities at each wake node due to bound, trailing and shed
circulationarecalculatedusingEqts.5.10.
r
iv. theabsolutevelocityateachwakenode, W ,isdeterminedbyvectoriallyaddingthe
flowfreestreamvelocityandthelocalinducedvelocitiesusingtheformula:
r r r
W = U + u (5.11)
Eqt. 5.11 is written in the XYZ reference frame and considering a uniform free
windspeeditisexpressedas
WX 0 u X
WY = U + uY (5.12)
WZ 0 uZ
v. thepositionofeachwakenodein3DspaceisdeterminedusingtheEulerexplicit
timeintegratingscheme:
r 1 r r
= r + (W +1 + W ) (5.13)
r
r +1
2
vi. therotorazimuthpositionandtimestepareupdated(andm).
TheaboveprocessisrepeateduntilmisequaltoMtot.Duringthelastrotorrevolution,(i.e.
from time step m=(Mtottot) to m=Mtot, the BiotSavart law is applied to both the generated
near free wake and the prescribed far wake to compute the 3D induced velocities at a
selectedplaneparalleltotherotorplane.These3Dinducedvelocitiesareobtainedinthex
yzreferenceframefordifferentrotorazimuthpositions().Thedescretizationofpointsin
this selected plane is determined in a similar way to HAWT_PVC, as shown in Fig. 4.40
(page107).ItcanbenotedthatinthetimeintegratingschemeusingEqt.5.13,theabsolute
velocityofeachnodeistakentobeequaltotheaverageofthenewlycalculatedvalueand
thatoftheprevioustimestep.Thisaveragingprocedureimprovesthenumericalaccuracy
andstabilityofthetimeintegratingscheme,asnotedin[5].
224
Chapter5DevelopmentofaFreeWakeVortexModel
BladeTrailing
Edge
a=m
a=m
a=m
a=0
a=1
a=0
i=n1 i=n1 i=n1
Rotor
Axis Timestep0 Timestep1 Timestep2
(m=0) (m=1) (m=2)
a=m
a=m
a=0
a=3
a=2
a=1
a=2
a=1
a=0
i=n1 i=n1
i=n2 i=n2
. .
. .
i=1 i=1
i=0 i=0
Timestep3 Timestep4
(m=3) (m=4)
Figure5.7Timemarchingsequenceforvortexlatticeformationofthenearwakeduetoeachblade.
Thefirstfivetimestepsareshown.
225
Chapter5DevelopmentofaFreeWakeVortexModel
5.2.5ProgramStructure
HAWT_FWC was written in MathCad version 11. Fig. 5.8 describes the structure of the
program. Like the other developed codes, there are three main modules: the Data Input
Module,inwhichtheparametersdescribingtherotorgeometryandoperatingconditionare
inputted,togetherwiththeprescribedsteady/unsteadyboundcirculationdistributions.The
farwakemodelparameters(pfw,nfwRevands)arealsoprescribedhere.Thedistancefrom
the rotorplane of the plane at which the 3D induced velocities are to be computed is also
inputted.TheDataProcessingModuleimplementsthevortexmodelandthetimemarching
solutionusingtheBiotSavartequationsasdescribedinsections5.2.1,5.2.2,5.2.3and5.2.4
togeneratethefreewakeandcalculatethe3Dinducedvelocitiesataplaneparalleltothe
rotorplane. Given that the aerofoil data are known, this module also calculates the
steady/unsteady aerodynamic blade loading distributions using the BET equations (Eqts.
3.16 and 3.20, Chapter 3) as well as the 3D global aerodynamic loads induced at the yaw
bearing (using procedure described in Appendix B). The Data Output Module outputs the
freewakegeometry,togetherwiththetrailingandshedcirculationinthewakeandthe3D
induced velocities in the required plane for different rotor azimuth positions. The blade
aerodynamicloaddistributionsandthegloballoadsattheyawbearingarealsooutputted
asafunctionofblade/rotorazimuthangle.
226
Chapter5DevelopmentofaFreeWakeVortexModel
DataInput:
(1)InputRotorGeometryDetails:
NumberofBlades(B)
BladeTipandRootRadii(Rt & Rr)
Bladechordandtwistdistributions(c& )
(2)InputRotorOperatingConditions:
RotorAngularSpeed()
RotorYawAngle()
WindSpeed(U)
(3)InputBoundCirculationatBlades:
BoundCirculationatBlades(B)
(4)Viscousmodelling parameters:
Nearwakeviscousparameters(v and Sc))
(5)FarWakeModel:
WakeGeometryDetails(pfw,nfwRev ands)
Numericalcutoffparametervalue
DataProcessing:
(1)Freewakenumericalsolution
(2)Computationsofinducedvelocities
atrequiredplaneparalleltorotorplane
(3)Aerodynamicloadcalculations
DataOutput:
(1)Inducedvelocitiesatplaneparallelto
rotorplane
(2)Aerodynamicloaddistributionsatblades
(3)Globalloads
Figure5.8StructureofcomputercodeHAWT_FWC.
227
Chapter5DevelopmentofaFreeWakeVortexModel
5.3VerificationandValidationofFreewakevortexmodel
As with all numerical models, assessing the accuracy and reliability of the solution in
reproducing results observed in the real world is an indispensable requirement. This
sectiondescribestheverificationandvalidationworkcarriedoutonthenewlydeveloped
freewake liftingline code HAWT_FWC for modelling wind turbines in both axial and
yawedconditions.Thisworkwasindispensableinordertobeabletoassessthereliability
ofthenewmethodbeingproposedinChapter6forderivingtheangleofattackandinflow
distributionsintheturbinewakefrombladepressuremeasurements.
Although freewake models have reached a reasonably good level of maturity, more
rigorous validation efforts are still required. It is important to remark here the distinction
betweenverificationandvalidation.AsdefinedbyOberkampfetal.[59],Verificationisthe
processofquantifyingthenumericalerrorinsolvingaconceptualmodel.Indoingso,the
accuracyofthecomputationalsolutionismeasuredrelativetotwotypesofhighlyaccurate
solutions:analyticalsolutionsandhighlyaccuratesolutions.Validation,ontheotherhand,is
theassessmentofaccuracyofacomputationalsolutionwhencomparedwiththerealworld,
i.e.theexperimentaldata.Inverificationthecomparisonofthecomputationalresultswith
the experimental data is not an issue. In validation this comparison is an issue. Briefly
speaking,verificationisprimarilyamathematicsissuewhilevalidationisaphysicsissue.
InthisstudyonHAWT_FWC,validationisbasedondirectcomparisonofnearwakeinflow
measurements.Thismakesthevalidationmorerigorousthenwhencomparingbladeloads.
Thereasonforthisregardsthebasicfactthatfreewakeliftinglinevortexmethodsstillrely
on the bladeelement theory for computing the aerodynamic loads on the blades. The
accuracy in load calculations will depend on the accuracy of the induced velocities
computed by the freewake code as well as the aerofoil data used. At low induction
conditions however, a large error in the calculated induced velocities can still yield
considerablyaccurateloads.Butthiswouldnotmeanthatthefreewakemodelisaccurate.
Alsoerrorsintheaerofoildatamaymaskerrorsintheinducedvelocitiescomputedbythe
freewakecode.
5.3.1Verification&ValidationMethodology
ThissectiondescribestheproceduresusedtoverifyandvalidatevortexcodeHAWT_FWC.
A.Verification
228
Chapter5DevelopmentofaFreeWakeVortexModel
procedureshere.Consequently,solutionconvergenceisonlyduetoinsufficientbladeand
vortex sheet (i.e. wake) discretization. Blade and wake discretization convergence implies
that,bysystematicallyusingafinerdiscretization,thesolutionshouldeventuallyconverge
asymptoticallytoanexactsolutionwhichisthecalculuslimit.
Thebladeliftinglinediscretizationisdeterminedbynalone.Increasingnwillmakeblade
discretizationfinerasitwillrepresentmoreaccuratelythechordandtwistvariationsalong
thebladespan.Alargernwillalsoyieldamoreaccuraterepresentationoftheprescribed
bound circulations (refer to Fig. 5.1). The discretizaton of the near wake is determined by
the number of vortex filaments to make up the mesh to represent the vortex sheet/s
extendingdownstreamfromtherotorplane.Thisdiscretizationisdependentonmodelling
parametersn, , andnRev. isdirectlyrelatedto ( = /).Forafixednumberof
rotor revolutions (nRev), increasing n and using a smaller will increase the number of
filaments,thusmakingthewakediscretizationfiner.Anotherparameteristhedistanceto
which the near wake extends downstream of the rotorplane. For fixed values of n and
, increasingMtotwillmakethenearwakeextendmoredownstream.Thediscretizationof
thefarwakeisdeterminedbythenumberofstraightlinesegmentstakentorepresentthe
prescribednumberofhelicesaswellas.
To be physically meaningful, the required outputs from the freevortex solution should
ideallybeinsensitivetodiscretization.Inverification,aconvergencestudyisconductedby
numericalexperimentsinwhichasystematicrefinementisappliedwithonediscretization
parameteratatimewhileholdingtheothersconstant.Inthisway,onemaydeterminethe
mostappropriatevaluesthatshouldbeassignedtothediscretizationparametersinorderto
achieve acceptable levels of numerical accuracy. Another fact that should be taken into
account is computational cost. A major drawback of freewake models is the very rapid
increaseincomputationalcostasthenumberofvortexfilamentsisincreased.Compromise
should therefore be reached between numerical accuracy and computational cost. It is
imperative that the verification procedure demonstrates the computational expenses
requiredtoachievetherequirednumericalaccuracy.
The higher the number of dicretization parameters, the more complex the convergence
studywillbe.Yetthereisanotherproblemthatcomplicatesthingsfurther:apartfromthe
discretisationparameters,thereareothermodellingparametersthatareincludedtomake
the conceptual (mathematical) model more realistic. The problem is that we do not know
exactly which values for such parameters to use for a simulation. It is best to leave the
choicearbitraryandtothesubjectivedecisionofcodeuser.Itisvalidationnotverification
thatwillhelpusidentifythemostrealisticvaluestouseinasimulation.Therearetwosets
of such arbitrary parameters in the subject freewake model: (1)Viscous core modelling
parameters(vandSc)and(2)Farwakemodellingparameters(cutoffparameter,prescribed
pitch(pfw)andnumberofrotationsforwhichthehelicesextenddownstream(nfwRev)).In
this situation, the verification procedure should ideally check whether the discretizaton
convergencecriterionismetoverawiderangeofvaluesofthearbitraryparameters.
229
Chapter5DevelopmentofaFreeWakeVortexModel
Inthisverificationprocedure,thecomputedaxialinducedvelocityatthebladesisusedasa
basis for comparison. Apart from providing the necessary evidence of dicretization
convergence and correct implementation of the conceptual model, the verification process
will provide insight into the numerical behaviour of the code and how the different
parameterswillinfluencetheresults.
Toverifythefreewakemodel,twodifferentcasestudieswereconsidered.Inthefirstcase
study,thefreewakecodemodelledasingleellipticalbladerotatingataverylargeradius
compared to its span. This approximated an elliptical wing moving linearly at constant
speed.AnellipticalboundcirculationwasprescribedtothebladeasshowninFig.5.9and
thecalculateddownwashwascomparedwiththatfromtheanalyticalsolutionwhichstates
thatthedownwashisconstantalongthewingspanandequalto([3],[42]):
Bmax
u a ,c = (5.14)
2bs
Thebladegeometricandoperatingparametersweresetasshownintable5.4.Thefarwake
modelwasnotincludedinthiscasestudy.
2
2z
B ( z ) = B , MAX 1
bs
BMAX B(z)
2
2z
c ( z ) = c0 1
bs
z
c0 c(z)
Figure5.9Ellipticalwingwithaprescribedboundcirculationdistribution.
230
Chapter5DevelopmentofaFreeWakeVortexModel
Table5.4Bladegeometryandoperatingconditionparametersforellipticalwing.
BladeGeometryParameters
B 1
R 1000m
Bladespan 10m
3deg
c0 2m
OperatingConditions
U 0m/s
0.1rev/min
1.225kg/m3
Forthesecondcasestudy,theverificationwascarriedoutconcurrentlywiththevalidation
process (see Part. B) when modelling the TUDelft wind tunnel turbine whose geometric
detailsaregiventable4.1,page46.HAWT_FWCmodelledtheTUDelftrotorforthesame
conditionsasthoseusedforhotfilminflowmeasurements(withU=5.5m/s,=8and tip=20).
The verification was carried out for =00 and 450. The bound circulation distributions
obtained from HAWT_LFIM using the inflow measurements and the unsteady aerofoil
theory(seeFigs.4.55(method1)and4.79(uncertaintylimitsnotincluded),Chapter4)were
prescribed to HAWT_FWC to generate the near freewake and compute the induced
velocities at the rotorplane or at 3.5cm downstream. The axial induced velocities were
computed for different values for the discretization parameters n, and nRev and the
numericalaccuracieswerecomputedandanalyzed.Tolimitthecomplexityresultingfrom
alargenumberofvariablestobeanalyzedseparatelywhilemaintainingtheotherconstant,
thefarwakemodellingparameterswhetherkeptconstantthroughtheverificationstudy.
Thedisadvantageofdoingverificationandvalidationconcurrentlyisthatanyerrorsdueto
insufficient numerical convergence may mask other errors when comparing with
experimental data. To avoid this problem, it was assured in this study that appropriate
numericalconvergencewasachievedbeforecomparingwithexperimentaldata.
B.Validation
As already outlined in Chapter 2 (Fig. 2.4), the experimental data from the TUDelft wind
tunnelturbinemodelwasusedforthevalidationworkonthefreewakevortexmodel.The
validation was performed for =00, 300 and 450. The bound circulation distributions
obtained from HAWT_LFIM using the inflow measurements and the unsteady aerofoil
theory(seeFigs.4.55,4.78and4.79,Chapter4)wereprescribedtoHAWT_FWCtogenerate
231
Chapter5DevelopmentofaFreeWakeVortexModel
thefreewakeandcomputetheinducedvelocitiesattherotorplaneorat3.5cmdownstream
atdifferentvaluesforviscousparameters(v,Sc).Twoindependentprocedureswerethen
adoptedforthevalidation:inthefirstprocedure,theinducedvelocitiesfromthefreewake
model were compared with those derived from the hotfilm measurements (see Figs. 4.59,
4.85and4.86);inthesecondprocedure,thetipvorticallocationsfromthefreewakemodel
were compared with those from the smoke visualization (see Figs. 4.24, 4.27 and 4.28,
Chapter 4). A limitation of this validation process is that it only considers attached flow
conditions. Validation of HAWT_FWC for stalled flow conditions over the blades is not
performed.
Fig.5.10summarizestheprocedurefortheverificationandvalidationworkofHAWT_FWC
whenmodellingtheTUDelftrotor.
232
Chapter5DevelopmentofaFreeWakeVortexModel
ModelTUDelft turbine
forU=5.5m/s,=8,tip=20
PrescribeB distributionsto
HAWT_FWCfor=00,300 and450
Verification
Computeaxialinducedvelocitiesatrotorplane orat3.5cmdownstream
andfindnumericalaccuracyatdifferentvaluesofdiscretization
parametersn, andnRev
Verificationcarriedoutat=00 and450
Validation
Computeaxialinducedvelocitiesatrotorplane orat3.5cmdownstream
withvaluesfordiscretization parametersn,,andnRev selectedtogive
sufficientnumericalaccuracyatreasonablecomputationalcost
Validationcarriedoutat=00,300 and450
(1)Validationusinghotfilmnearwakemeasurements
Computeaxialinducedvelocitiesatrotorplane
orat3.5cmdownstreamatdifferentvaluesof
viscousmodelling parameters(v,Sc)andcomparethem
withthoseobtainedfromhotfilmmeasurements
(2)Validationusingsmokevisualizationmeasurementsof
tipvortexpaths
Computefreewakegeometryatdifferentvaluesof
viscousmodelling parameters(v,Sc)andcomparethetip
vortical locationswiththosefromsmokevisualization
Figure5.10ProcedureforVerificationandValidationofHAWT_FWCusingtheTUDelftrotor
model.
233
Chapter5DevelopmentofaFreeWakeVortexModel
5.3.2ResultsandDiscussion
This section describes the results obtained from this verification and validation study on
HAWT_FWC.Theresultsareorganizedintwoparts:PartAdescribestheresultsobtained
fromtheverificationstudywhichmodelledtheellipticalwing;PartBdescribestheresults
from the verification and validation which modelled the TUDelft wind tunnel turbine for
theoperatingcondition=8,=20at=00,300and450.
A.VerificationbyModellingEllipticalWing
Aconstantellipticalcirculationdistribution(timeindependent)wasprescribedattheblade
rotatingattheverylargeradiuswith B,MaxinEqt.5.14setto10m2/s.Thebladewasrotated
until a constant spanwise distribution for the downwash was obtained at the lifting line
(ua,c).Itwasfoundthatbyrotatingthebladebyonly30wasenoughtoyieldadownwashat
theliftinglinethatwasindependentofvortexsheetlength.Thiswasequivalenttoratioy/c0
equaltoabout26whereyisthedistancetravelledbythewing.Sincetheboundcirculation
isconstantwithtime,therearenoshedvortexfilamentsinthewakeexceptthoseincluded
at the first time step to account for Kelvins condition. For B,Max equal to 10m2/s, the
analyticalvalueforthedownwashattheliftinglineasgivenbyEqt.5.14isequalto0.5m/s.
Fig. 5.11 shows a typical downwash distribution predicted by the freewake code at
differentvaluesofn.Itcanbeobservedthatthepredictionsareveryclosetotheanalytical
result. The freewake numerical solution predicts a downwash that is fairly constant for
mostofthespanexceptatthewingstipswherethedownwashincreasesrapidlytopositive
values.
3.5
1.5
0.5
0
-6 -4 -2 0 2 4 6
-0.5
-1
z (m)
Figure5.11Spanwisedistributionofdownwashforellipticalwing.Freewakecoderesults
computedforMtot=30,=0.10,=0.167s,v=10,Sc=1.
234
Chapter5DevelopmentofaFreeWakeVortexModel
Afinerdiscretizationfortheliftingline(i.e.usingalargern)reducesthedeviationfromthe
analyticalresultatallradiallocations,especiallythewingtips.Thediscrepanciesbetween
the analytical and the freewake code results are mainly due to the fact in the analytical
solution, the wake is assumed to be rigidly flat. In the freewake solution, the wake is
allowedtodevelopfreelyin3Dspaceundertheactionoftheselfinducedvelocityandthat
fromthebladeliftingline.Theresultingfreewakeisnotflat,especiallytowardsthewing
tipsweretherearehigherlevelsoftrailingvorticity.
Fig.5.12showsthebuildupofdownwashua,c attheliftinglineasthewingissuddenlyset
intomotionatconstantspeedandwithaconstantellipticalboundcirculation.Asthewing
moves,itleavesawakebehindthatcausesanevolvingdownwashatthewing.Atstartup,
thedownwashiszerobutthisisfollowedbyaninstantaneousincreaseduetotheshedding
ofthestartingshedvortex.Withtime,thestartingvortexmovesawayfromthewingand
thedownwashapproachestothesteadyvalues.
Fig.5.13displaysatypicalresultingwakeplotfromthefreewakesolution.Thepresenceof
rollupofthevortexsheetatthewingtipsisevidentandthetipvortexformationateach
wingtipshowsupunmistakably.
0.25
0.05
ua,c (m/s)
0 5 10 15 20 25
-0.15
-0.35
-0.55
-0.75
y/c 0
Figure5.12Buildupofdownwashatdifferentspanwiselocationsaspredictedbyfreewakecode.
n=25,Mtot=30,=0.10,=0.083s,v=500,Sc=10.
235
Chapter5DevelopmentofaFreeWakeVortexModel
Elliptical wing
trailing edge Directio
n of win
g motio
n
B.VerificationandValidationbyModellingTUDelftWindTunnelTurbineModel
B.1.AxialConditions
The results obtained from the verification and validation procedures applied for the
TUDelftturbineat =00arenowpresented.TheboundcirculationdistributionofFig.4.55
(usingmethod1)wasprescribedtoHAWT_FWCandthecomputationswerecarriedoutfor
different values of nRev, , n, v and Sc. The computed axial induced velocities were
compared with those obtained experimentally (Figs. 4.59(a) and (b)). The wake vortical
locations predicted by the freewake code were compared with the tip vortex locations
measuredduringthesmokevisualizationexperiments.
Tolimitthenumberofvariablesintheverificationandvalidationstudy,theprescribedfar
wakemodelparametersinHAWT_FWC(pitch(pfw),numberofhelicalrevolutions(nfwRev)
andcutoffdistance)werekeptconstantthroughoutthestudy.Todetermineanappropriate
pitch (pfw) for the far wake, preliminary freewake computations were carried out at
differentfarwakepitchvaluesandthepitchvaluewasextractedfromthecomputednear
freewakegeometry.Thenumberofhelicalrevolutions(nfwRev)waskeptconstantat5.A
constantcutoffdistanceof0.5mmwasused.
Verification
EffectofnRevandfarwakemodel
In order to obtain realistic results for the induced velocities at the blade lifting lines, one
shouldmakesurethatthefreewake(i.e.thenearwakeinthiscontext)extendsfarenough
downstream from the rotor. When treating rotors in steady axial flows, two conditions
236
Chapter5DevelopmentofaFreeWakeVortexModel
mustbesatisfiedbeforeapplyingtheBiotSavartlawtocomputetheinflowdistributionat
therotorplane:Condition1isthatthecomputedinducedvelocitydistributionattheblades
should ideally be independent of the number of revolutions (nRev) used to generate the
freewake;Condition2isthatthisvelocitydistributionshouldideallybeindependentofthe
farwake, i.e. the computed induced velocities at the rotorplane with and without thefar
wake vortex model should be the same. This is because the farwake model is only a
prescribedwakevortexmodelandisthereforelessreliable.Condition1ismetbyselecting
nRev large enough to reach a suitable degree of wake periodicity. Wake periodicity is
achievedif,whencomputingtheinflowattherotorplaneduringthelastrotorrevolution,
the induced velocity at the blade lifting line of the first blade at a rotor azimuth angle of
3600isequaltothatatarotorazimuthangleof00.Asuitabledegreeofwakeperiodicityis
achieved if the relative error is close to zero. The percentage relative error for wake
periodicityisdefinedhereas
ua ,c = 00
ua ,c = 3600
wp = 100 (5.15)
ua ,c = 360 0
Condition2isalsometbyusingasufficientlylargevaluefornRev.IncreasingnRevwillpush
thefarwakehelicalvortexmodelawayfromtherotorandthusitscontributiontothetotal
inductionatthebladesislessinfluential.
IninvestigatingquantitativelytheeffectofnRevandthefarwake,freewakecomputations
werecomputedatdifferentvaluesofnRev(equalto1,2,3,4and5)whilekeepingtheother
parameters, n, , v and Sc, constant (at 21, 100, 10 and 5 respectively). The reasons for
selectingthelattervaluesforsuchparameterswillbediscussedlater.Theinducedvelocities
atthebladeswerecomputedwithandwithoutthefarwakemodelincluded.Notehowever
that the far wake model was still included to generate the near wake. Fig. 5.14 shows the
computedaxialinducedvelocitydistributionsatthebladeliftinglinesforthedifferentnRev
values.Fig.5.15showsthecorrespondingvariationoftherelativepercentageerrorforwake
periodicityagainstnRevcomputedusingEqt.5.15.Fromthisfigure,itmaybeobservedthat
wake periodicity is achieved rapidly after the first three rotor revolutions and the
percentage discrepancy converges steadily towards zero over the following revolutions.
With nRev equal to 3, the relative percentage error is below 5 percent at all radial blade
locations.Fig.5.16illustratestheincreaseincomputationaltimewhenincreasingnRevfrom
1 up to 5. The Relative Computational Time Factor (RCTF) is plotted here instead of the
actual computational time in hours, since the latter varies depending on the computer
processorspeed.Inthiscase,theRCTFisdefinedasthetimerequiredtocomputethefree
wake solution with a given value of nRev divided by the time required to do the same
computationwithnRevsettoone.Therapidincreaseincomputationaltimeisevidentfrom
Fig. 5.16. Selecting nRev equal to 4 instead of 3 will reach a higher degree of wake
periodicity but this also implies that more than double the computation time is required
(RCTFis36insteadof16).
237
Chapter5DevelopmentofaFreeWakeVortexModel
0.0
0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
-0.2
-0.4
-0.6
-0.8
ua,c (m/s)
-1.0
-1.2
-1.4
-1.6
-1.8
Experiment nRev=1
-2.0 nRev=2 nRev=3
nRev=4 nRev=5
-2.2
r/R
Figure5.14Axialinducedvelocitydistributionatbladeliftinglinecomputedbyfreewakecodefor
different values of nRev. Blade is at 3600 azimuth. (Far wake model included in velocity
calculations).Barsinexperimentalcurverepresent+/onestandarddeviationintheexperimental
data.
45
40
35
r/R = 0.4
30 r/R = 0.6
r/R = 0.8
25
wp(%)
r/R = 0.9
20
15
10
0
1 2 3 4 5
nRev
Figure5.15VariationofRelativePercentageErrorinWakePeriodicitywithnRev.
Althoughgoodagreementwasachievedwiththeexperimentaldatawhenincludingthefar
wake model in the induced velocity calculations for all nRev (as shown in Fig. 5.14), the
agreementwasnotasgoodwhenperformingthesamecalculationsexcludingthefarwake
model(refertoFig.5.17).Yet,inFig.5.17betteragreementwiththeexperimentswasstill
238
Chapter5DevelopmentofaFreeWakeVortexModel
70
60
50
RCTF
40
30
20
10
0
1 2 3 4 5
nRev
Figure5.16VariationofRelativeComputationalTimeFactorwithnRev
0.4
0.2
0.0
0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
-0.2
-0.4
-0.6
ua,c (m/s)
-0.8
-1.0
-1.2
-1.4
-1.6
-1.8 Experiment nRev=1
-2.0 nRev=2 nRev=3
nRev=4 nRev=5
-2.2
r/R
Figure5.17Axialinducedvelocitydistributioncomputedbyfreewakecodefordifferentvaluesof
nRev(Farwakenotincluded).Barsinexperimentalcurverepresent+/onestandarddeviationin
theexperimentaldata.
obtained as nRev was increased.In analyzing the discrepancy in calculating the induction
withandwithoutthefarwakemodelatdifferentnRevvalues,theresultsofFigs.5.14and
5.17wereusedtocalculatethepercentagerelativeerrorgivenby:
ua , c ua , c
fw =
with far wake without far wake
100 (5.16)
ua , c without far wake
239
Chapter5DevelopmentofaFreeWakeVortexModel
TheerrorsareplottedinFig.5.18.Asexpected,asthenumberofnearwakerevolutionsis
increased,therelativeerrordecreasesandconvergencessteadilytoapproachzero.Thisis
due to the reduced influence of the far wake on the rotorplane induction as the former is
convecteddownstreambythegrowingnearwake.Itshouldbenotedthatateachvalueof
nRev, the relative errors in Fig. 5.18 are significantly larger than those in Fig. 5.15. This
implies that, as nRev is increased, condition 1 (i.e. that wake periodicity is reached) is
satisfiedbeforecondition2(i.e.independenceoffarwakemodel).Thisprovesthatselecting
nRevbasedonsatisfyingcondition1aloneisinsufficient,unlesscondition2isalsosatisfied.
Inthisanalysis,thecriteriontomeetconditions1and2arebasedonnRevbecausethelatter
isdirectlyrelatedtocomputationalcost.Inthiswayonecouldobtaininsightregardingthe
computational cost required to reach different levels of wake periodicity and farwake
independence. But one should realize the fact that this also depends on the operating
condition of the rotor, in particular the tip speed ratio (). A safer criterion is to base the
selection on the distance the near wake should extend downstream from the rotor. The
higherthetipspeedratio,thesmalleristhepitchofthevortexsheetandthereforeahigher
nRevwouldberequiredtoextendthenearwakesuitablydownstream.Inthissense,free
wakemodelsrequesthighercomputationalcostswhenmodellinghightipspeedsthanfor
lowervalues.Inthisstudy,inwhichtherotorwasoperatingat=8,thecorrespondingnear
wake distance from the rotorplane for nRev equal to 1, 2, 3, 4 and 5 are 0.33d, 0.75d, 1d,
1.25d,and1.5drespectively.
90
80
70 r/R=0.4
r/R=0.6
60
r/R=0.8
50 r/R=0.9
fw(%)
40
30
20
10
0
1 2 3 4 5
nRev
Figure5.18VariationofRelativePercentageError(fw)fordiscrepanciesbetweencomputationsfor
ua,ccarriedoutwithandwithoutfarwakemodel.
240
Chapter5DevelopmentofaFreeWakeVortexModel
This verification study therefore suggests that to reach wake periodicity and far wake
independence with an error of less than 10% at all radial locations at the blades, the near
wakeshouldatleastextend1.25diametersbeforecomputingtheinducedvelocitiesatthe
bladeliftinglines.
Effectof
Toinvestigatetheinfluenceofthetimestepsizeonthecomputedinducedvelocitiesatthe
blades,thecalculationsusingHAWT_FWCwerecarriedoutusingtheboundcirculationof
Fig. 4.55 (using method 1) for different values of , keeping the other parameters fixed.
Four values for were chosen, equal to 7.5, 10, 15 and 300 while keeping parameters n,
nRev, v and Sc fixed at 21, 3, 10 and 5 respectively. In these calculations, the induced
velocityatthebladeswascomputed,takingintoaccountalsothecontributionfromthefar
wake.TheresultsaredisplayedinFig.5.19.Thedifferencesareverysmall,thelargestbeing
whenisequalto300.
Toquantifythesedifferences,therelativepercentageerrorwasfoundusingtheequation:
ua ,c ua ,c
= 100 (5.17)
l s
ua ,c s
where suffix s and l denote the small and large value of respectively. Eqt. 5.17 was
appliedbytaking(s,l)successivelyequalto(7.50,100),(100,150)and(150,300).Thespanwise
variationof atthedifferentvaluesof(s,l)isshowninFig.5.20.Itmaybeobservedthat
therelativeerrorissmall,reachingamaximumtowardsthebladerootandtip.
0
0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
-0.2
-0.4
-0.6
-0.8
ua,c (m/s)
-1
-1.2
-1.4
-1.6
-1.8
= 7.5 deg
= 10 deg
= 15
deg = 30 deg
-2
r/R
Figure5.19Axialinducedvelocitydistributionatbladeliftinglinecomputedbyfreewakecodefor
differentvaluesof.Bladeisat3600azimuth.(Farwakemodelincludedinvelocitycalculations).
241
Chapter5DevelopmentofaFreeWakeVortexModel
becomessmallerandapproacheszeroassmallervaluesof(s,l)areused.Thishappensat
all spanwise blade positions and provides evidence of discetization convergence in the
numerical freewake model when using gradually smaller rotor time steps. Reducing the
valueof improvesnumericalaccuracybecauseitreducesthetimestep()usedinthe
timemarchingscheme.AlsoatagivenvalueofnRev,byusingsmallervaluesfor , the
wakevorticesaresegmentedintosmallersegments.Thisimprovestheaccuracywithwhich
thevortexsystemrepresentstherotorwakeasdiscussedGuptaetal.[34].Unfortunately,a
reducedtimestepwillrequireincreasedcomputationalcosts,asillustratedinFig.5.21.The
RCTFisdefinedhereasthetimetakentocarryoutthefreewakewithagivenvalueof
dividedbythetimetakenforthesamecomputationwithequalto7.50.
0
0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
-2
-4
(%)
-6
-8
-10
7.5 - 10deg 10 - 15deg 15 - 30deg
-12
r/R
Figure5.20Spanwisedistributionofpercentagerelativeerrorresultingfromdifferentvaluesof.
0.8
0.6
RCTF
0.4
0.2
0
5 10 15 20 25 30
(Deg)
Figure5.21VariationofRelativeComputationalTimeFactorwith.
242
Chapter5DevelopmentofaFreeWakeVortexModel
Effectofn
ThefreewakecalculationsusingHAWT_FWCwerealsocarriedoutatdifferentvaluesofn
(equal to 11, 21 and 31) while maintaining nRev and constant (3 revolutions and 100
respectively).Ateachvalueofn,differentsetsofvaluesfortheviscousparameters(v,Sc)
were taken. The main objective of these calculations was to provide evidence of
discretization convergence when increasing n over a wide range of values of (v, Sc). Fig.
5.22showstheaxialinducedvelocitydistributionatthebladeliftinglinecalculatedforthe
threevaluesofnwith(v,Sc)takenas(10,5).Thepresenceofdiscretizationconvergenceis
apparent,sincethediscrepancybetweenthedistributionsatn=21andn=31islessthanthat
betweenn=11andn=21.
The Relative Computation Time Factor (RCTF) variation is displayed in Fig. 5.23. In this
case,theRCTFisdefinedasthecomputationtimerequiredforagivennvaluedividedby
thatrequiredtocarryoutthesamecomputationwithn=11.Itisimportanttonotethatthe
computationaltimeisnotaffectedbythechoiceof(v,Sc)becausetheseparametersarenot
blade or wake discretization parameters. Taking n equal to 21 instead of 31 reduces the
computationaltimerequiredbyabout60%andwhiledecreasingthenumericalaccuracyby
onlyamaximumof4%atallradiallocations.
0
0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
-0.2
n=11 n=21 n=31 Experiment
-0.4
-0.6
ua,c (m/s)
-0.8
-1
-1.2
-1.4
-1.6
-1.8
n
Figure5.22Axialinducedvelocitydistributionatbladeliftinglinecomputedbyfreewakecodefor
differentvaluesofn.(v,Sc)areequalto(10,5).
243
Chapter5DevelopmentofaFreeWakeVortexModel
10
RCTF
6
1
10 15 20 25 30
n
Figure5.23VariationofRelativeComputationalTimeFactorwithn.
Figs. 5.24(a) and (b) illustrate how the calculated value of ua,c at the blade lifting line (at
r/R=0.65)varieswithnatdifferentvaluesof(v,Sc).Convergencewasobservedatallvalues
of (v,Sc). Similar convergence trends were seen at other blade spanwise locations. Higher
convergenceratesareachievedatlargervaluesof vandSc.Toquantifythedifferencesin
ua,cthroughvaryingn,therelativepercentageerrorwasfoundusingtheequation:
ua ,c l ua ,c
n = 100 (5.18)
s
ua ,c s
wheresuffixsandldenotethesmallandlargevalueofnrespectively.Eqt.5.18wasapplied
bytaking(s,l)successivelyequalto(21,11)and(31,21)forthedifferentvaluesof(v,Sc).Figs.
5.24 (a) and (b) also show the resulting values of n. The decreased relative error from
(21,11)to(31,21)providesevidenceofnumericalconvergence.
244
Chapter5DevelopmentofaFreeWakeVortexModel
-1.2
11 13 15 17 19 21 23 25 27 29 31
-1.25
Relative % Error at r/R =0.65
-1.3 n n
(21,11) (31,21)
-1.5
-1.55
-1.6
-1.65
dv10,Sc0.1
v10, Sc0.1 dv10,Sc1
v10, Sc1 dv10,Sc5
v10, Sc5 dv10,Sc10
v10, Sc10
-1.7
n
Figure5.24(a)Variationofcomputedinducedvelocityatbladeliftinglineatr/R=0.65withnfor
differentvaluesofSckeepingvconstant.(n=11,21and31).
-1
11 13 15 17 19 21 23 25 27 29 31
-1.1
-1.2
Relative % Error at r/R =0.65
n n
-1.3 (21,11) (31,21)
ua,c (m/s)
-1.6
v10, Sc10
dv10,Sc10 dv100,Sc10
v100, Sc10 v10, Sc5
dv10,Sc0.1 dv100,Sc0.1
v100, Sc0.1
-1.7
n
Figure5.24(b)Variationofcomputeinducedvelocityatbladeliftinglineatr/R=0.65withnfor
differentvaluesofvkeepingScconstantat10andthenat0.1.(n=11,21and31).
245
Chapter5DevelopmentofaFreeWakeVortexModel
Validation
After investigating the convergence trends in ua,c when increasing n, a study was then
carried out to determine how different values of (v,Sc) influence the correlation with the
experimental data of Figs. 4.59(a) and (b). Figs. 5.25(a) and (b) show the spanwise
distributions of the axial induced velocity at the blade lifting lines at different (v,Sc). The
experimentalresultsofFigs.4.59(a)and(b)arealsoincludedintheseplots.Thefollowing
observationswerenoted:
Very good agreement was obtained with the experiments for the middle blade sections
(from r/R=0.55 to 0.75) at all (v,Sc) values, except that at (100,10). In fact, at these middle
sections,thecalculatedinducedvelocityisquiteinsensitivetovaluesofvrangingfrom1to
10andtovaluesofScrangingfrom0.01to10seconds.
Theagreementisnotgoodatthebladetipandrootregion.HighsensitivitytovandScis
observedatthebladerootandtipregions.
Whenassigningsmallvaluesof vandSc,theinducedvelocitydistributionwilltendtoa
convergeddistribution.
WithreferencetoEqt.5.8,increasingboth vandScincreasetheviscouscoreradiusofthe
vortex filaments that modelthe near wake. As it may be noted from Figs. 5.25(a) and (b),
when assigning small values for these parameters (i.e use small core radii), the induced
velocity distribution at the blade will converge to one that is very close to the real
(experimental)valuesatthemiddlebladesections,butoverpredictstheinducedvelocities
atthebladetipandroot.Ontheotherhand,whenassigningincreasingvaluesforvandSc,
the discrepancy at the blade tip and root is reduced. But if these two parameters are
increased too much, the induced velocities will reduce to unrealistic low values at all
positionsalongtheblades.
InanattempttodeterminearealisticestimationofvandSc,Eqt.5.8wasusedtogetherwith
theassumptionthatthethicknessoftheviscouslayerofthevortexsheetatthepointwhen
itisshedfromthetrailingedgeofeachbladeisontheorderofthebladethickness.Thisis
equalto0.12cfortherotorconsideredinthiscasestudy.Basedonthisassumption,thecore
radii of the vortex filaments emerging from the blades (i.e. at t =0) would therefore be
approximately equal to half the blade thickness. Using these criteria, Eqt. 5.8 then yields
thattheproductof vXScequals0.332.Giventhat visfixedat10(tofollowtheguidelines
inreferences[2,10]forsmallscalerotors),thenScwouldbeequalto0.0332seconds.This
impliesthattheaxialinducedvelocitydistributionwouldbesomewherebetweenthoseat
(10,0.01)and(10,0.1).Therefore,referringtoFig.5.25(a),thedistributionwouldhavegood
agreementwiththeexperimentaloneatthemiddlebladesectionsbutnotatthebladetip
androot.Althoughthereissomelevelofuncertaintyinthecomputedinducedvelocitydue
tothefactthattheboundcirculationdistribution(refertoFig.4.55)atthebladetipandroot
regionwasextrapolated(refertosection4.3.6),oneshouldalsorecallthattheflowatblade
tip/root regions is highly 3D in character and modelling the blade as a lifting line is
insufficienthere.Thisisamajorlimitationofliftinglinemethods,asopposedtothemore
246
Chapter5DevelopmentofaFreeWakeVortexModel
accurateliftingsurfacemethodswhichcaterforthe3Deffectsaroundthebladeends.The
significanceoftheerrorfromtheliftinglinemethodsatthebladetip/rootisdependenton
theaspectratiooftherotorblades:thehighertheaspectratio,thelessisthesignificanceof
theerror.Forthisstudy,therotorbladeshadalowaspectratio(equalto5.25)andtheblade
root/tips were rectangular in shape. Consequently these errors are large. Luckily enough,
modern wind turbine (and helicopter) blades have higher aspect ratios and consequently
thislimitationofliftinglinemodelsisnothighlyinfluential.
Figs.5.26(a)and(b)plottheazimuthallyaveragedaxialinducedvelocitiesattherotorplane.
Thisplotshowsasimilarbehaviourasfortheinducedvelocityattheliftingline,i.e.thereis
alowsensitivityforvandScatthemiddlebladesections,butahighsensitivityattheblade
tipandrootregions.
In effect, v and Sc determine the viscous core radius of the wake filaments at different
vortexages.Thevariationofthecoreradiuswithvortexage(i.e.time)atvariousvaluesof
vandScmaybeobservedinFig.5.27.Sccontrolstheinitialcoresizeatthepointwherea
vortex is shed from the blade. v controls the initial core size as well as the rate of core
growthwithtime.Recallthataisanindexdenotingthevortexageandtheloweritsvalue,
theolderistheageofthevortex.ItmaybeobservedfromFig.5.27thattheoldervortices
suffer from large changes in core radius. This is because they lie on the edge of the near
wakevortexsheet,closetotheinterfacewiththefarwakewherelargefilamentstrainsare
experienced. This abrupt variation is quite unrealistic and is only due to the limitation in
whichthewakeisbeingmodelled.
247
Chapter5DevelopmentofaFreeWakeVortexModel
2.5 dv10,Sc0.1
v10, Sc0.1 dv10,Sc1
v10, Sc1
dv10,Sc5
v10, Sc5 dv10,Sc10
v10, Sc10
2
Experiment dv10,Sc0.01
v10, Sc0.01
1.5
1
ua,c (m/s)
0.5
0
0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
-0.5
-1
-1.5
-2
-2.5
r/R
Figure5.25(a)Axialinducedvelocitydistributionatbladeliftinglinecomputedbyfreewakecode
fordifferentvaluesofSckeepingvconstant.(n=21).
2.5
v1, Sc10
dv1,Sc10 dv10,Sc10
v10, Sc10 v100, Sc10
dv100,Sc10 Experiment
2 v1, S
dv1, c0.1
Sc0.1 dv10,Sc0.1
v10, Sc0.1 v100, Sc0.1
dv100,Sc0.1
1.5
1
ua,c (m/s)
0.5
0
0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
-0.5
-1
-1.5
-2
-2.5
r/R
Figure5.25(b)Axialinducedvelocitydistributionatbladeliftinglinecomputedbyfreewakecode
fordifferentvaluesofvkeepingScconstantat10andthenat0.1.(n=21).
248
Chapter5DevelopmentofaFreeWakeVortexModel
1.5
dv10,Sc0.1
v10, Sc0.1 dv10,Sc1
v10, Sc1
1 dv10,Sc5 dv10,Sc10
v10, Sc5 v10, Sc10
Experiment v10,Sc0.01
dv10,Sc0.01
0.5
0
0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
ua (m/s)
-0.5
-1
-1.5
-2
-2.5
r/R
Figure5.26(a)Azimuthallyaveragedaxialinducedvelocitydistributionatrotorplanecomputedby
freewakecodefordifferentvaluesofSckeepingvconstant.(n=21).
1.5
dv1,Sc10
v1, Sc10 dv10,Sc10
v10, Sc10 dv100,Sc10
v100, Sc10 Experiment
1
dv1,
v1, S c0.1
Sc0.1 dv10,Sc0.1
v10, Sc0.1 dv100,Sc0.1
v100, Sc0.1
0.5
0
0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
ua (m/s)
-0.5
-1
-1.5
-2
-2.5
r/R
Figure5.26(b)Azimuthallyaveragedaxialinducedvelocitydistributionatrotorplanecomputedby
freewakecodefordifferentvaluesofvkeepingScconstantat10andthenat0.1.(n=21).
249
Chapter5DevelopmentofaFreeWakeVortexModel
0.12
v10, Sc0.1
dv10, Sc1 v10, Sc5
dv10, S c5 dv100,
v100, SSc0.01
c0.01
v100, Sc1
dv100, S c1 v10, Sc0.01
dv10, Sc0.01
0.1
0.08
rc (m)
0.06
0.04
0.02
0
0 20 40 60 80 100
a
Figure5.27Variationofviscouscoreradiusofnearwaketrailingvortexfilamentswithvortexage
atr/R=0.65.(n=21, =100,nRev=3).
Figs.5.28(a)and(b)plottheazimuthaldistributionoftheaxialinducedvelocityforr/R=0.6
and 0.9 at 3.5cm downstream from the rotorplane for a fixed blade position at =00. The
freewake results from HAWT_FWC are plotted at different viscous parameters and are
compared with those measured in the wind tunnel. As a repeatability check of the
experimentaldata,twomeasuredtracesareshown:atrotorazimuthpositions00and1800,
sincetheseshouldbethesameinatwobladedrotor.ItmaybenotedinFig.5.28(a)that,
althoughthecomputedinducedvelocityinlinewiththebladeliftinglineatr/R=0.6agrees
wellwiththeexperimentalvalue,thecomparisonofthepeaktopeakvelocityisnotgood.
The peaktopeak signal predicted by the lifting line model is only a fraction of the
measured value. The main reason for this is the fact that the measured velocity signal
includestheeffectsofarealbladesectionpassingby.Thustheeffectsofsectionthickness
and a nonzero chord length are present in the measurements. In Mast et al. [55], these
effects are quantified and a method is proposed to adapt the measured velocities
accordingly.Asonemovesawayfromtherotorplane,thepeaktopeakvelocitypredicted
by the lifting line becomes more realistic. In fact, better predictions were obtained when
computingthesamevelocitytracesfurtherdownstream,at9cmfromtherotorplane(refer
toFigs.5.29(a)and(b)).
250
Chapter5DevelopmentofaFreeWakeVortexModel
1
ExpRotor0deg
ExpRotor0 ExpRotor180deg
ExpRotor180 FreeWake(v10,Sc0.01)
FreeWake(dv10,Sc0.01)
0.5
FreeWake(v10, Sc1)
FreeWake(dv10,Sc1) FreeWake(v10, Sc5)
FreeWake(dv10,Sc5) FreeWake(v100, Sc1)
FreeWake(dv100,Sc1)
0
0 30 60 90 120 150 180 210 240 270 300 330 360
-1
-1.5
-2
-2.5
-3
-3.5
-4
Azimuth Angle (deg)
figure(a):r/R=0.6
1.5
ExpRotor0deg
ExpRotor0 ExpRotor180deg
ExpRotor180 FreeWake(v10,Sc0.01)
FreeWake(dv10,Sc0.01)
FreeWake(v10, Sc1)
FreeWake(dv10,Sc1) FreeWake(v10, Sc5)
FreeWake(dv10,Sc5) FreeWake(v100, Sc1)
FreeWake(dv100,Sc1)
0.5
Axial induced velocity (m/s)
-1.5
-2.5
-3.5
-4.5
Azimuth Angle (deg)
figure(b):r/R=0.9
Figuree5.28Comparisonofaxialinducedvelocitytracescomputedbyfreewakemodelwiththat
measuredbyhotfilmat3.5cmdownstreamfromrotorplane.Rotorbladeisat0/180degreesazimuth.
fig.(a):r/R=0.6;fig(b):r/R=0.9.(n=21, =100,nRev=3).
251
Chapter5DevelopmentofaFreeWakeVortexModel
0
0 30 60 90 120 150 180 210 240 270 300 330 360
ExpRotor0
ExpRotor0deg ExpRotor180
ExpRotor180deg FreeWake(v10,Sc0.01
FreeWake(dv10,Sc0.01)
-0.5
FreeWake(v10, Sc1)
FreeWake(dv10,Sc1) FreeWake(v10, Sc5)
FreeWake(dv10,Sc5) FreeWake(v100, Sc1)
FreeWake(dv100,Sc1)
-1.5
-2
-2.5
-3
Azimuth Angle (deg)
figure(a):r/R=0.6
0
0 30 60 90 120 150 180 210 240 270 300 330 360
-0.5
ExpRotor0deg
ExpRotor0 ExpRotor180deg
ExpRotor180 FreeWake(v10,Sc0.01
FreeWake(dv10,Sc0.01)
FreeWake(v10, Sc1)
FreeWake(dv10,Sc1) FreeWake(v10, Sc5)
FreeWake(dv10,Sc5) FreeWake(v100, Sc1)
FreeWake(dv100,Sc1)
Axial induced velocity (m/s)
-1
-1.5
-2
-2.5
-3
-3.5
Azimuth Angle (deg)
figure(b)r/R=0.9
Figure 5.29 Comparison of axial induced velocity traces computed by freewake model with that
measuredbyhotfilmat9cmdownstreamfromrotorplane.Rotorbladeisat0/180degreesazimuth.
fig.(a):r/R=0.6;fig.(b):r/R=0.9.(n=21, =100,nRev=3).
Fig. 5.30 compares the vortical wake positions predicted by HAWT_FWC for different
values of (v,Sc) with the tipvortex locations measured using the smoke visualization
experiments (see section 4.2.3). It may be easily observed that for (v,Sc) equal to (10,0.1),
verygoodagreementisobtainedforthewakeexpansion,eventhoughsuchparametersdid
notyieldanaccuratepredictionfortheaxialinducedvelocityatthebladerootandtip(refer
toFigs.5.25(a)and(b)).Recallthattheselectionof(10,5)resultedinabettercorrelationfor
252
Chapter5DevelopmentofaFreeWakeVortexModel
0.70
0.65
0.60
0.55
0.50
0.45
0.40
0.35
0.30
0.25
0.20
0.4
0.6
0.8
0.2
0
Figure(a):(v,Sc)=(10,0.1)
0.70
0.65
0.60
0.55
0.50
0.45
0.40
0.35
0.30
0.25
0.20
0.2
0.4
0.6
0.7
0
Figure(b):(v,Sc)=(10,5)
0.70
0.65
0.60
0.55
0.50
0.45
0.40
0.35
0.30
0.25
0.20
0.4
0.6
0.8
0.2
0
Figure(c):(v,Sc)=(100,1)
Figure 5.30 Comparison of tip vortical wake positions computed by freewake code at different
(v,Sc) with measured tip vortex locations measured using smoke visualizations (represented by
whitedots).n=21, =100,nRev=3.
253
Chapter5DevelopmentofaFreeWakeVortexModel
theinducedvelocityatthebladerootandtip.Yetthisdoesnotyieldabetterpredictionfor
thewakeexpansion,asseeninFig.5.30(b).Ineffect,parameters(v,Sc)controltheamount
of rollup and expansion in the computed freewake geometry. Increasing both (v,Sc)
reducethewakerollupaswellasthewakeexpansion.
Wakecirculationdistribution
Fig. 5.31 illustrates the freewake plot resulting from the bound circulation at the blades
giveninFig.4.55(method1).Colourcodingisintroducedintheplottodenotethetrailing
circulation in the wake. The plot provides a better understanding of how the bound
circulation at the blades is eventually diffused into the wake. The figure shows that the
highest circulation occurs at the blade tip and root, with the circulation at the tip being
positive,whilethatattherootbeingnegative.
Figure5.31Freewakeplotwith(v,Sc)=(10,5)andn=21, =100,nRev=5.Colourcoding
representstrailingcirculationinwake.
254
Chapter5DevelopmentofaFreeWakeVortexModel
B.2.YawedConditions
This section describes the results from verification and validation process of HAWT_FWC
bymodellingtheTUDelftrotorinyawedconditions.Throughoutthisprocess,thefarwake
model parameters were kept constant. The number of the far wake helical revolutions
(nfwRev)wassetequalto10.Todeterminereasonablevaluesfortheprescribedhelicalpitch
(pfw)andwakeskewangle(s)forthefarwakemodelatboth =300and450,preliminary
calculationswerecarriedoutwiththesamefreewakemodelandanestimatewasobtained
fromthenearwakegeometry.Thecutoffdistancewaskeptequalto0.5mm.
Theverificationascarriedoutonlyfor=450whilethevalidationwasaccomplishedatboth
=300and450.
Verification
TheverificationofHAWT_FWCat =450wasperformedusinganapproachsimilartothat
applied for axial conditions. The unsteady bound circulation distribution of Fig. 4.79
(Chapter4),wereprescribedtothevortexmodel.Theuncertaintylimitsofthisdistribution
wherenotconsideredinthecalculationsrequiredforthisverificationprocess.Parameters
nRev, andnwerevariedoneatatime,whilekeepingalltheotherparametersconstant.
The induced velocities at the blade lifting lines (ua,c) at Ya=3.5cm were used as a basis for
comparison. In all these calculations, the far wake contribution was included when
computingboththenearwakegeometryandtheinducedvelocitiesattheliftinglines.
EffectofnRev
Asinaxialconditions,whenmodellingarotoroperatinginyawatconstantspeedandina
uniform wind speed, the freewake solution should attain a reasonable level wake
periodicity to be realistic. Therefore parameter nRev should be large enough to obtain the
axial induced velocity at the lifting line of the first blade at a rotor azimuth angle of 3600
equal to that at 00 and hence have wake periodicity. nRev should also be large enough to
pushthefarwakemoredownstreamandthusreducingitsinfluenceontheinductionatthe
blades. In investigating the effect of nRev, the freewake computations with HAWT_FWC
werecomputedatdifferentvaluesofnRev(equalto1,2,3,4and5)whilekeepingtheother
parameters n, , v and Sc fixed at 21, 100, 10 and 5, respectively. Figs. 5.32 and 5.33
compare the induced velocities at the blade lifting lines (at Ya=3.5cm) for the different
valuesofnRev.ToquantifythelevelofwakeperiodicityattainedateachvalueofnRev,the
error wpwascalculatedusingEqt.5.15atvariousradiallocations.Someoftheresultsare
shown in Fig. 5.34 where it may be realised that wake periodicity is achieved rapidly,
similar to what was found in axial conditions (see Fig. 5.15). Yet, when comparing the
255
Chapter5DevelopmentofaFreeWakeVortexModel
results of Fig. 5.34 with those of Fig. 5.15 it is found that the values of wp at nRev=1 are
muchhigherat=450thanat =00.Thisisbecauseoftheskewedwakeeffectinyawwhere
alargervalueofnRevisrequiredtoachieveagivenlevelofwakeperiodicitythaninaxial
conditions.
Notethatinthesecalculations,thefarwakemodelwasincludedbothtogeneratethefree
wakegeometryandtocomputetheinducedvelocitiesattheblades.Thesamecalculations
wererepeatedbutexcludingthefarwakewhenfindingtheinducedvelocitiesattheblades
and the error fw was determined using Eqt. 5.16. The results also had the trends very
similartothoseobtainedinaxialconditions.
0.4 0.4
0.2 0.2
0.0 0.0
0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
-0.2 -0.2
-0.4 -0.4
ua,c (m/s)
ua,c (m/s)
-0.6 -0.6
-0.8 -0.8
-1.0 -1.0
-1.2 -1.2
r/R r/R
Fig.(a):=900 Fig.(b):=1800
0.4 0.4
0.2 0.2
0.0 0.0
0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
-0.2 -0.2
-0.4
-0.4
ua,c (m/s)
ua,c (m/s)
-0.6 -0.6
-0.8 -0.8
-1.0 -1.0
-1.2 -1.2
-1.4 -1.4
nRev=1 nRev=2 nRev=3 nRev=4 nRev=5 nRev=1 nRev=2 nRev=3 nRev=4 nRev=5
-1.6 -1.6
r/R r/R
Fig.(c):=2700 Fig.(d):=3600
Figure 5.32 Spanwise distributions of the axial induced velocities at the blades at Ya=3.5cm
calculatedbyfreewakecodeat =450fordifferentvaluesofnRev.Thedistributionsareshownfor
differentbladeazimuthangles.
256
Chapter5DevelopmentofaFreeWakeVortexModel
Effectof
To investigate the influence of the time step size on the computed induced velocities at
=450,thefreewakecomputationswereperformedatdifferentvaluesof whilekeeping
all other parameters fixed. Four values of were chosen, equal to 7.50, 100, 150 and 300.
Parametersn,nRev, vandScweresetequalto31,3,10and5respectively.Theresultsare
showninFig.5.35.Thedifferencesaresmall,similartowhatwasobservedat =00(seeFig.
5.19).Toquantifythesedifferences,therelativepercentageerror wasfoundatdifferent
radiallocationsandbladeazimuthanglesusingEqt.5.17.Thisequationwasappliedtaking
(s,l)successivelyequalto(7.50,100),(100,150)and(150,300).Thevariationsof with at
variousradiallocationsforthedifferentvaluesof(s,l)isshowninFig.5.36.Itisevidentthat
the numerical accuracy is improved as small values of are used, thus showing
discretizationconvergenceinthenumericalfreewakemodelwhenusinggraduallysmaller
rotortimesteps.However,whencomparingthevaluesfor obtainedfor =00(Fig.5.20)
and450(Fig.5.36),itcanbeseenthattherelativeerrorsfor=450areingeneralhigherthan
thecorrespondingvaluesat=00.
0 0
0 60 120 180 240 300 360 0 60 120 180 240 300 360
-0.2 -0.2
-0.4 -0.4
-0.6 -0.6
-0.8 -0.8
ua,c (m/s)
ua,c (m/s)
-1 -1
nRev=1 nRev=1
-1.2 nRev=2 -1.2 nRev=2
nRev=3 nRev=3
-1.4 -1.4
-1.6 nRev=4
-1.6
nRev=4
nRev=5 nRev=5
-1.8 -1.8
-2 -2
(deg) (deg)
Fig.(a):r/R=0.4 Fig.(b):r/R=0.6
0 0
-0.2 0 60 120 180 240 300 360 0 60 120 180 240 300 360
-0.2
-0.4
-0.4
-0.6
-0.6
-0.8
ua,c (m/s)
ua,c (m/s)
-1 -0.8
nRev=1 nRev=1
-1.2
nRev=2 -1 nRev=2
-1.4
nRev=3 nRev=3
-1.2
-1.6 nRev=4 nRev=4
nRev=5 nRev=5
-1.8 -1.4
-2 (deg)
-1.6
(deg)
Fig.(c):r/R=0.7 Fig.(d):r/R=0.9
Figure 5.33 Variations of the axial induced velocities at the blades at Ya=3.5cm as a function of
blade azimuth angle as calculated by freewake code at =450. The results are shown for different
valuesofnRev.
257
Chapter5DevelopmentofaFreeWakeVortexModel
180
160
140 r/R=0.4
r/R=0.6
120
r/R=0.8
100
r/R=0.9
wp(%)
80
60
40
20
0
1 2 3 4 5
-20
nRev
Figure5.34VariationoftherelativepercentageerrorinwakeperiodicitywithnRevat=450.
0 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
0.4
0.2
-0.2
0
-0.4 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
-0.2
ua,c (m/s)
ua,c (m/s)
-0.6 -0.4
-0.8 -0.6
-0.8
-1 -1
= 7.5
deg = 10 deg
= 15 deg
= 30deg
= 7.5 deg
= 10
deg = 15
deg = 30 deg
-1.2 -1.2
r/R r/R
Fig.(a):=00 Fig.(b):=900
0.4 0.4
0.2 0.2
0 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
0
0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
-0.2 -0.2
ua,c (m/s)
ua,c (m/s)
-0.4 -0.4
-0.6 -0.6
-0.8 -0.8
-1 -1
= 7.5 deg
= 10
deg = 15
deg = 30
deg = 7.5
deg = 10 deg
= 15 deg
= 30 deg
-1.2 -1.2
r/R r/R
Fig.(c):=1800 Fig.(d):=2700
Figure 5.35 Spanwise distributions of the axial induced velocities at the blades at Ya=3.5cm
calculated by freewake code at =450 for different values of . The distributions are shown for
differentbladeazimuthangles.
258
Chapter5DevelopmentofaFreeWakeVortexModel
6 4
4
2
2
0
0 60 120 180 240 300 360
0
0
60 120 180 240 300 360
-2
-2
7.5 - 10deg -4
-4 10 - 15deg 7.5 - 10deg
10 - 15deg
-6 15 - 30deg -6
15 - 30deg
-8 -8
(deg) (deg)
Fig.(a):r/R=0.4 Fig.(b):r/R=0.6
6 8
4 6
4
2
2
0
0 60 120 180 240 300 360 0
-2
-2 0 60 120 180 240 300 360
-4
-4
-6 -6
-8
-8
7.5 - 10deg -10
-10 7.5 - 10deg
10 - 15deg -12 10 - 15deg
-12 15 - 30deg
-14 15 - 30deg
-14 -16
(deg) (deg)
Fig.(c):r/R=0.7 Fig.(d):r/R=0.9
Figure 5.36 Variations of the percentage relative error resulting from different values of as a
functionofbladeazimuthangleascalculatedbyfreewakecodeat=450.
Effectofn
The freewake computations were then carried out at values of n (11, 21 and 31) while
maintainingnRevand constantat3and100,respectively.Viscousmodellingparameters
(v, Sc) were kept at (10, 5). Fig. 5.37 shows the axial induced velocity distributions at the
bladesatYa=3.5cmforthedifferentvaluesofn.Thepresenceofdiscretizationconvergence
when increasing n is clear. Eqt. 5.18 was applied to determine the relative error n at the
different blade azimuth angle and radial locations. The results are displayed in Fig. 5.38.
Thevaluesfornwereingeneralfoundtobeofthesameorderasthoseobtainedat=00.
259
Chapter5DevelopmentofaFreeWakeVortexModel
0.2 0.4
n=11 n=21 n=31
0.2 n=11 n=21 n=31
0
0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 0
0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
-0.2 -0.2
ua,c (m/s)
ua,c (m/s)
-0.4 -0.4
-0.6
-0.6
-0.8
-0.8 -1
-1 -1.2
r/R r/R
Fig.(a):=00 Fig.(b):=900
0.2 0.2
ua,c (m/s)
ua,c (m/s)
-0.2
-0.3
-0.4 -0.3
-0.5
-0.4
-0.6
-0.5
-0.7
-0.6
-0.8
-0.7
-0.9
r/R r/R
Figure 5.37 Spanwise distributions of the axial induced velocities at the blades at Ya=3.5cm
calculated by freewake code at =450 for different values of n. The distributions are shown for
differentbladeazimuthangles.
260
Chapter5DevelopmentofaFreeWakeVortexModel
10 0
0 60 120 180 240 300 360
0 -2
0 60 120 180 240 300 360
-10 -4
n: 21, 11
n: 31, 21
-20 -6
n (%)
n (%)
-30 n: 21, 11 -8
n: 31, 21
-40 -10
-50 -12
-60 -14
(deg) (deg)
Fig.(a):r/R=0.4 Fig.(b):r/R=0.6
0 2
-4
-6
-6
n (%)
n (%)
-8 -8
-10
-10
-12
-12
n: 21, 11 -14 n: 21, 11
-14 n: 31, 21
n: 31, 21
-16
-18
-16 (deg) (deg)
Fig.(c):r/R=0.7 Fig.(d):r/R=0.8
Figure 5.38 Variations of the percentage relative error resulting from different values of n as a
functionofbladeazimuthangleascalculatedbyfreewakecodeat=450.
261
Chapter5DevelopmentofaFreeWakeVortexModel
Validation
Wenowcomparetheaxialinducedvelocitydistributionsatthebladeliftinglinespredicted
by HAWT_FWC at =300 and 450 with those obtained from the hotfilm measurements. A
disadvantage of the approach for the validation lies in the basic fact that the bound
circulation distributions used where derived from unsteady aerofoil theory and inflow
measurements. As alreadydiscussed in Chapter 4, thesederived circulation distributions
(refer to Figs. 4.78 and 4.79) are subject to considerable uncertainties. Recall that such
uncertainties resulted from experimental errors in wa,c which were estimated to be
approximately 0.26m/s for =300 and 0.21m/s for =450, see section 4.3.6, page 128). To
avoid any ambiguity in the validation using the hotfilm measurements in yawed
conditions,theboundcirculationsdistributionsofFigs.4.78and4.79wereprescribedwith
their uncertainty estimates, i.e. for each yaw angle (300 and 450), three different bound
circulation distributions were prescribed separately: (1) central B distribution, i.e.
circulation that was computed assuming no uncertainty in the wa,c values derived at the
rotorplanefromtheinflowmeasurements,(2)upper Bdistribution,i.e.assuminga0.26m/s
or0.21m/suncertainty(for=300and=450,respectively)inthewa,cvaluesand(3),lowerB
distribution i.e. assuming a 0.26m/s or 0.21m/s uncertainty (for =300 and =450,
respectively) in the wa,c values. In each computation, HAWT_FWC parameters n, and
nRevweresetequalto21,100and5respectively.Tobeabletoinvestigatetheinfluenceof
the viscous parameters, these computations were also repeated for different values of (v,
Sc), equal to (10, 0.1), (10, 5) and (100, 1). Throughout this validation, the values of ua,c at
3.5cm downstream (which is one of the measurement planes at which the hotfilm
measurementsweretaken)wereusedforthecomparisonratherthanthoseexactlyonthe
bladeliftinglines.
Figs. 5.39 and 5.40 compare the HAWT_FWC results for ua,c at Ya=0.035cm with the
corresponding experimental results for =300 and =450. In these plots, the spanwise
variationofua,cisplottedatfourdifferentbladeazimuthangles(00,900,1800and2700).The
freewakemodelresultsfordifferentvaluesof(v,Sc)areshown.Itisnotedthat,similarto
what was observed in axial conditions, the freewake model results for the middle blade
sectionsareinsensitivetoviscousparameters(v,Sc).Butconsiderablesensitivityispresent
atthebladerootandtipregions.However,unlikeforaxialconditions,thereisingeneral
considerable disagreement between the freewake model results and those from the
experiments. Quantitatively, this disagreement is significantly larger than the differences
obtainedinthemodelpredictionsduetodifferentvaluesof (v,Sc).
Figs.5.41and5.42plotthevariationsofua,catYa=0.035cmwiththebladeazimuthangle()
at different radial locations as predicted by the freewake model for =300 and 450. The
resultsareshownforthethreedifferentboundcirculationdistributions(central,upperand
lower distributions). Thus it is possible to analyse the uncertainty in induced velocities
predicted by HAWT_FWC resulting from the uncertainty in the prescribed bound
262
Chapter5DevelopmentofaFreeWakeVortexModel
circulations (which uncertainty is due to errors in the hotfilm measurements). Figs. 5.41
and 5.42 include also the experimental distributions, together with their respective error
bars.Itisnotedthatthevortexmodelpredictionsandtheexperimentalresultsagreevery
well at =00. But for the other blade positions, the disagreement is not that good and the
percentagediscrepancymaysometimesbeabove100%.Thepercentagediscrepancycanbe
veryhighforthesimplereasonthattheinducedvelocitiesareverysmallandclosetozero.
Yetitshouldbeemphasizedthatthispercentagediscrepancyisverysensitivetothelevelof
uncertaintyestimatedinthehotfilmmeasurements.Itshouldbekeptinmindthatthetwo
uncertaintiesbothariseduetotheerrorsinthemeasurements.Theupperuncertaintyvalue
fortheinflowvelocitywa,c(i.e.wa,c+0.26or+0.21m/s,for=300and450respectively)resulted
in a higher bound circulation distribution, which in turn predicted a higher (negative)
distribution for ua,c. Also, this upper uncertainty yielded the lower distribution for ua,c (as
derivedbyEqt.4.57).Forthisreason,ifinFigs.5.41and5.42wecomparethevortexmodel
1.2
v 10,Sc0.1
dv10 Sc 0.1 v 10,Sc0.1
dv10 Sc 0.1
0.8
v 10,Sc5
dv10 Sc 5 0.8 v 10,Sc5
dv10 Sc 5
0.4 v 100,Sc1
dv100 Sc 1 v 100,Sc1
dv100 Sc 1
0.4
Experiment Experiment
0 0
0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
ua,c (m/s)
Fig.(a):=00 Fig.(b):=900
1.2 1.2
v 10, Sc0.1
dv10 Sc 0.1 v 10,Sc0.1
dv10 Sc 0.1
0.8 v 10, Sc5
dv10 Sc 5 0.8 v 10,Sc5
dv10 Sc 5
v 100, Sc1
dv100 Sc 1 v 100,Sc1
dv100 Sc 1
0.4 0.4
Experiment Experiment
0 0
0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
ua,c (m/s)
ua,c (m/s)
-0.4 -0.4
-0.8 -0.8
-1.2 -1.2
-1.6 -1.6
-2 -2
-2.4 -2.4
r/R r/R
Fig.(c):=1800 Fig.(d):=2700
Figure 5.39 Comparison of the spanwise distributions of ua,c at Ya=3.5cm predicted by freewake
vortexmodelwithexperimentalresultsfor=300.Vortexmodelresultsareshownfordifferentblade
azimuthanglesanddifferentvaluesfor(v,Sc).Thecentral Bdistributionisusedforcomputations
withn=21,=100,nRev=5.
263
Chapter5DevelopmentofaFreeWakeVortexModel
results HAWT_FWC_Upper with the upper extreme value of the experimental values,
then the correlation of the vortex model results with the experimental values improves
considerably.Insomecases,theuncertaintylimitsoftheHAWT_FWCpredictionandthose
oftheexperimentalresultsoverlaponeanother.Thisisapositivefactorthatprovidesmore
confidence that, given the uncertainties in the hotfilm measurements were smaller the
agreementcouldhavebeenbetter.Thisoverlappingismostlyobservedat =450(Fig.5.42)
between blade azimuth angles 00 and 1800. However, between blade azimuth angles 1800
and 3600, the agreement is still not good even when the uncertainties in the hotfilm
measurements are taken in account. This is because the amount of disagreement is
considerably larger that the uncertainty levels. However there is considerable evidence to
show that the large disagreement is due to the flow obstruction from the centre body
structure of wind tunnel model resulting from rotor yaw. The photos of the smoke
visualisationhelpsupportthisevidence.
0.6 0.8
v 10,Sc0.1
dv10 Sc 0.1 v 10,Sc0.1
dv10 Sc 0.1
0.4 v 10,Sc5
dv10 Sc 5 v 10,Sc5
dv10 Sc 5
0.4
v 100, Sc1
dv100 Sc 1 v 100, Sc1
dv100 Sc 1
0.2
Experiment Experiment
0 0
0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
ua,c (m/s)
-0.6 -0.8
-0.8
-1.2
-1
-1.2 -1.6
r/R r/R
Fig.(a):=00 Fig.(b):=900
0.8 1
v 10, Sc0.1
dv10 Sc 0.1 v 10, Sc0.1
dv10 Sc 0.1
v 10, Sc5
dv10 Sc 5 v 10, Sc5
dv10 Sc 5
0.6
0.4
v 100, Sc1
dv100 Sc 1 v 100, Sc1
dv100 Sc 1
Experiment 0.2 Experiment
0
0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
-0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
ua,c (m/s)
ua,c (m/s)
-0.4
-0.6
-0.8
-1
-1.2
-1.4
-1.6 -1.8
r/R r/R
Fig.(c):=1800 Fig.(d):=2700
Figure 5.40 Comparison of the spanwise distributions of ua,c at Ya=3.5cm predicted by freewake
vortexmodelwithexperimentalresultsfor=450.Vortexmodelresultsareshownfordifferentblade
azimuthanglesanddifferentvaluesfor(v,Sc).Thecentral Bdistributionisusedforcomputations
withn=21,=100,nRev=5.
264
Chapter5DevelopmentofaFreeWakeVortexModel
1.2 1.2
Experiment Experiment
0.8 HAWT_FWC_Central 0.8 HAWT_FWC_Central
0.4
HAWT_FWC_Upper
HAWT_FWC_Lower
0.4
HAWT_FWC_Upper
HAWT_FWC_Lower
0 0
0 60 120 180 240 300 360 0 60 120 180 240 300 360
ua,c (m/s)
ua,c (m/s)
-0.4 -0.4
-0.8 -0.8
-1.2 -1.2
-1.6 -1.6
-2 -2
-2.4 -2.4
(deg) (deg)
Fig.(a):r/R=0.4 Fig.(d):r/R=0.7
1.2 1.2
Experiment
Experiment
0.8 HAWT_FWC_Central 0.8
HAWT_FWC_Upper
HAWT_FWC_Central
HAWT_FWC_Upper
0.4 HAWT_FWC_Lower 0.4
HAWT_FWC_Lower
0 0
0 60 120 180 240 300 360 0 60 120 180 240 300 360
ua,c (m/s)
ua,c (m/s)
-0.4 -0.4
-0.8 -0.8
-1.2 -1.2
-1.6 -1.6
-2 -2
-2.4 -2.4
(deg) (deg)
Fig.(b):r/R=0.5 Fig.(e):r/R=0.8
1.2 1.2
Experiment Experiment
0.8 HAWT_FWC_Central 0.8 HAWT_FWC_Central
HAWT_FWC_Upper HAWT_FWC_Upper
0.4 HAWT_FWC_Lower 0.4 HAWT_FWC_Lower
0 0
0 60 120 180 240 300 360 0 60 120 180 240 300 360
ua,c (m/s)
ua,c (m/s)
-0.4 -0.4
-0.8 -0.8
-1.2 -1.2
-1.6 -1.6
-2 -2
-2.4 -2.4
(deg) (deg)
Fig.(c):r/R=0.6 Fig.(f):r/R=0.9
Figure 5.41 Comparison of the distributions of ua,c with at Ya=3.5cm predicted by freewake
vortexmodelwithexperimentalresultsfor =300.Vortexmodelresultsareshownasafunctionof
blade azimuth angle and at different radial location. The central, upper and lower B distributions
areusedforcomputationswithn=21,=100,nRev=5,(v,Sc)=(10,5).
265
Chapter5DevelopmentofaFreeWakeVortexModel
1.2 1.2
Experiment Experiment
0.8 0.8 HAWT_FWC_Central
HAWT_FWC_Central
HAWT_FWC_Upper HAWT_FWC_Upper
0.4
HAWT_FWC_Lower
0.4
HAWT_FWC_Lower
0 0
0 60 120 180 240 300 360 0 60 120 180 240 300 360
ua,c (m/s)
ua,c (m/s)
-0.4 -0.4
-0.8 -0.8
-1.2 -1.2
-1.6 -1.6
-2 -2
-2.4 -2.4
r/R r/R
Fig.(a):r/R=0.4 Fig.(d):r/R=0.7
1.2 1.2
Experiment Experiment
0.8 HAWT_FWC_Central 0.8 HAWT_FWC_Central
0.4
HAWT_FWC_Upper
0.4
HAWT_FWC_Upper
HAWT_FWC_Lower HAWT_FWC_Lower
0 0
0 60 120 180 240 300 360 0 60 120 180 240 300 360
ua,c (m/s)
ua,c (m/s)
-0.4 -0.4
-0.8 -0.8
-1.2 -1.2
-1.6 -1.6
-2 -2
-2.4 -2.4
r/R r/R
Fig.(b):r/R=0.5 Fig.(e):r/R=0.8
1.2 1.2
Experiment Experiment
0.8 HAWT_FWC_Central 0.8 HAWT_FWC_Central
HAWT_FWC_Upper HAWT_FWC_Upper
0.4 0.4
HAWT_FWC_Lower HAWT_FWC_Lower
0 0
0 60 120 180 240 300 360 0 60 120 180 240 300 360
ua,c (m/s)
ua,c (m/s)
-0.4 -0.4
-0.8 -0.8
-1.2 -1.2
-1.6 -1.6
-2 -2
-2.4 -2.4
r/R r/R
Fig.(c):r/R=0.6 Fig.(f):r/R=0.9
Figure 5.42 Comparison of the distributions of ua,c with at Ya=3.5cm predicted by freewake
vortexmodelwithexperimentalresultsfor =450.Vortexmodelresultsareshownasafunctionof
blade azimuth angle and at different radial location. The central, upper and lower B distributions
areusedforcomputationswithn=21,=100,nRev=5,(v,Sc)=(10,5).
266
Chapter5DevelopmentofaFreeWakeVortexModel
The photo in Fig. 5.43 shows that for a positive yaw angle of =450, at blade positions
=1800and3600,thetowerstructurelieswithinthetipvortexpath.Itisshowninthisphoto
thatthethirdtipvortexcore,aroundwhichthereisaconsiderablyhighvorticitycontent,is
infactcollidingwiththetowerstructure.Thisvortexcoreisstilltooclosetotherotorplane
(withinanaxialdistanceofaboutoneR).AsimpleapplicationoftheBiotSavartlawonthe
vortexcoresuggeststhatthisvorticallocationinfluencesconsiderablythetotalinducedat
the rotorplane, and even more at Ya=3.5cm, at which the comparison of the induced
velocitiesisbeingmade.
Theobstructionfromthemodelcentrebodycausedareductionintheflowvelocitiesonthe
upwindside side of the rotor. Consequently the values for wa,c obtained from the hotfilm
experimentsbetween =1800and3600(seeFig.4.19and4.20,Chapter4)arelowerthanwhat
wouldhavebeenmeasuredifnosuchcentrebodystructurewasnotpresent.Thisresulted
in lower experimental axial induced velocities (ua,c) when applying Eqt. 4.57. This is the
reasonwhyinFigs.5.41and5.42theexperimentalvaluesareconsiderablylowerthanthe
HAWT_FWC predictions between =1800 and 3600 and thus resulting in a large
disagreement.Theinterferenceoftheturbinecentrebodywasnotmodelledinthefreewake
vortex model. Looking at Fig. 5.43 explains that for the upwind side, the inboard radial
locations of the blade suffer the most from the flow obstruction since these are closer to
turbinecentrebody.ThisisinfactnotedinFig.5.42for =450,whereitisevidentthatthe
inboard radial locations (r/R= 0.4, 0.5 and 0.6, Figs. (a), (b), (c)) have the largest
disagreement between =1800 and 3600, larger then the disagreement obtained at the
outboard blade regions (Figs. (d), (e), (f)). On the other hand, for blade azimuth angles
between =00 and1800(i.e.onthedownwindsideofrotorplane),thetipvorticesarenot
obstructedbythecentrebody(asmaybeseeninFig.5.43).Thisexplainswhyamuchbetter
correlationwasachievedbetweentheexperimentalandfreewakevortexmodelresults.
The above arguments provide more confidence that the freewake vortex model is
reasonably accurate even in yawed conditions, despite the fact that considerable
disagreementswiththeinducedvelocitiesfromtheexperimentswerefound.
267
Chapter5DevelopmentofaFreeWakeVortexModel
Downwindside
(00<<1800)
Turbinecentrebody
Upwindside
(1800<<3600)
Figure 5.43 Smoke visualization photos of the TUDelft turbine at =450. Note that on upwind
side,thetipvortexcoretrajectoryisobstructedbythecentrebodyoftheturbine.
ThetipvorticallocationspredictedbyHAWT_FWCarenowcomparedwiththetipvortex
core locations measured during the smoke visualisation experiments. The vortex plots at
=300and450areshowninFigs.5.44and5.45,respectively.Thewhitedotsintheseplots
indicatethetipvortexcorelocationsmeasuredfromthesmokevisualization.Thefreewake
plotsareshownfortheinstantinwhichthefirstandsecondbladesarelocatedat =900and
1800,andthusatthesamepositionsatwhichthesmokevisualisationphotoswheretaken.
In these vortex model computations, parameters n, and nRev were set equal to 21, 100
and 3.5 respectively and the central bound circulation distribution were used, without
taking the corresponding uncertainties into account. To investigate the influence of the
viscous parameters, these computations were also repeated for different values of (v, Sc),
equal to (10, 0.1), (10, 5) and (100,1). As already explained before, for axial conditions
parameters(v,Sc)controltheamountofrollupinthemodelledfreewake.Thelargertheir
values, the less is the rollup and the deformation of the wake. It can be noted from Figs.
5.44and5.45thatthevortexsheetpitchiswellpredictedatallthethreesetsofvaluesof(v,
Sc).Yetitisseenthat,atboth =300and450,(v,Sc)valuesequalto(10,0.1)yieldthebest
predictedwakeexpansioncorrelationwiththeexperimentalmeasurements.Recallthatthe
samewasalreadynotedearlierinaxialconditions,seeFig.5.30.Recallalsothatfromthe
validationforaxialconditions(section5.3.2,page246)thatparametervwastakenequalto
268
Chapter5DevelopmentofaFreeWakeVortexModel
Fig.(a):
(v,Sc)=(10,0.1)
Fig.(b):
(v,Sc)=(10,5)
Fig.(c):
(v,Sc)=(100,1)
Figure 5.44 Comparison of tip vortical wake positions computed by freewake code at different
(v,Sc)withtipvortexlocationsmeasuredusingsmokevisualizations(representedbywhitedots)at
=300.n=21, =100,nRev=3.5.
269
Chapter5DevelopmentofaFreeWakeVortexModel
Fig.(a):
(v,Sc)=(10,0.1)
Fig.(b):
(v,Sc)=(10,5)
Fig.(c):
(v,Sc)=(100,1)
Figure 5.45 Comparison of tip vortical wake positions computed by freewake code at different
(v,Sc)withtipvortexlocationsmeasuredusingsmokevisualizations(representedbywhitedots)at
=450.n=21, =100,nRev=3.5.
270
Chapter5DevelopmentofaFreeWakeVortexModel
10basedontheresultsfromreferences[2,10].Scwastakenequalto0.1inordertoyielda
vortexsheetviscouslayerthicknessatthebladestrailingedgesapproximatelyequaltothe
bladethickness.
The very good agreement achieved in the correlation of tip vortical locations with the
smokevisualisationmeasurementsaddstotheconfidenceaboutthevalidityofthenewly
developedfreewakevortexmodelHAWT_FWC.
Wakecirculationdistribution
Figs. 5.46 and 5.47 show the free wake plots from the unsteady bound circulation at the
blades (Figs. 4.78 and 4.79, uncertainty limits not taken into account) at =300 and 450.
Colourcodingisusedtorepresentthetrailingorshedcirculationdistributioninthewake.
Foreachyawangle,thewakeplotisshowntwice:Fig.(a)illustratesthetrailingcirculation
whileFig.(b)illustratestheshedcirculationinthenearfreewake.Twoobservationscanbe
madehere.Firstofall,thetrailingcirculationtendstobeconcentratedatthetipandroot
regions,withthecirculationatthetipbeingpositivewhilethatattherootbeingnegative.
This is very similar to what was seen in axial conditions (see Fig. 5.31). Secondly, since
yawedconditionsarebeingdealtwithandalsobecausetheboundcirculationattheblades
is unsteady, some shed circulation results in the wake. However, in general, the shed
circulation levels are much smaller that the trailing circulation, as it may be seen in Figs.
5.46and5.47.
271
Chapter5DevelopmentofaFreeWakeVortexModel
Fig.(a)Trailingcirculation
Fig.(b) Shedcirculation
Figure5.46Freewakeplotwith(v,Sc)=(10,5)andn=21, =100,nRev=5at=300.Colour
codingrepresentsthetrailingorshedcirculationinwake.
272
Chapter5DevelopmentofaFreeWakeVortexModel
Fig.(a)Trailingcirculation
Fig.(b) Shedcirculation
Figure5.47Freewakeplotwith(v,Sc)=(10,5)andn=21, =100,nRev=5at=450.Colour
codingrepresentsthetrailingorshedcirculationinwake.
273
Chapter5DevelopmentofaFreeWakeVortexModel
5.4Conclusions
A freewake vortex model developed in this project was verified and validated by
modellingtheTUDelftmodelturbineinaxialandyawedconditions.Inboththeverification
andvalidation,theboundcirculationestimatedfromthehotfilminflowmeasurementsand
the unsteady aerofoil model (as described in section 4.3.1) was prescribed to the vortex
modeltogenerateafreewakeplotandcalculatetheinducedvelocitiesattherotorplaneor
at3.5cmdownstream.Intheverification,aparametricanalysiswascarriedouttoquantify
thenumericalerrorsandcomputationaltimeofthevortexmodelatdifferentlevelsofblade
andwakediscretization.Inthevalidation,theresultsfromthefreewakevortexmodelwere
comparedwiththeexperimentaldataobtainedfromthemeasurementstakenintheopen
jetwindtunnel.Twoseparatecomparisonswerecarriedout:(1)theinducedvelocitiesfrom
the freewake model were compared with those obtained from the inflow measurements;
(2) the tip vortical locations from the freewake model plots were compared with the tip
vortex location obtained in the smoke visualization measurements. A disadvantage of the
approachadoptedforthisvalidationwasthattheprescribedboundcirculationdistribution
wasderivedfromtheinflowmeasurements.Thus,thefreewakevortexmodelpredictions
were not fully independent of the experimental data. For axial conditions, the derived
bound circulation was reliable but for yawed cases considerable uncertainty was present
duetoerrorsintheinflowmeasurements.Toavoidambiguity,inthevalidationinyawed
cases(at =300and450),theinducedvelocitiesfromthefreewakemodelwerecomputed
takingintoaccounttheuncertaintylimitsoftheprescribedboundcirculationsthatresulted
fromerrorsintheinflowmeasurements.
The validation process examined how the choice of viscous modelling parameters
influences the accuracy of the calculated induced velocities when comparing it with the
experimentaldata.Itwasfoundthat,forthemiddlebladesections,thecalculatedinduced
velocities at the blade lifting lines are quite insensitive to a wide range of values of the
viscousparameters.However,highsensitivitytotheviscousparameters wasnotedatthe
bladeroot/tipregions.Thisbehaviourwasobservedinbothaxialandyawedconditionsof
therotor.
When modelling axial conditions, very good correlation with the experimental induced
velocitiesdatawasobtainedatthemiddlesectionsoftheblade.However,atthebladetip
and root regions, this correlation was very much dependent on the arbitrary choice of
viscousparameters.Inthebladetipandrootregions,theflowishighly3Dbynatureand
thisisnotusuallycateredforaccuratelybyliftinglineblademodels.Luckily,thissourceof
inaccuracyisnotveryproblematicwhenmodellingbladeswithahighaspectratio.
Inyaw,thecorrelationoftheinducedvelocitiespredictedbythefreewakemodelwiththe
correspondingexperimentalvalueswasnotasgoodasinaxialconditions.Thecorrelation
wasreasonablygoodforbladeazimuthpositions0<<1800(downwindsideofrotordisc).
274
Chapter5DevelopmentofaFreeWakeVortexModel
Butfor180<<3600 (upwindsideofrotordisc),thefreewakepredictionswereconsiderably
differentthanthosefromtheexperimentsatboth =300and450.Theexperimentalinduced
velocities were higher than the freewake values and it is very probable that this resulted
from the blockage from the centrebody structure of the model turbine. The smoke
visualization photos revealed that the centrebody obstructed the flow for 180<<3600
(upwind side of rotor disc) when the rotor was yawed. This flow interference was not
includedinthefreewakevortexmodelling.
Finally, the vortical locations of the free wake geometries predicted at different values of
viscousmodellingparameterswerecomparedwiththetipvortexlocationsmeasuredusing
smokevisualizationexperiments.Ingeneral,theagreementwasverygood,althoughitwas
better for those calculated using low values of the viscous parameters that correspond to
smallerfilamentcoreradii.
275
Chapter5DevelopmentofaFreeWakeVortexModel
276
Chapter6AerodynamicAnalysisoftheNRELPhaseVIWindTurbine
6.AerodynamicAnalysisoftheNRELPhaseVI
WindTurbine
6.1Introduction
ThischapterdescribesindetailtheaerodynamicanalysiscarriedoutontheNRELPhase
VIrotor.TheexperimentaldatacollectedintheNASAAmeswindtunnelduringtheUAE
experiments was used for the analysis. This data mainly consisted of blade pressure
measurements from which the normal and tangential coefficients (Cn and Ct) could be
found,togetherwiththelocalflowanglemeasurements(LFA)(seeFigs.2.1and2.2,Chapter
2).Torqueandbladebendingmomentmeasurements,takenusingstraingaugetechniques,
werealsousedfortheanalysis.
In this chapter, a novel approach is described for coupling the blade pressure
measurements andafreewakevortexmodeltoderivetheangleofattackdistributionsat
the blades together with the inflow distribution at rotorplane. The developed freewake
vortexmodelHAWT_FWC(seeChapter5)wasusedforthisanalysis.Bothaxial(i.e.steady)
andyawed(i.e.unsteady)conditionswereinvestigated.
Theangleofattackdistributionsandinflowdistributionsattherotorplanederivedfromthe
bladepressuremeasurementsusingHAWT_FWCwereusedtoobtainnewaerofoildatafor
BEMbased codes. For yawed conditions, new inflow corrections were also derived, apart
fromtheaerofoildata.Thenewaerofoildataandinflowcorrections(foryawedconditions)
were then used in BEM code HAWT_BEM to recompute the aerodynamic loads on the
NREL rotor. The results from the BEM code were compared with those measured in the
tunnel. In thisway, it was possible to assess the reliability of the BEM theory when more
accurateaerofoildataandinflowcorrectionmodelsareavailable.Thederivedaerofoildata
and inflow corrections could be very useful to develop improved engineering models for
BEMbaseddesigncodesaswillbeexplainedlateroninChapter7.
The new aerofoil data is more accurate than that from 2D static wind tunnel experiments
since the former was directly derived from the rotor experimental measurements. It
accounts for both rotationalaugmentation (stalldelay) andunsteadyeffects(e.g.dynamic
stall) at high angles of attack. Thus the new aerofoil data can be regarded as 3D aerofoil
data.Yetthereliabilityofthisnewdatadependsontheaccuracywithwhichthefreewake
vortexmodelcalculatestheangleofattackfromtheknownbladepressuremeasurements.
Althoughthefreewakevortexmodelwasalreadyverifiedandvalidatedearlierusingthe
277
Chapter6AerodynamicAnalysisoftheNRELPhaseVIWindTurbine
TUDelftrotor(seeChapter5),itwasfeltthatadditionaltestswererequiredtoevaluatethe
methodbeingproposedtofindtheangleofattack.Thefollowingtestswerecarriedout:
1. ForeachoperatingconditionontheNRELrotorinbothaxialandyawedconditions,
the freewake model computed the local inflow angles (LFA) at the positions where
inflow probes where installed. The LFAs predicted by the freewake model were
comparedwiththosemeasuredbytheinflowprobes.
2. For axial conditions only, the new 3D aerofoil data derived by HAWT_FWC was
comparedwiththatderivedfromCFDatRisobyJohansenetal.[40].
3. For one operating condition in yaw, the inflow distribution by HAWT_FWC was
comparedwiththatpredictedbyAWSM,afreewakevortexmodeldevelopedatECN
[26].
Thischapterisbeingorganizedinthreeparts:
A. Section6.2describesbrieflytheNRELPhaseVIexperimentsintheNASAAmes
tunnelandtheexperimentaldataselectedforthisstudy.
B. Section6.3describesindetailtheapproachusedforfindingtheangleofattack
from the blade pressure measurements using the freewake vortex model
HAWT_FWC. The results for the angle of attack, aerofoil data and inflow
distributionsarepresentedanddiscussed.
C. Section6.4presentsthecalculationsperformedontheBEMmodelHAWT_BEM
with the new aerofoil data and inflow corrections. The results are compared
withthoseperformedwithstandard2Dstaticaerofoildata.
278
Chapter6AerodynamicAnalysisoftheNRELPhaseVIWindTurbine
6.2NASA/AmesUAEWindTunnelDataUsed
The NREL Phase VI rotor is a twobladed 10.1 m diameter wind turbine rotor with a
ratedpowerof19.8kW(Fig.6.1).ThebladegeometryisbasedontheS809aerofoil.Fig.6.2
gives the chord and twist distributions of the blade. Further details regarding the blade
designmaybefoundin[29].Oneofthebladeswasequippedwithpressuretapsat0.30R,
0.47R,0.63R,0.80Rand0.95Rtoacquiredetailedsurfacepressuredata(Fig.6.3).Ateachof
the full pressure tap distributions, pressures were integrated to obtain Cn, Ct and Cm. The
blade was also equipped with fivehole pressure probes at 0.34R, 0.51R, 0.67R, 0.84R and
0.91R to measure the local inflow angle (LFA) (see Fig. 6.3 and 6.4). Various loads were
measured using straingauge techniques. These included the blade root flap and edge
momentsandthelowspeedshafttorque.
Figure6.1TheNRELPhaseVIwindturbineintheNASAAmeswindtunnel(Source:[35]).
279
Chapter6AerodynamicAnalysisoftheNRELPhaseVIWindTurbine
25 0.8
twist 0.7
20 chord
0.6
Blade twist (deg) 15
10 0.4
0.3
5
0.2
0
0 1 2 3 4 5 6 0.1
-5 0.0
r (m)
Figure6.2Bladetwistandchorddistributions(Source:[35]).
Figure6.3Locationofbladesurfacepressuretappingsandfiveholeprobes(Source:[35]).
280
Chapter6AerodynamicAnalysisoftheNRELPhaseVIWindTurbine
5holedprobe S809Aerofoil
Probemount
isaprobespecificangle
Figure6.4Installationoffiveholepressureprobe(Source:[35]).
In the UAE Phase VI experiments, the turbine was tested over a very wide range of
operatingconditions.Detailsoftheseexperimentalcampaignsmaybefoundinreferences
[35,73].Thisstudyconsideredbothaxialandyawedconditionswiththerotoroperatedin
theupwindconfigurationforthecasesshownintable6.1.Thebladepitchanglewassetto
30andtherotorconinganglewas00.Yawanglesabove300werenotincludedinthisstudy
since their measurements were expected to be considerably affected by the flow
disturbancesfromthebox/boominstalledupstreamoftherotor.Thisstudyonlyconsidered
theexperimentaldatacollectedwiththeSandHconfigurations(refertoreference[35]).
Table6.1Operationalconditionsconsideredinthisstudy.
DataFile Yaw Air Windspeed Rotational Tip Measured Measured
Angle Density (m/s) Speed(rpm) Speed RotorThrust RotorPower
(kg/m3) Ratio Coefficient Coefficient
S0500000 0 0 1.244 5.0 71.7 7.58 0.563 0.361
S0700000 00 1.246 7.0 71.9 5.43 0.481 0.362
S1000000 0 0 1.246 10.0 72.1 3.81 0.344 0.207
S1300000 00 1.227 13.1 72.1 2.91 0.245 0.092
S1500000 00 1.224 15.1 72.1 2.53 0.208 0.054
S2000000 0 0 1.221 20.1 72.0 1.90 0.153 0.021
S2500000 00 1.220 25.1 72.1 1.52 0.126 0.015
S0500300 30 0 1.244 5.0 71.7 7.58 0.43 0.23
S0700300 300 1.246 7.0 71.8 5.43 0.4 0.26
S1000300 30 0 1.246 10.1 72.0 3.81 0.29 0.17
S1300300 300 1.227 13.0 72.2 2.91 0.23 0.11
S1500300 300 1.225 15.1 72.2 2.53 0.19 0.07
281
Chapter6AerodynamicAnalysisoftheNRELPhaseVIWindTurbine
Inthetableabove,thevaluesshownforthethrustandpowercoefficientsaretheaveraged
valuesofthosemeasuredoveronewholerotorrevolution.
For each operating condition, data records of the surface pressures were taken for 36
individualrotorrotations(cycles).Fig.6.5showstypicalvariationsforCnobtainedateach
individualcycleat0.3Ratyaw300andawindspeedof15m/s.Inthisfigure,datafromfive
different cycles are shown. It may be observed that for blade azimuth angles between
approximately1200and2400,thecycletocyclevariationisverysmall.Asitwillbenoted
lateron,atthisregionofbladeazimuthangle,theangleofattackissmallandtheflowis
attached.Fortheotherbladeazimuthangles,thecycletocyclevariationislargeandthis
is due to the high angles of attack where flow separation influences reduce the
repeatabilityofthemeasureddata.Inallcases,themean(indicatedbythesolidblackline
inFig.6.5)isconsideredfortheanalysisusingthefreewakecode.
3.5
2.5
Cn
1.5
0.5
0
0 60 120 180 240 300 360
Blade Azimuth Angle (deg)
Figure6.5Variationofthenormalcoefficient(derivedfrombladesurfacepressuremeasurements)
withbladeazimuthangleforyaw30deg,U=15m/sandr/R=0.3.Thevaluesforfiveseparatecycles
areshownandthecyclemeanisindicatedbytheboldcurve.
282
Chapter6AerodynamicAnalysisoftheNRELPhaseVIWindTurbine
6.3EstimatingtheAngleofAttackfromBladePressure
MeasurementsusingtheFreewakeVortexModel
6.3.1Methodology
ThecomputationalprocedureusedtoderivetheunsteadyangleofattackusingtheNREL
experimentaldataandthefreewakemodelisnowdescribed.Themethodwasappliedto
bothaxialandyawedconditions.Inthelatterconditions,thesituationisunsteadyandthe
derivedangleofattackisafunctionofthebladeazimuthangle().Theexperimentaldata
wasonlyavailableforoneblade.Thesamedatawasappliedfortheotherblade,obviously
takingintoaccounttherespectivebladeazimuthangles.TheprocedureisillustratedinFig.
6.6.Itconsistedofthefollowingsteps:
Step1:ThemeasuredvaluesofQNORM,CnandCtatthedifferentradiallocationsandblade
azimuthalangleswerefirstloadedintothecodefromtheNRELexperimentaldatabase(S
configuration[35]).
Step2:AnangleofattackdistributionwasthenassumedandtheequationshowninFig.2.1
(refertoChapter2)wasusedtoobtainanestimateforClatthedifferentradiallocationsand
bladeazimuthalangles.
Step3:Therelativeflowvelocityattheblades,Vr,wasfoundfromQNORMusingtheequation
fordynamicpressuregivenby
1
QNORM = Vr (6.1)
2
2
Step 4: The KuttaJoukowksi theorem for a lifting line was used to find the bound
circulationsatthepressuremeasurementstations:
1
B = ClVr c (6.2)
2
Step 5: A double interpolation was made to obtain a bound circulation at the blades as a
function of radial position and blade azimuth position (B[r/R,]), with the bound
circulationatthebladetipandrootsettozero.Forinterpolating,asplineinterpolationwas
initiallyusedtoobtain asafunctionofr/R.Butitwasfoundthatthenumberofpressure
measuring stations along the blade (equal to 5) was not enough to give an accurate
representationoftheboundcirculationdistributionalongthebladewithasplinemethod.
This caused the spline method to yield unrealistic jumps. Alternatively, a linear
interpolatingmethodhadtobeadopted.Asplineinterpolationwasusedonlytointerpolate
fordifferentbladeazimuthpositions.
283
Chapter6AerodynamicAnalysisoftheNRELPhaseVIWindTurbine
InputQNORM,Cn,Ct frommeasurements
Assumedistributionfor
FindB fromEqt.6.2
PrescribeB tofreewakevortexmodel
ComputeCl,Cdp,and
Inductionatrotorplane
andcarryoutfreewakecomputations
Figure6.6ComputationalprocedureusedtoderivetheangleofattackfromCnandCtusingfree
wakevortexmodel.
Step6:theinterpolatedboundcirculationdistribution(B[r/R,])wasthenprescribedtothe
freewake code and the modelled rotor was set to rotate to generate the freewake
extending downstream for a given number of rotations. This was used to calculate the
unsteady 3D induced velocities at the lifting line of each rotor blade from which a new
angleofattackestimatewasobtainedusingtheequationderivedfromFig.2.1(Chapter2):
= tan 1 ( )
Vn
Vt
(6.3)
Step 7: the new angle of attack was used in Eqt. 6.2 to yield a new bound circulation
distribution and the whole process was repeated until convergence in the angle of attack
wasobtainedatallradiallocationsandrotorazimuthangles.
Onceaconvergedsolutionwasobtained,thefinaldistributionsoftheunsteadyvaluesofCl
andCdpwerederivedinaccordancewiththeequationsofFig.2.1.Fromknowledgeofthe
angleofattackvariationswithrotorazimuthangle,itwaspossibletoderivethetimerateof
change of angle of attack, together with the hysterisis loops for Cl and Cdp. The
computedfreewakegeometrieswereusedtocalculatetheunsteadyinductiondistributions
at the rotorplane. The latter were then used together with the derived lift and drag
284
Chapter6AerodynamicAnalysisoftheNRELPhaseVIWindTurbine
coefficients to evaluate the rotor/blade loads as a function of rotor azimuth angle. These
loads were compared with those measured during the wind tunnel experiments. Finally,
usingthefreewakevortexmodel,itwaspossibletocomputethelocalinflowangles(LFA)
at the measuring tip of the flow direction probes. These were also compared with the
correspondingexperimentalvalues(Hconfiguration[35]).TheLFAwascomputedusing
the same equation for the angle of attack (Eqt. 6.3) but with flow components Vn and Vt
calculatedatapointonthechordline0.8cinfrontoftheleadingedge(refertoFigs.2.1,2.2
and6.4)
The above procedure was repeated for each of the operating conditions listed in table 6.1
(page 281). In these calculations the flow disturbances due to the tower and nacelle were
ignored.Structuraldynamicinfluenceswerealsoneglected.
6.3.2AxialConditions
This section describes the computational work carried out for axial conditions (=00) for
windspeedsU=5,7,10,13,15,20and25m/s.
.
A.SelectingFreeWakeModelParameters
SelectionofHAWT_FWCparametersn,andnRev
nRevwassetequalto3forwindspeedsequalto13m/sandhigher.Forlowerwindspeeds
(5, 7 and 10m/s), the pitch of the wake vortex sheets is smaller due to a higher tip speed
ratio. Consequently nRev had to be increased to be able to generate a near wake that
extended enough downstream. For U=5m/s, nRev was set equal to 5 while for U=7 and
10m/s,nRevwassetto4.WiththesevaluesfornRev,thenearwakeextendeddownstream
by3.6R,4R,6R,6R,7R,9.4R,12Ratwindspeeds5,7,10,13,15,20and25m/srespectively.
Atthesedistances,thevaluesforwatallradiallocationswere<1.5%.Also,thecontribution
285
Chapter6AerodynamicAnalysisoftheNRELPhaseVIWindTurbine
of the far wake to the axial induction at the lifting lines was found to be only <3% of the
total axial induction. Thus it may be assumed that the total induction is due to the near
wakealone.
SelectionofHAWT_FWCparametersvandSc
It should be appreciated that, considering the limitations of vortex models to cater for
viscous effects, it is very difficult to select the optimum values for (v,Sc). Due to this
difficulty,ananalysiswasthereforecarriedoutinthisstudytoinvestigatethesensitivityof
ua,ctodifferentvaluesof(v,Sc).Theanalysiswasperformedatawindspeedof13m/s.Four
differentsetsofvaluesof(v,Sc)wereused:(10,0.1),(100,1),(500,1)and(500,10).
Figs. 6.7(a) and (b) display the axial induced velocities at the blade lifting line and the
corresponding azimuthally averaged values computed using the four sets of values for
(v,Sc).Forthecaseshown,nwaskeptfixedat21.Atthemiddlebladesections,extending
formostoftheblades(from0.4Rto0.85R),thecalculatedinducedvelocitywasfoundtobe
almostinsensitivetochangesin(v,Sc).Highsensitivitywasnotedthoughinthebladeroot
andtipregions,thereasonbeingthehightrailingcirculationinthenearwakethatresults
fromrapidradialchangesinboundcirculation.Sinceitisverydifficulttoestablishrealistic
values of (v, Sc), the insensitivity of the freewake solution for the induced velocities for
mostofthebladespanisaverypositivefactor.
AsoutlinedinChapter5,page246,increasingboththevalueof vandScwillincreasethe
coreradiusofthemodelledtrailingandshedvorticesofthenearwake.Scdeterminesthe
initialcoresizeatthepointavortexisshedfromabladestrailingedge.vcontrolsnotonly
theinitialcoresizebutmoreovertherateofcoregrowthwithtime.Fig.6.8showsatypical
variationoftheviscouscoreradiuswiththevortexageindexaontheNRELrotor.Infact
thisfigureissimilartoFig.5.27.Recallthatthehigherthevalueofa,theloweristhevortex
age.Asitmaybeobserved,theolderwakevorticeshavealargercoreradius,asmodelled
bythefreewakesolution.Theoldestvorticessufferlargechangesintheircoreradius.This
is because they lie on the edge of the vortex sheet of the near wake where large vortex
filamentstrainispredicted.Thisabruptvariationisnotphysicalandisonlyduetotheway
thewakeismodelled.
Both vandSccontroltheamountofdeformationofthefreewakegeometry,especiallyat
the inboard and outboard edges of the vortex sheets were rollup is known to occur. The
highertheirvalues,thelessisthedeformation.Figs.6.9(a)and(b)illustratetwowakeplots
ofthenearwakecomputedusingthefreewakecodefortwodifferentsetsofvaluesfor(v,
Sc). The reduced wake rollup with larger values of these two arbitrary parameters is
evident.
286
Chapter6AerodynamicAnalysisoftheNRELPhaseVIWindTurbine
dv=10,sc=0.1
v = 10, Sc = 0.1 dv=100,sc=1
v = 100, Sc = 1 vdv= 500,sc=1
500, Sc = 1 dv
v = 500,sc=10
500, Sc =10
4
ua,c (m/s)
0
0.15 0.25 0.35 0.45 0.55 0.65 0.75 0.85 0.95 1.05
-2
-4
r/R
Figure6.7(a)Variationofinducedaxialvelocityatbladeliftinglineforn=21atdifferentvalues
of(v,Sc).
delta v 10
v = 10, Sc Sc0.1
= 0.1 delta v 100
v = 100, Sc Sc
= 11 vdelta v 500
= 500, Sc =Sc
11 delta
v = 500, ScSc
v 500 =10
10
1.5
0.5
0
ua (m/s)
-0.5
-1
-1.5
-2
r/R
Figure6.7(b)Variationoftheazimuthallyaveragedaxialinducedvelocityatbladeliftinglinefor
n=21atdifferentvaluesof(v,Sc).
287
Chapter6AerodynamicAnalysisoftheNRELPhaseVIWindTurbine
deltav,Sc
v = 10, Sc= =10,
0.10.1 v = 100, Sc = 1
deltav,Sc=100,1 deltav,Sc=500,1
v = 500, Sc = 1 v = 500, Sc =10
deltav,Sc=500,10
0.9
0.8
0.7
0.6
rceff (m)
0.5
0.4
0.3
0.2
0.1
0
0 10 20 30 40 50 60 70
a
Figure6.8Variationofviscouscoreradiusofnearwaketrailingvortexfilamentsatr/R=0.712.
Figure6.9Wakeplotscomputedwithn=31fordifferentvaluesof(v,Sc).
Duetothewakedeformation,thevortexfilamentsexperiencestrain.Thisstrainismostly
significantinthetipandrootregionsofthevortexsheets.Fig.6.10plotsthevariationofthe
strainwiththevortexageexperiencedbythetrailingvortexfilamentsoriginatingfromr/R
=0.9.ThestrainiscomputedusingEqt.E.3,AppendixE.Thevariationisplottedatdifferent
valuesof(v,Sc).Sincelargervaluesof(v,Sc)yieldlessrollup(asillustratedinFig.6.9),it
followsthattheyalsocausethefilamentstoexperiencelessstrain.
288
Chapter6AerodynamicAnalysisoftheNRELPhaseVIWindTurbine
deltav,Sc
v = 10, Sc= =10,
0.10.1 deltav,Sc=100,1
v = 100, Sc = 1 deltav,Sc=500,1
v = 500, Sc = 1 deltav,Sc=500,10
v = 500, Sc =10
0.15
0.1
0.05
-0.05
-0.1
5 15 25 35 45 55 65
a
Figure6.10Variationofstrainofnearwaketrailingvortexfilamentsatr/R=0.96.
Inattemptingtoselectrealisticvaluesfortheseviscousparameters(v,Sc)forthisstudy,it
was assumed that twice the value of rc at the blade trailing edge (i.e. at time t=0) should
approximatelybeequaltothethicknessoftheviscouslayerofthewakevortexsheets.This
thicknessdependsontheboundarylayersformedontheupperandlowersurfacesofthe
aerofoilwhichareinturnafunctionofthelocalReynoldsnumberandtheangleofattack.
The blade sections of the rotor at the operating conditions stated in table 6.1 have a local
Reynolds number in the range 0.6 1.4X106. For attached flow conditions for which the
anglesofattackaresmall,theboundarylayersarethin.ItwasestimatedusingRFOIL[63]
that,forthisReynoldsnumberrange,themaximumboundarylayerthicknessatthetrailing
edge is only 7.82X103c up to an angle of attack of 100. Consequently, the thickness of the
viscouslayerofthevortexsheetatthebladestrailingedgeinattachedflowconditionsis
alsosmallanditdoesnotexceedthelocalbladethickness(20.95c%)(seeFig.6.11).Athigh
angles of attack however, flow separation on the blades upper surface may cause the
thickness of the viscous layer to be larger than the local blade thickness. The maximum
angle of attack observed during the subject experiments was about 430 at U=25m/s and
r/R=0.3(refertoFig.6.13)).Atthisangleofattack,howeveritisunlikelythatthethickness
of the viscous layer exceeds the blade chord length. Detailed investigations of the blade
surface pressure distributions on NREL rotor [74, 75] revealed that as the angle of attack
increases and the flow separation point arrives at the blade leading edge, boundary layer
reattachment commences at the same time due to a high suction at the aft region of the
289
Chapter6AerodynamicAnalysisoftheNRELPhaseVIWindTurbine
upperbladesurface.Thisphenomenonisreferredtoshearlayerimpingement[75]andthe
resulting flow topology is illustrated in Fig. 6.11. Consequently the flow reattachment
causes the thickness of the viscous layer of the vortex sheet to become thin again at the
bladetrailingedge.
Thevaluesof2*rcatthetrailingedge(t=0)predictedbyEqt.5.8(Chapter5)forthedifferent
setsofvaluesof(v,Sc)werecomparedwiththebladethickness(whichisequalto20.95%c
for the S809 aerofoil) and the blade chord spanwise distributions. This comparison is
illustratedinFig.6.12.Itisshownthatparameters(500,10)yieldavalueof2*rcthatismuch
largerthanthemaximumbladechord.Thisisnotrealistic.Thus,alowervalueof2*rcwas
suggested for this study. Values (500, 1) tend to be suitable for separated flow conditions
sincetheresultingvalueof2*rcisbetween20.95%candc.Forattachedflowconditionsthe
values vand/orScshouldstrictlyspeakingbedecreasedsoastoreduce2*rctobelessthan
20.95%c.However,sincethisdoesnotaffectconsiderablythecalculateddistributionsofua,c
(refertoFig.6.7),(v,Sc)weresetto(500,1)forattachedflowconditionsalso.
sat
gedg hickness
ess
r )
c
kn
trailing ayerthicknes
e(=2
ness
c
flowseparationpoint thi
2rc )
boundarylayer
attr slayert
d e
ck
Bla
ethi
edge(=
ailin
Blad
ou
boundarylayer
sl
Visc
Viscou
Lowangleofattack(attachedflow) Highangleofattack(separatedflow)
flowreattachment
boundarylayer
Highangleofattack(shearlayerimpingement)
Figure6.11Influenceoftheangleofattackonthevariationoftheviscouslayerthicknessofthe
wakevortexsheetatthetrailingedge.
290
Chapter6AerodynamicAnalysisoftheNRELPhaseVIWindTurbine
1.4
1.2
1
2*rc, 20.95%c or c (m)
(v 100, Sc 1)
2*rc (dv100,Sc1)
0.6
(v 500,Sc 1)
2*rc (dv500,Sc1)
0.4
(v 500,Sc10)
2*rc (dv500,Sc10)
20.95%c
0.2
c
0
0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
r/R
Figure6.12Comparisonof2*rccalculatedatdifferentvaluesforviscousparameters(v,Sc)withthe
spanwisedistributionofbladethickness(20.95%c)andbladechord(c).
SelectionofHAWT_FWCparametersforthefarwakemodel
As regards the far wake parameters, the azimuthal step was always equal to (i.e. 100),
while the number of helical revolutions (nfwRev) was kept constant equal to 10 for all
calculations.Theprescribedhelicalpitch(pfw)wasvarieddependingonthewindspeed.To
determineareasonablevaluefortheprescribedhelicalpitch,preliminarycalculationswere
carried out with the same freewake model and an estimate was obtained from the near
wakegeometry.
291
Chapter6AerodynamicAnalysisoftheNRELPhaseVIWindTurbine
B.ResultsandDiscussion
Thissectionpresentstheresultsobtainedbyapplyingtheiterativeprocedure(Fig.6.6)for
findingtheangleofattackontheNRELrotorforU=5,7,10,13,15,20and25m/sat =00.
TheparametersgiveninsectionAabovewereusedinHAWT_FWC.
It was found that, for all the seven operating windspeeds, only two iterations were
necessary to obtain a converged angle of attack distribution at the blades with 5%
maximum error. The converged angle of attack distributions are shown in Fig. 6.13. The
corresponding bound circulation distributions are shown in Fig. 6.14. These distributions
are very similar to those derived by Tangler [90] using a prescribedwake vortex code. In
Fig.6.13,ariseintheangleofattackisnotedatthebladetip.Thisresultsfromtheviscous
effects being modelled in the rolling up vortex sheet of the freewake vortex code. At a
windspeedof15m/s,ariseintheboundcirculationisobservedinthemiddlesectionofthe
blades. This increases the trailing vorticity towards the middlesections of the blades.The
operating range of Reynolds numbers is around 0.6 1.4 X106 as shown in Fig. 6.15. In
checkingthecalculations,theflowrelativevelocityvariations,Vr,wereestimatedusingthe
inducedvelocitiesoutputtedbythefreewakemodelandthesewerecomparedwiththose
derivedexperimentallyfromQNORM(fromEqt.6.1).Goodagreementwasobtained,asmay
beseeninFig.6.16.
50
45
40
35
30
(deg)
25
20
15
10
0
0.25 0.35 0.45 0.55 0.65 0.75 0.85 0.95
r/R
Figure6.13Convergedangleofattackdistributionsatdifferentwindspeeds.
292
Chapter6AerodynamicAnalysisoftheNRELPhaseVIWindTurbine
20
18
16
14
12
B (m2/s)
10
0
0.15 0.25 0.35 0.45 0.55 0.65 0.75 0.85 0.95
r/R
Figure6.14Convergedboundcirculationdistributionsatdifferentwindspeeds.
1400000
1200000
1000000
Re
800000
600000
400000
0.25 0.35 0.45 0.55 0.65 0.75 0.85 0.95
r/R
Figure6.15VariationofReynoldsnumberatbladesfordifferentwindspeeds.
293
Chapter6AerodynamicAnalysisoftheNRELPhaseVIWindTurbine
45
40
35
Vr (m/s)
30
25
20
15
10
0.25 0.35 0.45 0.55 0.65 0.75 0.85 0.95
r/R
Figure 6.16 Comparison of Vr calculated by the freewake model with that derived from QNORM
(denotedbyexp).
WakePlots
The resulting near wake plots computed by the freewake code are illustrated for each
windspeedinFigs.6.17(a)(g).Colorcodingisusedtorepresentthecirculationdistribution
inthewake.Suchcomputationsnotonlygiveinsightonthewakegeometrydevelopedbut
also provide a pictorial representation of how the bound circulation formed around the
blades (see Fig. 6.14) is eventually diffused into the wake under the action of complex 3D
flows. Since all plots are for axial flow conditions, then the wake only consists of trailing
circulation.Intheseplots,highpositivetrailingcirculationisdisplayedinred,whilehigh
negativetrailingcirculationisdisplayedinblue.Atlowwindspeeds(U=5and7m/s),the
waketrailingcirculationtendstobeconcentratedatthebladetipandrootlocations.InFigs.
6.17(a) and (b) high positive trailing circulation in red is not always visible due to the
folding of the vortex sheets resulting from rollup. At higher wind speeds however
(U=10m/sandhigher),considerablelevelsoftrailingcirculationarereleasedforthemiddle
sectionsofthebladesintheformofhorseshoevortices.
294
Chapter6AerodynamicAnalysisoftheNRELPhaseVIWindTurbine
Figure6.17(a)FreewakeplotatU=5m/s.
Figure6.17(b)FreewakeplotatU=7m/s.
295
Chapter6AerodynamicAnalysisoftheNRELPhaseVIWindTurbine
Figure6.17(c)FreewakeplotatU=10m/s.
Figure6.17(d)FreewakeplotatU=13m/s.
296
Chapter6AerodynamicAnalysisoftheNRELPhaseVIWindTurbine
Figure6.17(e)FreewakeplotatU=15m/s.
Figure6.17(f)FreewakeplotatU=20m/s.
297
Chapter6AerodynamicAnalysisoftheNRELPhaseVIWindTurbine
Figure6.17(g)FreewakeplotatU=25m/s.
298
Chapter6AerodynamicAnalysisoftheNRELPhaseVIWindTurbine
A summary of the vortex sheet pitch variation derived from the above freewake
geometries is shown in Figs. 6.18. Fig. 6.18(a) shows the variation of the pitch against the
tunnelwindspeedwhileFig.6.18(b)isasimilarplotbutwiththenondimensionalvalues
being shown instead. It may be noted that the vortex pitch increases almost linearly with
windspeed. Wake expansion in the vicinity of the rotorplane was found to be small and
hardly to quantify accurately from the freewake plots. This small wake expansion is a
consequenceofthefactthattherotorwasoperatingatlowaxialthrustcoefficients.
25
Pitch of Wake Vortex Sheets (m)
20
15
10
0
5 7.5 10 12.5 15 17.5 20 22.5 25
U (m/s)
Figure 6.18(a) Vortex sheet pitch variation with tunnel windspeed as estimated from freewake
plotsat=00.
4.5
4
Non-dimensional pitch (p/Rt)
3.5
2.5
1.5
0.5
0
1 2 3 4 5 6 7 8
Figure6.18(b)Nondimensionalvortexsheetpitchvariation(p/Rt)withrotortipspeedratio()as
estimatedfromfreewakeplotsat=00.
299
Chapter6AerodynamicAnalysisoftheNRELPhaseVIWindTurbine
AerofoilCoefficients
The derived aerofoil data are summarized in Figs. 6.19(a) and (b). Both the lift and drag
coefficientsaredependentontheradiallocation.TheDelft2DaerofoildatafortheS809is
plottedforaReynoldsnumberof1X106.Atlowanglesofattack,theinboardbladesections
have a lower lift coefficient than the 2D values, but this increases continuously for high
angles of attack. The presence of stall delay is evident at the inboard blade sections.
However, the tip region has the lowest lift coefficients, considerably lower than the 2D
values.Thisisobservedevenforsmallanglesofattack.ThisreductioninClatthebladetip
resultsfromthe3Dfloweffectsinthisregionaswellasinductioneffectsfromthestrongtip
vortex. Apart from increasing Cl, stall delay also increases Cdp. This increase is also more
significant at the inboard sections. Lower values for Cdp are obtained for the tip region at
0.95R.
3DInductionatRotorplane
HAWT_FWC calculated the 3D induced velocity fields across the whole rotorplane for
different blade azimuth angles for the blade pressure measurements taken in the tunnel.
The contour plots displaying the induced velocity fields for U=5 and 15m/s are plotted in
Fig. 6.20. In these plots, the positions of the blades (lifting lines) is recognized from the
negatively large axial induced velocities (uy) (dark blue zone). From such plots it was
possibletoobtaintheaxialinductionfactordistributionsatthebladeliftinglinestogether
withthecorrespondingazimuthally(annular)averagedinductionfactors.Theseinduction
factordistributionsateachwindspeed(U=5,7,10,13,15,20and25m/s)areillustratedin
Figs. 6.21(a)(f). The axial, tangential and radial induction factors at the blade lifting lines
areshowninFigs.(a)(c),whileFigs.(d)(f)showthecorrespondingazimuthallyaveraged
values at each radial location. In general, higher windspeeds reduce the axial induction
factors at the rotorplane. This is a consequence of the fact that, for a fixed rotor angular
speed, higher windspeeds increase the pitch of the tip vortices and this reduces the axial
inductionexperiencedbytherotor.WhenlookingatFigs.(a)and(d),itcanbenotedthat,
for the blade root and tip regions, the axial induction factor at the blades is considerably
higherthantheazimuthallyaveragedvalue.Thisiscausedbythetipandrootvorticesthat,
at any instant, are closer to the blades than to the other regions in the rotorplane not
occupiedbytheblades.Thehigherinductioneventuallyreducesthebladeloadingatthetip
androotregions.
The tangential induction at each blade section increases with windspeed, reaching the
highest values at the blade root. However the tangential induction factor remains small
comparedtotheaxialinductionfactoratallwindspeeds.
300
Chapter6AerodynamicAnalysisoftheNRELPhaseVIWindTurbine
r/R = 0.3 r/R = 0.47 r/R = 0.63 r/R = 0.80 r/R = 0.95 2D Delft
2.5
1.5
Cl
0.5
0
0 5 10 15 20 25 30 35 40 45
(deg)
Figure6.19(a)Derivedliftcoefficientsfordifferentradialpositions.
r/R = 0.3 r/R = 0.47 r/R = 0.63 r/R = 0.80 r/R = 0.95 2D Delft
1.4
1.2
0.8
Cdp
0.6
0.4
0.2
0
0 5 10 15 20 25 30 35 40 45
(deg)
Figure6.19(b)Deriveddragcoefficientsfordifferentradialpositions.
301
Chapter6AerodynamicAnalysisoftheNRELPhaseVIWindTurbine
= 3450 = 45 0
Tangential induced
velocity field (ux)
= 2250
= 135 0
= 3450
= 45 0
Axial induced
velocity field (uy)
= 2250
= 135 0
= 3450
= 45 0
Radial induced
velocity field (uz)
= 2250
= 135 0
Figure6.20(a)3DinducedvelocityfieldatrotorplaneatU=5m/s, =00.Bladeisatanazimuth
angleof0/1800.Viewislookingupstreamfrombehindtherotorplane.
302
Chapter6AerodynamicAnalysisoftheNRELPhaseVIWindTurbine
= 3450
= 45 0
Tangential induced
velocity field (ux)
= 2250
= 135 0
= 3450 = 45 0
Axial induced
velocity field (uy)
= 2250
= 135 0
= 3450 = 45 0
Radial induced
velocity field (uz)
= 2250
= 135 0
Figure6.20(b)3DinducedvelocityfieldatrotorplaneatU=15m/s, =00.Bladeisatanazimuth
angleof0/1800.Viewislookingupstreamfrombehindtherotorplane.
303
Chapter6AerodynamicAnalysisoftheNRELPhaseVIWindTurbine
0.07
0.02
U = 5 m/s
-0.13
U = 7 m/s
U = 10 m/s
-0.18
U = 13 m/s
U = 15 m/s
-0.23
U = 20 m/s
U = 25 m/s
-0.28
r/R
Figure6.21(a)variationofaxialinductionfactoratbladeliftingline.
0.02
0
0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
-0.02
a2,c
U = 5 m/s
-0.04 U = 7 m/s
U = 10 m/s
U = 13 m/s
-0.06
U = 15 m/s
U = 20 m/s
U = 25 m/s
-0.08
r/R
Figure6.21(b)variationoftangentialinductionfactoratbladeliftingline.
0.1
0.08
0.06
a3,c
0.04
U = 5 m/s
U = 7 m/s
0.02
U = 10 m/s
U = 13 m/s
0 U = 15 m/s
0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 U = 20 m/s
U = 25 m/s
-0.02
r/R
Figure6.21(c)variationofradialinductionfactoratbladeliftingline.
304
Chapter6AerodynamicAnalysisoftheNRELPhaseVIWindTurbine
0.05
0
0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
-0.05
-0.1
a1
-0.15
U = 5 m/s
U = 7 m/s
-0.2
U = 10 m/s
-0.25 U = 13 m/s
U = 15 m/s
-0.3 U = 20 m/s
U = 25 m/s
-0.35
r/R
Figure6.21(d)variationofazimuthallyaveragedaxialinductionfactor.
0.000
0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
-0.005
-0.010
-0.015
-0.020
a2
U = 5 m/s
-0.025
U = 7 m/s
-0.030 U = 10 m/s
U = 13 m/s
-0.035
U = 15 m/s
-0.040 U = 20 m/s
U = 25 m/s
-0.045
r/R
Figure6.21(e)variationofazimuthallyaveragedtangentialinductionfactor.
0.22
0.17
0.12
a3
0.07
U= 5 m/s
U = 7 m/s
0.02 U = 10 m/s
U = 13 m/s
0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
-0.03 U = 15 m/s
U = 20 m/s
U = 25 m/s
-0.08
r/R
Figure6.21(f)variationofazimuthallyaveragedradialinductionfactor.
305
Chapter6AerodynamicAnalysisoftheNRELPhaseVIWindTurbine
WithreferencetoFig.(f),thefreewakecodeispredictinganazimuthallyaveragedradial
induction factor that is positive at the outboard sections, but negative at the inboard
sections for 5, 7 and 10m/s. A positive radial induction factor means that radial flow is
outwards towards the rotor tip radius and vice versa. This factor therefore gives an
indicationonthewaythewakeisexpandingattherotorplane.Itappearsthatforlowwind
speeds(5,7and10m/s),theflowattheouterradialpositionsisexpandingoutwardswhile
thatattheinnerradialpositionsisexpandinginwardsinthedirectionoftherotorhub.For
thehigherwindspeeds(13m/sandabove),thevortexcodeispredictinglowervaluesfora3
andwakeexpansionisoutwards.ItcanbenotedfromFig.(c)thatatthebladeliftinglines,
theradialinductionisingeneralmuchlowerinmagnitudethattheazimuthallyaveraged
value. A more detailed study would be necessary to investigate how the wakeinduced
radial flow patterns influence the aerodynamic behaviour of the blades at both low and
highanglesofattack.
Globalloads
The lift and drag coefficients and the induced velocities derived by the freewake vortex
model were used to calculate the lowspeed shaft torque and the blade flap/edgewise
bending moments using the bladeelement theory Eqts. 3.16. These results are shown in
Figs. 6.22 (a)(c) were they are noted by plots Freewake. As a crosscheck for these
calculations, the same loads were computed directly from the experimental data at each
rotor/bladetimestep byusingalinearinterpolationtoobtainadistributionforCn,Ctand
QNORMacrossthewholebladespanandusingthefollowingequations:
( )
b = B 1 Rtip
LSSTQ = QNORM b ,i
Ct
b ,i
* cos( i ) + Cn
b ,i
* sin( i ) ci ri dr (6.4)
b=0 Rhub
( )
Rtip
RFM = QNORM b ,i
Cn * cos( i tip ) Ct * sin( i tip ) ci ri rstrain gauge dr
b ,i b ,i
( )
Rhub
(6.5)
( )
Rtip
(6.6)
whererstraingaugeistheradiallocationofthestraingaugesmeasuringthebladerootflap/edge
moments. tipisincludedintheaboveequationsbecausethestraingaugesinstalledatone
oftherotorbladeswereorientedsuchthattheymeasuretheflappingmomentaboutanaxis
paralleltothechordlineofthebladetipsection[35].Inthelinearinterpolation,CnandCtat
thebladetip/rootweresetequaltozero.Thetrapeziumrulewasadoptedforintegrating
306
Chapter6AerodynamicAnalysisoftheNRELPhaseVIWindTurbine
1600
1400
1200
1000
LSSTQ (Nm)
800
600
400
Torque Free-Wake
200 Exp
Strain Gauges
0
4 6 8 10 12 14 16 18 20 22 24 26
U (m/s)
Figure6.22(a)Variationlowspeedshafttorquewithwindspeed.
5000
4500
Flap
4000 Moment
3500
RFM (Nm)
3000
2500
2000
1500
Free-Wake
1000 Exp
Strain Gauges
500
4 6 8 10 12 14 16 18 20 22 24 26
U (m/s)
Figure6.22(b)Variationofbladerootflapmomentwithwindspeed.
307
Chapter6AerodynamicAnalysisoftheNRELPhaseVIWindTurbine
1400
Edge
1200
Moment
1000
800
600
REM (Nm)
400
200
0
4 6 8 10 12 14 16 18 20 22 24 26
-200
-400 Free-Wake
Exp
-600
Strain Gauges
-800
U (m/s)
Figure6.22(c)Variationofbladerootedgemomentwithwindspeed.
numerically across the whole blade span. These results are included in Figs. 6.22 and are
referredtoExp.TheloadsmeasureddirectlyusingstraingaugesarealsoincludedinFigs.
6.22. These are plotted together with the corresponding +/one standard deviation. The
largestandarddeviationsintheREMwereduetothecyclicgravitationalloads.Theresults
calculated by the freewake model and those derived directly from the pressure
measurements(Exp)agreeverywell.HoweverfortheLSSTQandREM,agreementwith
the loads measured by the strain gauges is not as good especially at higher wind speeds.
Although one cannot ignore the fact that structural dynamic effects of the rotating rotor
couldhaveinfluencedtheloadsmeasuredbythestraingauges,amoreprobablesourceof
errorwasinmeasuringCt.WheninvestigatingthevariationofthemeasuredCtvalueswith
bladeazimuthangleoverawholerevolution,aconsiderablylargestandarddeviationwas
found.Anotherpossiblesourceoferrorwasduetothefactthatthepressuremeasurements
were only performed at five radial locations. Thus the linear extrapolated distributions of
theexperimentaldatausedinthefreewakemodelcouldyieldnumericalerrors.
308
Chapter6AerodynamicAnalysisoftheNRELPhaseVIWindTurbine
ComparisonofInflowAngle
The local flow angles (LFA) predicted by HAWT_FWC are now compared with those
measured by the flow direction probes during the wind tunnel experiments (see Fig. 6.4).
ThecomparisonisshowninFigs.6.23(a)(g).Theangleofattackisalsoincluded.Ideallythe
HAWT_FWCpredictionsfortheLFAshouldbeequaltothosemeasuredbytheprobes.But
this is not always the case, as it may be seen in Figs. 6.23. One source of error could be
because the blades were modelled as rigid in HAWT_FWC and during the experiments
blade deflection effects could influence the LFA measurements. However calculations
performed by the ECN on aeroelastic code PHATAS confirm that the deflection of the
blades was small, even at the high windspeeds (maximum blade tip deflection<10cm),
(ECN, Gerard Schepers, personal communication). To assess the uncertainty due to
arbitrary choice of viscous parameters, the LFA distribution at U =25m/s was computed
using the freewake vortex model at two extreme sets of values of (v, Sc): at (10, 0.1) and
(500,1).ThemaximumdeviationintheLFAduetothedifferent(v,Sc)valueswasfoundto
beonly0.340atr/R=0.91.
At all windspeeds, the freewake vortex model predicts an LFA that is larger than the
derived angle of attack. This is due to the presence of upwash created by the bound
circulationattheblade.ThisisconsistentwiththemeasuredvaluesofLFAsincetheseare
alsolargerthantheangleofattackbeingpredictedbythevortexmodel.Thecorrelationof
the LFA is very good at low windspeeds (U=5 and 7m/s) at which the angle of attack is
smallacrossthewholebladespan.Athigherwindspeeds,theagreementdoesnotremain
good, especially at the inboard sections. At U=25m/s, the largest discrepancy between the
predictedandmeasuredLFAisabout90whichisconsiderablyhigh.
ItwasfoundthatHAWT_FWCisaccurateinpredictingtheLFAwhenattheoutboardblade
sections(r/R>0.67),evenatthehighwindspeedswherethelocalangleofattackexceedsthe
2Dstallingangle(around100).Butattheinboardbladesections,thecapabilityofthecode
in predicting the LFA accurately degrades steadily at larger windspeeds. The problem is
mostlikelytobeassociatedwiththewaythevortexmodelmodelscirculationaroundthe
blades.InHAWT_FWC,thebladesaremodelledasliftinglinesandtheboundcirculationis
forcedtoleavefromthetrailingedgeasasinglevortexsheet(seeFig.6.24(a)).Theliftand
bound circulation are related by the KuttaJoukowski law (Eqt. 6.2). Such a model works
well where the flow over the blades is fully attached. Yet it is known that this model has
limitedaccuracywhenitcomestomodellinghighanglesofattackwheretheflowoverthe
upperbladesurfaceexperiencesseparation.Insuchconditions,notallboundcirculationis
released into the wake through the trailing edge. More accurate vortex models exist, as
describedin[42,pp505516].Suchmodelswouldincludemorethanjustoneboundvortex,
withalltheboundvorticesdistributedaroundthebladessurface(refertoFig.6.24(b)).In
addition, to account for flow separation, a secondary vortex sheet in the near wake is
includedemergingfromtheupperbladesurface.Asalreadydescribedinsection6.3.2,
309
Chapter6AerodynamicAnalysisoftheNRELPhaseVIWindTurbine
20
18
16
Free-wake, angle of attack
Free-wake, LFA
14
or LFA (deg)
Exp, LFA
Fig.(a): U = 5m/s 12
10
8
6
4
0
0.25 0.35 0.45 0.55 0.65 0.75 0.85 0.95
r/R
20
18 Free-wake, angle of attack
16 Free-wake, LFA
Exp, LFA
or LFA (deg)
14
Fig.(b): U = 7m/s 12
10
8
6
0
0.25 0.35 0.45 0.55 0.65 0.75 0.85 0.95
r/R
30
Exp, LFA
or LFA (deg)
20
Fig.(c): U = 10m/s
15
10
0
0.25 0.35 0.45 0.55 0.65 0.75 0.85 0.95
r/R
40
Fig.(d): U = 13m/s 25
20
15
10
0
0.25 0.35 0.45 0.55 0.65 0.75 0.85 0.95
r/R
Figure6.23ComparisonofLFAdistributionpredictedbyHAWT_FWCwiththosemeasuredby
fiveholeprobe.AngleofattackdistributioncomputedbyHAWT_FWCisalsoshown.
310
Chapter6AerodynamicAnalysisoftheNRELPhaseVIWindTurbine
50
45
Free-wake, angle of attack
40 Free-wake, LFA
35 Exp, LFA
or LFA (deg)
Fig.(e): U = 15m/s 30
25
20
15
10
5
0
0.25 0.35 0.45 0.55 0.65 0.75 0.85 0.95
r/R
60
55
50
45
or LFA (deg)
Fig.(f): U = 20m/s 40
35
30
25
20
15
Free-wake, angle of attack
Free-wake, LFA
10
Exp, LFA
5
0
0.25 0.35 0.45 0.55 0.65 0.75 0.85 0.95
r/R
60
55
50
Fig.(g): U = 25m/s 45
or LFA (deg)
40
35
30
25
20 Free-wake, angle of attack
15 Free-wake, LFA
10 Exp, LFA
5
0
0.25 0.35 0.45 0.55 0.65 0.75 0.85 0.95
r/R
Figure6.23Contd.frompreviouspage.
variousstudiesonthesubjectNRELPhaseVIexperiments[74,75]howeverconcludedthat
at sufficiently high angles of attack, the flow topology over the inboard sections of the
blades is that of an impinging shear layer (see Fig. 6.11). The vortex lifting line model
embedded in HAWT_FWC is also limited for modelling this situation. A more accurate
vortex model could be with a second wake vortex sheet that emerges from the blades
leadingedgeandreattachingitselfwiththebladesupperchamberatadefinedposition,as
depictedinFig.6.24(c).
311
Chapter6AerodynamicAnalysisoftheNRELPhaseVIWindTurbine
Fig.(a)
Fig.(b) Fig.(c)
Figure6.24Threedifferentvortexmodelsusedtomodelflowoverblades.Fig.(a)showsthevortex
liftinglinemodelembeddedinHAWT_FWC;Fig.(b)isamoreaccuratevortexmodelforhighangles
of attack where separation only occurs. Fig. (c) is a more accurate vortex model for high angles of
attackwhereflowseparationattheleadingedgeisfollowedbyashearlayerimpingement.
ComparisonwithResultsfromEllipSys3D
The NREL Phase VI rotor has been modelled by Johansen et al. [40] using the CFD code
EllipSys3D which is used to model wind turbine rotors. The code was developed by
Michelsen et al. [57, 58] and Srensen et al. [87]. In the code, a multiblock finite volume
discretizationoftheincompressibleReynoldsAvergedNavierStokesequationsisadopted.
The code uses a collocated variable arrangement, and Rhei/Chow interpolation is used to
avoid odd/even pressure decoupling. The SIMPLE algorithm is used to enforce the
pressure/velocity coupling. The momentum equations are solved using a second order
upwindinterpolationscheme(SUDS)fortheconvectiveterms.Theturbulenceismodelled
usingtheDetachedEddySimulation(DES)modelaccordingtoStretlets[89]).
Johansenetal.[40]usedtheresultsfromEllipSys3DfortheNRELrotortoderive3Daerofoil
coefficients (Cl and Cdp) for BEM models at different wind speeds (U) for =00. The
azimuthally(annular)averagedaxialflowvelocityasafunctionofaxialdistancefromthe
rotorplane was obtained from the CFD wake computations. The axial velocity at the
rotorplanecouldthenbeobtained,togetherwiththeazimuthallyaveragedaxialinduction
factor (a1). The tangential induction factor was taken as zero since it was found to be
negligiblysmall.a1wasthenusedtoevaluatetheangleofattack().UsingCnandCtwith
thederived ,theaerofoilcoefficientsforthedifferentradiallocationswerederivedusing
312
Chapter6AerodynamicAnalysisoftheNRELPhaseVIWindTurbine
theequationsinFig.2.1.Theangleofattackwasdeterminedusingasimilarmethodasfor
HAWT_FWCwiththeonlydifferencethatitwasbasedontheazimuthallyaveragedaxial
inductionfactor(a1)insteadofthatattheliftinglineoftheblades(a1,c).
TheresultsfromHAWT_FWCarenowcomparedwiththoseobtainedusingEllipSys3Dfor
=00 and wind speeds U=7, 10, 13, 15, 20 and 25m/s. It is important to recall that
HAWT_FWCmakes use of the experimental data (Cn, Ct and QNORM) tobe able to find the
angleofattackandinductionfactordistributions.Ontheotherhand,EllipSys3Distotally
independentanddoesnotrequiresuchexperimentaldataasinput.Theanalysisstartedby
comparing the spanwise distributions of Cn and Ct. The comparison is illustrated in Figs.
6.25and6.26.Intheseplots,experimentaldatafromNRELisshown.Themeanvaluesfor
CnandCtareplotted(meanoveronewholebladerevolution)togetherwiththeerrorsbars
denotingtheonestandarddeviations.ThedistributionsforCnandCt(extrapolatedusing
linearinterpolationasexplainedalreadyinsection6.3.1)thatwereinputtedtoHAWT_FWC
arealsoshown.Ingeneral,thepredictionsforCnandCtbyEllipSys3Dagreeverywellwith
themeasurementsandthisdemonstratesthecapabilityofCFDmethodsincapturingstall
delay phenomena at the inboard blade sections. It is noted in Figs. 6.25 and 6.26 that the
agreement is best atU=7m/s,for which the flow is known to be attached. At higher wind
speeds however, Cn predictions by EllipSys3D tend to be slightly overpredicted at the
inboard sections. This is attributed to the flow turbulence model embedded in the CFD
code.At10m/s,adipinbothCnandCtispredictedbetweenr/R=0.5and0.7byEllipSys3D.
However in the measured data, this dip is not observed in Cn. This may be due to the
deficiencies of the CFD models in modelling accurately laminartoturbulent transition of
theflowontheupperbladesurface.
Theinductionfactorsa1anda1,cobtainedbyHAWT_FWC(plottedinFigs.6.21(a)and(d))
werecomparedwitha1derivedbyEllipSys3D.ThecomparisonispresentedinFig.6.27.At
thelowwindspeedofU=7m/s(smallanglesofattack),theagreementina1predictedbythe
two different codes is very good (see Fig. 6.27(a)). But discrepancies in a1 are observed at
higher windspeeds. Since such discrepancies are only found at high windspeeds which
causelargeanglesofattackattheblades,itisexpectedthatthediscrepanciesareduetothe
deficiencies of the vortex model embedded in HAWT_FWC. As already described in the
previoussection,avortexmodelinwhichthebladesaremodelledasliftinglineswithonly
asinglevortexsheetemergingfromthebladestrailingedgemaybeinsufficienttomodel
theflowaroundthebladesathighanglesofattack,especiallywhentheflowismassively
separated (see Fig. 6.24). Recall also that the largest discrepancies in the LFA comparison
(refer to previous section) were also noted at the higher windspeeds. On the other hand,
CFDismorephysicallycomprehensiveinmodellingtheflowlocallyaroundthebladesat
highanglesofattack.
SincetheangleofattackinHAWT_FWCisbasedona1,cwhilethatinEllipSys3Disbasedon
a1, then the differences between these induction factors (see Fig. 6.27) result in different
313
Chapter6AerodynamicAnalysisoftheNRELPhaseVIWindTurbine
spanwisedistributionsfortheangleofattack.ThelatterdistributionsareshowninFig.6.28.
Becausetheresultsfora1,cfromHAWT_FWCanda1inEllipSys3Ddifferconsiderablyfrom
eachother(especiallyathighwindspeeds),itisexpectedthattheangleofattackdistribution
would be very much different. On the other hand, these differences are negligible for
r/R>0.5atallwindspeeds.Thisisbecausetheinducedvelocitiesareverysmallcomparedto
Uandr.
Fig. 6.29 compare the Cl and Cd curves for the different spanwise locations being
predicted by HAWT_FWC and EllipSys3D. In general, the relations were found to be
reasonablyclose.Itshouldbekeptinmindthatthetwocodesuseadifferentdefinitionfor
the angle of attack. But since for r/R>0.5 the angles of attack from both methods are very
close (see Fig. 6.28), then the Cl and Cd values can be compared. This is not the case for
spanwiselocationsr/R=0.3and0.47sincedifferencesinresultindifferencesinClandCd.
314
Chapter6AerodynamicAnalysisoftheNRELPhaseVIWindTurbine
1 3
0.9
2.5 NREL Exp
0.8 HAWT_FWC
0.7 EllipSys3D
2
0.6
Cn
Cn
1.5
0.5 NREL Exp
0.4 HAWT_FWC
0.3 EllipSys3D 1
0.2 0.5
0.1
0 0
0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
r/R r/R
1.6 3
1.5
0.6 1
0.4 NREL Exp
HAWT_FWC 0.5
0.2 EllipSys3D
0 0
0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
r/R r/R
3
2.5
NREL Exp
2 NREL Exp
2.5
HAWT_FWC
EllipSys3D
HAWT_FWC
2
EllipSys3D
1.5
Cn
Cn
1.5
1
1
0.5
0.5
0 0
0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
r/R r/R
Figure6.25Comparisonofthespanwisedistributionsofthenormalcoefficientalongthebladesat
=00.
315
Chapter6AerodynamicAnalysisoftheNRELPhaseVIWindTurbine
0.25
0.16 NREL Exp
NREL Exp 0.2 HAWT_FWC
0.14
HAWT_FWC EllipSys3D
0.12 EllipSys3D 0.15
0.1
0.1
0.05
Ct
0.08
Ct
0
0.06
-0.05
0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
0.04
-0.1
0.02
-0.15
0
0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 -0.2
r/R r/R
Fig. (d): U = 15m/s
Fig. (a): U = 7m/s
0.3 0.25
NREL Exp
0.25 NREL Exp
0.2 HAWT_FWC
HAWT_FWC
0.2
EllipSys3D 0.15 EllipSys3D
0.15 0.1
0.05
0.1
Ct
Ct
0
0.05 -0.05
0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
0
-0.1
0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
-0.05 -0.15
-0.1 -0.2
-0.15 -0.25
r/R r/R
Fig. (b): U = 10m/s Fig. (e): U = 20m/s
0.6 0.25
0.5 NREL Exp 0.2 NREL Exp
HAWT_FWC HAWT_FWC
0.4 EllipSys3D
0.15
EllipSys3D
0.1
0.3
0.05
0.2
Ct
Ct
0
0.1
-0.05
0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
0
-0.1
0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
-0.1 -0.15
-0.2 -0.2
-0.3 -0.25
r/R r/R
Figure6.26Comparisonofthespanwisedistributionsofthetangentialcoefficientalongtheblades
at=00.
316
Chapter6AerodynamicAnalysisoftheNRELPhaseVIWindTurbine
0.1 0.15
HAWT_FWC, a1 HAWT_FWC, a1
0.05 HAWT_FWC, a1c 0.1 HAWT_FWC, a1c
EllipSys3D, a1 0.05 EllipSys3D, a1
0
0
0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
a1 and a1,c
a1 and a1,c
-0.1
-0.1
-0.15
-0.15
-0.2
-0.2 -0.25
-0.25 -0.3
r/R
r/R
0.1 0.1
HAWT_FWC, a1 HAWT_FWC, a1
0.05 HAWT_FWC, a1c 0.05 HAWT_FWC, a1c
EllipSys3D, a1 EllipSys3D, a1
0
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
a1 and a1,c
a1 and a1,c
-0.1
-0.15
-0.15 -0.2
-0.2 -0.25
r/R r/R
0.1 0.1
HAWT_FWC, a1
0.05 HAWT_FWC, a1c
0.05
EllipSys3D, a1
0
0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 0
a1 and a1,c
a1 and a1,c
-0.25 -0.2
r/R r/R
Figure6.27Comparisonofthespanwisedistributionsoftheinductionfactorsalongthebladesat
=00.
317
Chapter6AerodynamicAnalysisoftheNRELPhaseVIWindTurbine
40 60
35
HAWT_FWC 50 HAWT_FWC
30 EllipSys3D EllipSys3D
40
25
(deg)
(deg)
20 30
15
20
10
10
5
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 0
0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
r/R r/R
50 70
45
HAWT_FWC
60
HAWT_FWC
40
EllipSys3D
50
EllipSys3D
35
30
(deg)
(deg)
40
25
30
20
15 20
10
10
5
0 0
0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
r/R r/R
60
70
50 60
HAWT_FWC HAWT_FWC
EllipSys3D EllipSys3D
50
40
(deg)
(deg)
40
30
30
20
20
10 10
0 0
0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
r/R r/R
Fig. (c): U = 13m/s Fig. (f): U = 25m/s
Figure 6.28 Comparison of the spanwise distributions of the angle of attack along the blades at
=00.
318
Chapter6AerodynamicAnalysisoftheNRELPhaseVIWindTurbine
2.4 2
2.2 1.8
2
1.8
1.6
1.4
r/R=0.3 1.4 1.6
r/R=0.3 1.2
Cd
Cl
1.2 1
1 0.8
0.8
HAWT_FWC 0.6 HAWT_FWC
0.6
EllipSys3D
0.4
EllipSys3D
2D Delft
0.4
2D Delft
0.2 0.2
0 0
0 5 10 15 20 25 30 35 40 45 50 0 5 10 15 20 25 30 35 40 45 50
(deg) (deg)
1.6 1.6
1.4 1.4
1.2 1.2
r/R=0.47 1 r/R=0.47 1
Cd
Cl
0.8 0.8
0.6 0.6
0.4
HAWT_FWC
0.4
EllipSys3D HAWT_FWC
0
0 5 10 15 20 25 30 35 40 45 50
0
0 5 10 15 20 25 30 35 40 45 50
(deg) (deg)
1.4 1.2
1.2
1
1
0.8
r/R=0.63 0.8 r/R=0.63
Cd
Cl
0.6
0.6
0.4
HAWT_FWC 0.4
EllipSys3D HAWT_FWC
0.2
2D Delft 0.2 EllipSys3D
2D Delft
0 0
0 5 10 15 20 25 30 35 40 45 0 5 10 15 20 25 30 35 40 45
(deg) (deg)
1.2 1
0.9
1
0.8 HAWT_FWC
EllipSys3D
0.8
0.7
2D Delft
r/R=0.8 r/R=0.8 0.6
Cd
Cl
0.6 0.5
0.4
0.4
HAWT_FWC 0.3
0.2
EllipSys3D
0.2
2D Delft
0.1
0 0
0 5 10 15 20
(deg)
25 30 35 40 0 5 10 15 20
(deg)
25 30 35 40
1.2 0.5
1
HAWT_FWC
EllipSys3D
0.45
0.4 HAWT_FWC
0.8
2D Delft
0.35
EllipSys3D
2D Delft
0.6 0.25
0.2
0.4
0.15
0.1
0.2
0.05
0 0
0 5 10 15 20 25 30 35 0 5 10 15 20 25 30 35
(deg) (deg)
Figure6.29Comparisonofthederivedaerofoildataat=00.
319
Chapter6AerodynamicAnalysisoftheNRELPhaseVIWindTurbine
6.3.3YawedConditions
Thissectiondescribesthecomputationalworkcarriedoutforyawedconditions(=300)for
windspeedsU=5,7,10,13and15m/s.
A.SelectingFreeWakeModelParameters
SelectionofHAWT_FWCparametersn,andnRev
Inallthefreewakecomputationsat=300,nandwerechosentobeequalto21and100,
respectively. These values were found to provide sufficient numerical accuracy, even in
yawedconditions.ThefollowingvaluesfornRevwerefoundtobeadequate:ForU=5m/s,
nRevwassetto5.ForU=7m/sand10m/s,nRevwassetto4whileforU=13and15m/s,nRev
wassetto3.Withthesevalues,thenearwakeextendeddownstream(alongtheYaxis)by
3.6R,4.4R,6R,6Rand7RatU=5,7,10,13,15m/srespectively.Theresultingvaluesforthe
percentagediscrepancyforwakeperiodicity(wp,computedusingEqt.5.15,Chapter5)at
all radiallocations were <1.5% while the far wake was enough downstream such that its
contributiontotheinductionattherotorplanewaskeptbelow3%ofthetotalinduction.
SelectionofHAWT_FWCparametersvandSc
As discussed already in section 6.3.2, Part A, it is very difficult to select the optimum
valuesfor(v,Sc)usedinthevortexmodelstocaterforviscouseffects.Thisisevenmore
difficult when modelling yawed conditions in which these effects become subject to
dynamic flow influences. Due to this difficulty, an analysis was therefore carried out to
investigate the sensitivity of ua,c to different values of (v,Sc) using a similar approach as
thatperformedinaxialconditions.Theanalysiswasperformedattwowindspeeds(7and
13m/s) which yielded attached flow and separated flow conditions at the blades,
respectively.Fourdifferentsetsofvaluesof(v,Sc)wereused:(10,0.1),(100,1),(500,1)and
(500,10). Figs. 6.30 and 6.31 illustrate the resulting spanwise variations at different blade
azimuthpositions(00,1200and2400).Itisobservedthatthesensitivityofua,catthefirst3
setsofvaluesof(v,Sc)isverysmall.Butanotablerelativediscrepancyisseenfor(500,10).
However, as explained in section 6.3.2, Part A. (see Fig. 6.12), the values (500,10) are too
largetoberealisticandinallthefreewakecomputationsat =300,(v,Sc)werealsotaken
tobeequalto(500,1),asforaxialconditions.
320
Chapter6AerodynamicAnalysisoftheNRELPhaseVIWindTurbine
v 10,Sc0.1
dv10, Sc 0.1 dv100, Sc 1
v 100, Sc 1 500,
dv500
v
, Sc1
S c 1 dv500,
v 500,S c 10
Sc10
3.5
2.5
2
ua,c (m/s)
1.5
0.5
0
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
-0.5
-1
-1.5
r/R
Fig(a): = 0 deg
dv10, Sc0.1
v 10, Sc 0.1 100,
dv100,
v
Sc1S
c1 dv500, Sc1
v 500, Sc 1 dv500, Sc10
v 500, Sc 10
2.5
1.5
ua,c (m/s)
0.5
0
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
-0.5
-1
-1.5
-2
r/R
Fig(b): = 120 deg
dv10,
v 10, Sc 0.1
Sc0.1 v 100,
dv100, Sc1Sc 1 v 500,
dv500, Sc1Sc 1 dv500,
v 500, Sc 10
Sc10
4
3.5
3
2.5
2
ua,c (m/s)
1.5
1
0.5
0
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
-0.5
-1
-1.5
r/R
Fig(c): = 240 deg
Figure6.30Effectofviscousparameters(v,Sc)onthespanwisedistributionofua,catdifferent
bladeazimuthangles,U=7m/s.
321
Chapter6AerodynamicAnalysisoftheNRELPhaseVIWindTurbine
10,
dv10,
v
Sc0.1
S 0.1
c dv100, Sc1
v 100, Sc 1 500,
dv500,
v
Sc1S
c dv500, Sc10
v 500, Sc 10
3
ua,c (m/s)
2
0
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
-1
-2
-3
r/R
Fig(a): = 0 deg
10,
dv10,
v
Sc0.1
S 0.1
c 100,
dv100,
v
Sc1
S c 1 500,
dv500,
v
Sc1S
c 1 dv500, Sc10
v 500, Sc 10
3
ua,c (m/s)
0
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
-1
-2
-3
r/R
3
ua,c (m/s)
0
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
-1
-2
-3
r/R
Fig(c): = 240 deg
Figure6.31Effectofviscousparameters(v,Sc)onthespanwisedistributionofua,catdifferent
bladeazimuthangles,U=13m/s.
322
Chapter6AerodynamicAnalysisoftheNRELPhaseVIWindTurbine
SelectionofHAWT_FWCparametersforthefarwakemodel
Forthefarwakeparameters,theazimuthalstepwasalwaysequalto (i.e.100),whilethe
number of helical revolutions (nfwRev) was kept constant equal to 10 for all calculations.
Theprescribedhelicalpitch(pfw)andwakeskewangle(s)werevarieddependingonthe
windspeed.Todeterminereasonablevaluesfortheprescribedhelicalpitchandwakeskew
angle, preliminary calculations were carried out with the same freewake model and an
estimatewastakenfromthenearwakegeometry.
B.ResultsandDiscussion
Thissectiondescribestheresultsobtainedfromaconvergedsolutionintheangleofattack
fromtheNRELbladepressuremeasurementsatwindspeeds5,7,10,13and15m/sandata
yawangleof300usingtheprocedureofFig.6.6.TheparametersgiveninsectionA.above
wereusedinHAWT_FWC.
Variationof, ,ClandCdpwithbladeazimuthangle()
Figs.6.32(a)(e)illustratethevariationsoftheconvergedangleofattack,therateofchange
oftheangleofattackandtheliftanddragcoefficientswiththebladeazimuthangleateach
of the different tunnel wind speeds. For all windspeeds, the angle of attack varies
periodicallywiththebladeazimuthangle.AtU=5m/s,agradualphaseshiftisobservedin
the angle of attack when moving outboard from r/R=0.3 to 0.95. At r/R=0.3, the angle of
attack is minimum at 2100 while at r/R=0.95 the minimum occurs at 1200. For the higher
windspeedshowever(U=7,10,13and15m/s),thisphaseshiftisnotnotedandtheangleof
attack is minimum when the blade is approximately at the 1800 position and reaches a
maximumwhenthebladeisatthe0/3600position.Foraparticularwindspeed,boththe
meanandcycliccomponentsoftheangleofattackarelargerattheinboardbladesections,
implying that the inboard sections experience higher unsteady effects and are first to
experiencestall.Increasingthewindspeedincreasesboththemeanandcycliccomponents
oftheangleofattack.
InFigs.6.32(a)(e)itmayalsobenotedthatthetimerateofchangeofangleofattack( )
variesalsoperiodicallywithbladeazimuthangleatallwindspeeds.Therateofchangeof
angleofattackisnegativeduringapproximatelybladeangles00to1800 whileitispositive
fromabout1800to3600.Theinboardbladesectionstendtoexperiencethehighestratesof
change in the angle of attack, occurring at approximately blade angles 900 and 2700.
Increasing the windspeed while keeping the rotor shaft speed constant results in higher
valuesfor .AtU=5m/s, atr/R=0.3isequaltoabout+/25deg/swhileatU=15m/s,this
reachesmaximumvaluesofabout+/200deg/s.
323
Chapter6AerodynamicAnalysisoftheNRELPhaseVIWindTurbine
With regards to the variation of the lift coefficient with blade azimuth angle, it may be
observed that, at each windspeed, the inboard regions of the blades experience higher
values in both the mean and cyclic components of Cl. At low windspeeds (U=5, 7 and
10m/s),thevariationofCl with isperiodicandquiteregular,reachingamaximumvalue
at blade position 0/3600 and a minimum value at about 1800. At the higher wind speeds,
(U=13and15m/s),thevariationisnolongerregular.
Thepressuredragcoefficients(Cdp)areverysmallatlowwindspeeds(U=5and7m/s).But
as the windspeed is increased, the pressure drag coefficient values increase rapidly, the
highestvaluesbeingattheinboardbladesectionsatbladeangles01200and2403600.
324
Chapter6AerodynamicAnalysisoftheNRELPhaseVIWindTurbine
5 r/R = 0.3 r/R = 0.47 r/R = 0.63 r/R = 0.8 r/R = 0.95 14
r/R = 0.3 r/R = 0.47 r/R = 0.63 r/R = 0.8 r/R = 0.95
4 12
3 10
(deg)
8
(deg)
2
6
1
4
0
0 60 120 180 240 300 360
2
-1
0
-2 0 60 120 180 240 300 360
(deg) (deg)
r/R = 0.3 r/R = 0.47 r/R = 0.63 r/R = 0.8 r/R = 0.95 r/R = 0.3 r/R = 0.47 r/R = 0.63 r/R = 0.8 r/R = 0.95
24 44
34
14 24
14
4
(deg/s)
(deg/s)
4
-6 0 60 120 180 240 300 360 -6 0 60 120 180 240 300 360
-16
-26
-16
-36
-26 -46
(deg) (deg)
r/R = 0.3 r/R = 0.47 r/R = 0.63 r/R = 0.8 r/R = 0.95 r/R = 0.3 r/R = 0.47 r/R = 0.63 r/R = 0.8 r/R = 0.95
0.7 1.2
0.6
1
0.5
0.8
0.4
Cl
0.6
Cl
0.3
0.4
0.2
0.1
0.2
0
-0.1 0 60 120 180 240 300 360 0
0 60 120 180 240 300 360
(deg) (deg)
r/R = 0.3 r/R = 0.47 r/R = 0.63 r/R = 0.8 r/R = 0.95 r/R = 0.3 r/R = 0.47 r/R = 0.63 r/R = 0.8 r/R = 0.95
0.02 0.08
0.018
0.07
0.016
0.06
0.014 0.05
0.012
0.01
Cdp
Cdp
0.04
0.008 0.03
0.006
0.02
0.004
0.002 0.01
0 0
0 60 120 180 240 300 360
0 60 120 180 240 300 360
(deg) (deg)
325
Chapter6AerodynamicAnalysisoftheNRELPhaseVIWindTurbine
25 r/R = 0.3 r/R = 0.47 r/R = 0.63 r/R = 0.8 r/R = 0.95 45
r/R = 0.3 r/R = 0.47 r/R = 0.63 r/R = 0.8 r/R = 0.95
40
20 35
30
15
(deg)
(deg)
25
20
10 15
5 10
0 0
0 60 120 180 240 300 360 0 60 120 180 240 300 360
(deg) (deg)
80 r/R = 0.3 r/R = 0.47 r/R = 0.63 r/R = 0.8 r/R = 0.95 r/R = 0.3 r/R = 0.47 r/R = 0.63 r/R = 0.8 r/R = 0.95
150
60
100
40
50
20
0 (deg/s)
0
(deg/s)
-40 -100
-60 -150
-80 -200
(deg) (deg)
r/R = 0.3 r/R = 0.47 r/R = 0.63 r/R = 0.8 r/R = 0.95
2.8
r/R = 0.3 r/R = 0.47 r/R = 0.63 r/R = 0.8 r/R = 0.95
2.4
2.4
2
2
1.6
1.6
1.2
Cl
Cl
1.2
0.8
0.8
0.4 0.4
0 0
0 60 120 180 240 300 360 0 60 120 180 240 300 360
(deg) (deg)
r/R = 0.3 r/R = 0.47 r/R = 0.63 r/R = 0.8 r/R = 0.95 r/R = 0.3 r/R = 0.47 r/R = 0.63 r/R = 0.8 r/R = 0.95
1 1.6
0.9
1.4
0.8 1.2
0.7
0.6 1
0.5
Cdp
Cdp
0.8
0.4 0.6
0.3
0.2 0.4
0.1 0.2
0 0
0 60 120 180 240 300 360 0 60 120 180 240 300 360
(deg) (deg)
326
Chapter6AerodynamicAnalysisoftheNRELPhaseVIWindTurbine
50 r/R = 0.3 r/R = 0.47 r/R = 0.63 r/R = 0.8 r/R = 0.95
45
40
35
30
(deg)
25
20
15
10
5
0
0 60 120 180 240 300 360
(deg)
250 r/R = 0.3 r/R = 0.47 r/R = 0.63 r/R = 0.8 r/R = 0.95
200
150
100
(deg/s)
50
00 60 120 180 240 300 360
-50
-100
-150
-200
(deg)
r/R = 0.3 r/R = 0.47 r/R = 0.63 r/R = 0.8 r/R = 0.95
3
2.5
2
1.5
Cl
0.5
r/R = 0.3 r/R = 0.47 r/R = 0.63 r/R = 0.8 r/R = 0.95
2
1.8
1.6
1.4
1.2
1
Cdp
0.8
0.6
0.4
0.2
00 60 120 180 240 300 360
(deg)
Fig.6.32(e):, ,ClandCdpatU=15m/s
327
Chapter6AerodynamicAnalysisoftheNRELPhaseVIWindTurbine
HysteresisloopsClandCd
From the data presented in Figs. 6.32, it is possible to derive the unsteady aerofoil
hysterisis loops for Cl and Cdp. Figs. 6.33(a)(j) illustrate the hysteresis loops for the
differentwindspeedsandradiallocations.The2Dsteadyaerofoildataobtainedfromthe
Delft wind tunnel experiments are also shown. The 3D steady aerofoil data, which were
derived using the blade pressure measurements for nonyawed conditions together with
the freewake code (see Figs. 6.19) are also included. The hysteresis plots are helpful in
establishing whether any particular blade section is operating in an attached or stalled
flowregime.Oneshouldkeepinmindthatinayawedrotor,theseloopsarenotonlydue
tocyclicpitchvariations,butalsoduetoacyclicvariationoftheflowvelocityrelativeto
the blades. In each of these hysteresis plots, the reduced frequency k is included. k is a
parametercharacterizingthedegreeofunsteadinessatwhichanaerofoilisoperating.The
reducedfrequencyisgivenbytheequation:
c
k = (6.7)
2Vr
where Vr is the relative flow velocity at the aerofoil. When k=0, the flow is steady. In
accordancewithLeishman([49],page306),when0<k<0.05,theunsteadinessintheflowis
minimal and the flow may be assumed to be quasisteady. Higher values of k are
consideredunsteady.Valuesofkequalto0.2andaboveareconsideredhighlyunsteady.
At U=5m/s (Figs. 6.33(a) and (b)), the angles of attack are small at all radial locations.
Consequently the flow over the blades is fully attached throughout the whole blade
revolution. Both the lift and drag hysteresis loops are considerably wide at the inboard
sections (r/R=0.3, 0.47) which indicates that unsteady effects here are significant even
thoughnostallispresent.Attheotherradiallocations(r/R=0.63,0.8and0.95),theloopsare
verynarrowandthereforeunsteadyeffectsherearenegligible.AtU=7m/s,thebladesare
operatingathigheranglesofattack.Nostallisyetpresent,eventhoughthe2Dstaticstall
angleisexceededattheinboardregions(r/R=0.3,0.47and0.63).TheCl loopsarelocated
aroundthe3Dsteadyvalueswhichexplainsthefactthatstalldelayispresent,eveninan
unsteadyenvironment.Atthehigherwindspeeds(U=10,13and15m/s),theangleofattack
goes beyond the 3D steady stalling angle, resulting in dynamic stall. It may be observed
thatmuchlargerphasevariationsinClandCdparepresentthanforattachedflowconditions
(U=5,7m/s).Thisisaresultofsignificanthysteresisintheflowstructures.ThevaluesofCl
and Cdp at the same angle of attack will vary on whether the flow is separating or re
attaching. Insome severe cases however, flow reattachment may not take placesincethe
minimum angle of attack will be too high. Figs. 6.33 (e)(j) indicate that dynamic stall is
mostsevereattheinboardbladesections.Onemayobservethattheunsteadyliftanddrag
coefficientsmaywellexceednotonlythe2Dsteadyvalues,butalsothe3Dsteadyvalues
derivedfornonyawedconditions.
328
Chapter6AerodynamicAnalysisoftheNRELPhaseVIWindTurbine
Someimportantcommentsconcerningthederivedhsyteresisloopsare:
at each wind speed the hysteresis loops for both Cl and Cdp are larger at the
inboardbladesectionsandunsteadyeffectsarelesspronouncedattheoutboard
regions. However severe unsteady effects may still take place at the blade tip
regions, especially at high angles of attack (refer to Figs. 6.33(g) and (i) for
r/R=0.95)
ingeneral,boththeClandCdphysteresisloopsareclosertothe3Dsteadyvalues
thantothecorresponding2Dsteadyvalues.Thisisobservedatbothattachedand
stall conditions and it implies that 3D effects have a significant role in the
unsteady flow behaviour of wind turbine blades, in particular stalldelay. This
impliesthatacorrectionfor3Deffectsisalsonecessarywhencorrecting2Dstatic
aerofoildataforunsteadyaerodynamicconditions.
The hysteresis loops often change direction from counterclockwise to clockwise
when moving outboard from r/R=0.3 to 0.95. This is mainly observed in the lift
hysteresisloopsatU=10and15m/s(seeFigs.6.33(c)and(e)).
Figs. 6.33 show that the obtained values for k at the different radial locations and wind
speedsoftheNRELrotor.Sinceinayawedrotortheflowrelativevelocityateachblade
section (Vr) is a function of blade azimuth angle (), then the maximum and minimum
valuesareincluded.Themaximumandminimumvaluesofkoccuratapproximateblade
azimuth angles of 00 and 1800 respectively. It may be observed that the highest reduced
frequencies occur at low r/R values. However, changing the windspeed does not alter
significantly the reduced frequencies at each radial position significantly. This is a
consequenceofthefactthattherotorangularspeediskeptconstantandthewindtunnel
speedismuchsmallerthantherotorangularspeed(U<<r).
More detailed analysis of the unsteady aerodynamic effects may be carried out by
analyzing these derived hysteresis loops in conjunction with the analysis of the unsteady
chordwisebladepressuredistributionsmeasuredinthewindtunnel.Thelatteranalysisis
veryhelpfulinidentifyingtheflowcharacteristicsofthebladesateachangleofattack,as
describedbySchrecketal.[74,76].
BoundCirculationDistributionsatBlades
Fig.6.34illustrates3Dplotsoftheboundcirculationdistributionsatthebladesasafunction
of radial position (r/R) and blade azimuth angle (). For low wind speeds (U=5, 7m/s), at
which the flow over the blades is considerably attached, the 3D plot takes the form of a
saddle.Ateachbladeazimuthangle,thepeakboundcirculationoccursclosetothemiddle
blade section. Also, the peak circulation is highest at blade position equal to 0/600 and is
lowestatabout1800.Athighwindspeeds,whereseparatedflowconditionsareknownto
occur (U=10, 13, 15m/s), the saddleshaped distribution is no longer present and a rather
329
Chapter6AerodynamicAnalysisoftheNRELPhaseVIWindTurbine
irregular distribution occurs. The peak bound circulation tends to shift towards the root
regionoftheblade,whereinfactthehighestliftcoefficientsarenoted.Howeverthepeak
bound circulation is still observed at blade position 0/3600. It may be observed from Fig.
6.34thathighwindspeedsresultinhigherlevelsofboundcirculation.
330
Chapter6AerodynamicAnalysisoftheNRELPhaseVIWindTurbine
1 0.1
0.9 0.09
0.08 3D Unsteady
0.8
C dp
0.05
Cl
0.4
kmin = 0.19
0.04
0.3 3D Unsteady
2D Steady Delft
0.03
0.2 0.02
0.1
3D Steady
0.01
-2
0
0 2 4 6 8 10 12 -2
0
0 2 4 6 8 10 12
(deg) (deg)
1 0.1
0.9 0.09
0.6 0.06 k
max = 0.15
k = 0.12
min
C dp
Cl
0.5 0.05
0.4
k
max
k
min
= 0.15
= 0.12
0.04
0.3
3D Unsteady
0.03
0.1 0.01
0 0
0 2 4 6 8 10 -2 0 2 4 6 8 10 12
(deg)
(deg)
1 0.1
0.9 0.09
3D Unsteady
0.8 0.08
2D Steady Delft
0.7 0.07 3D Steady
r/R=0.63 0.6 r/R=0.63
0.06 k
max = 0.1
k = 0.08
min
C dp
Cl
0.5 0.05
k
max = 0.1
0.4
k
min = 0.08 0.04
0.3 0.03
3D Unsteady
0.2 2D Steady Delft 0.02
3D Steady
0.1 0.01
0 0
0 2 4
(deg)
6 8 10 -2 0 2 4
(deg)
6 8 10 12
1.0 0.08
0.9 0.07
0.8
3D Unsteady
2D Steady Delft
0.06
3D Steady
r/R=0.8 0.7 r/R=0.8
0.05
0.6 k = 0.06
max
k = 0.05
min
Cdp
0.04
Cl
0.5
k
max = 0.06
0.4 k
min = 0.05 0.03
0.3
0.2
3D Unsteady 0.02
2D Steady Delft
0.01
0.1
3D Steady
0.0 0
0 2 4 6 8 10 0 2 4 6 8 10 12
(deg) (deg)
1.0
0.08
0.9
0.8
k
max
k
min
= 0.04
= 0.04
0.07
3D Unsteady
k = 0.04
Cl
0.4
0.03
0.3
0.02
3D Unsteady
0.2
2D Steady Delft
3D Steady
0.01
0.1
0.0 0
0 2 4 6 8 10 0 2 4 6 8 10 12
(deg) (deg)
Fig.6.33(a):ClhysteresisloopsatU=5m/sFig.6.33(b):CdphysteresisloopsatU=5m/s
331
Chapter6AerodynamicAnalysisoftheNRELPhaseVIWindTurbine
1.2 0.08
1.1 3D Unsteady
1
0.07
2D Steady Delft
3D Steady
r/R=0.3 0.9 r/R=0.3 0.06
k = 0.27 max
0.05
0.8 k = 0.17 min
C dp
Cl
0.7 0.04
0.6
3D Unsteady
2D Steady Delft 0.03
0.5 3D Steady
0.02
0.4
k = 0.27
max
0.3 k = 0.17
min 0.01
0.2
1 3 5 7 9 11 13 15
0
-2 0 2 4 6 8 10 12 14
(deg)
(deg)
1.2 0.08
1.1
0.07
3D Unsteady
r/R=0.47 1 r/R=0.47 0.06 2D Steady Delft
0.9 3D Steady
0.8 0.05
k = 0.16 max
C dp
k = 0.11 min
Cl
0.5
3D Steady 0.03
k = 0.16
max 0.02
0.4 k = 0.11
min
0.3 0.01
0.2 0
1 3 5 7 9 11 13
(deg)
-2 0 2 4 6
(deg)
8 10 12 14
1.1 0.08
1 0.07
0.9
3D Unsteady
2D Steady Delft
r/R=0.63 r/R=0.63 0.06
0.8 3D Steady
0.05
k = 0.1
C dp
0.7 max
k = 0.07
Cl
0.04 min
3D Unsteady
0.6
2D Steady Delft
3D Steady 0.03
0.5
0.4 k = 0.1
max 0.02
k = 0.07
min
0.3 0.01
0.2
1 3 5 7 9 11 13
0
0 2 4 6 8 10 12 14
(deg) (deg)
1.1 0.08
1
0.07
3D Unsteady
2D Steady Delft
0.9
r/R=0.8 r/R=0.8 0.06 3D Steady
0.8
k = 0.06 max
0.7
0.05
k = 0.05 min
Cdp
Cl
3D Unsteady 0.04
0.6 2D Steady Delft
3D Steady 0.03
0.5
k = 0.06
max
0.02
0.4 k = 0.05
min
0.3 0.01
0.2 0
1 3 5 7 9 11 13 0 2 4 6 8 10 12 14
(deg) (deg)
1.1 0.08
1
0.07
3D Unsteady
r/R=0.95 0.9
r/R=0.95 0.06 2D Steady Delft
0.8 3D Steady
0.05
Cdp
Cl
0.04
k = 0.04 min
0.6
3D Unsteady 0.03
0.5
2D Steady Delft
3D Steady 0.02
0.4
0.3
k = 0.04
max
0.01
k = 0.04
min
0.2 0
1 3 5 7 9 11 13 0 2 4 6 8 10 12 14
(deg) (deg)
Fig.6.33(c):ClhysteresisloopsatU=7m/sFig.6.33(d):CdphysteresisloopsatU=7m/s
332
Chapter6AerodynamicAnalysisoftheNRELPhaseVIWindTurbine
2.2 0.8
2
3D Unsteady
0.7
3D Unsteady
2D Steady Delft 2D Steady Delft
r/R=0.3 1.8
3D Steady
r/R=0.3 0.6
3D Steady
k = 0.27
1.6 max
k = 0.15
min 0.5 kmax = 0.27
1.4 kmin = 0.15
Cl
C dp
0.4
1.2
1
0.3
0.2
0.8
0.6 0.1
0.4
0
2 7 12 17 22
(deg)
5 10 15
(deg)
20 25
1.6 0.6
3D Unsteady 3D Unsteady
1.4
2D Steady Delft 0.5 2D Steady Delft
3D Steady
r/R=0.47 3D Steady r/R=0.47
1.2
0.4
kmax = 0.15
1
kmin = 0.1
C dp
Cl
0.3
0.8
0.2
kmax = 0.15
0.6
kmin = 0.1
0.1
0.4
0
2 7 12
(deg)
17 22
5 10 15 20 25
(deg)
1.4 0.3
0.9 0.15
0.8
0.1
0.7
kmax = 0.1
0.6
kmin = 0.07
0.5
0.05
0.4 0
5 7 9 11 13 15 17 19
2 4 6 8 10 12 14 16
(deg) (deg)
1.2 0.12
1.1
3D Unsteady
3D Unsteady
2D Steady Delft
0.1 2D Steady Delft
r/R=0.8 1 3D Steady
r/R=0.8 3D Steady
0.08
0.9 kmax = 0.06
kmin = 0.05
Cdp
0.8
0.06
Cl
0.7
0.6
kmax = 0.06
kmin = 0.05
0.04
0.5
0.02
0.4 0
2 4 6 8 10 12 14 16 5 7 9 11 13 15
(deg)
(deg)
1 0.06
0.9
3D Unsteady
2D Steady Delft 3D Unsteady
0.05
r/R=0.95 0.8
3D Steady
r/R=0.95
2D Steady Delft
3D Steady
0.04
0.7 kmax = 0.04
kmin = 0.03
Cdp
Cl
0.6 0.03
0.5
0.02
0.4 kmax = 0.04
0.3
kmin = 0.03
0.01
0.2 0
2 4 6 8 10 12 5 6 7 8 9 10 11 12
(deg)
(deg)
Fig.6.33(e):ClhysteresisloopsatU=10m/sFig.6.33(f):CdphysteresisloopsatU=10m/s
333
Chapter6AerodynamicAnalysisoftheNRELPhaseVIWindTurbine
3
2
3D Unsteady
2.5 2D Steady Delft
1.8 3D Unsteady
3D Steady 1.6 2D Steady Delft
r/R=0.3 r/R=0.3
1.4
3D Steady
2 kmax = 0.24
kmax = 0.24
kmin = 0.13
1.2 kmin = 0.13
Cl
C dp
1.5
1
0.8
1
0.6
0.5 0.4
0.2
0
0
0 5 10 15 20 25 30 35 40
0 5 10 15 20 25 30 35 40
(deg) (deg)
3
2
3D Unsteady 1.8
2.5
3D Unsteady
2D Steady Delft
r/R=0.47 3D Steady r/R=0.47
1.6 2D Steady Delft
3D Steady
2 1.4
kmax = 0.15
1.2
kmax = 0.15
Cl
C dp
1.5 1
0.8
1
0.6
0.5 0.4
0.2
0 0
0 5 10 15 20 25 30 35 40
(deg)
0 5 10 15 20
(deg)
25 30 35 40
3
2
1.8
2.5 3D Unsteady
2D Steady Delft 1.6
3D Unsteady
r/R=0.63 r/R=0.63 2D Steady Delft
3D Steady
2 1.4 3D Steady
kmax = 0.1 kmax = 0.1
Cl
1.2
kmin = 0.07 kmin = 0.07
C dp
1.5
1
1
0.8
0.6
0.4
0.5
0.2
0
0 5 10 15 20 25 30 35 40
0
0 5 10 15 20 25 30 35 40
(deg) (deg)
3
2
2.5
3D Unsteady
2D Steady Delft
1.8
3D Unsteady
1.6 2D Steady Delft
r/R=0.8
2
3D Steady r/R=0.8
1.4
3D Steady
kmax = 0.06
kmin = 0.04 1.2 kmax = 0.06
kmin = 0.04
C dp
Cl
1.5
1
0.8
1
0.6
0.5 0.4
0.2
0
0 5 10 15 20 25 30 35 40
0
0 5 10 15 20 25 30 35 40
(deg)
3
(deg)
2
3D Unsteady 1.8
2.5 2D Steady Delft 3D Unsteady
r/R=0.95
3D Steady
r/R=0.95
1.6 2D Steady Delft
3D Steady
2 1.4
kmax = 0.04
kmin = 0.03 1.2 kmax = 0.04
Cdp
kmin = 0.03
Cl
1.5
1
0.8
1
0.6
0.4
0.5
0.2
0
0 5 10 15 20 25 30 35 40
0
0 5 10 15 20 25 30 35 40
(deg)
(deg)
Fig.6.33(g):ClhysteresisloopsatU=13m/sFig.6.33(h):CdphysteresisloopsatU=13m/s
334
Chapter6AerodynamicAnalysisoftheNRELPhaseVIWindTurbine
3
2
2.5
3D Unsteady
2D Steady Delft
1.8
3D Unsteady
1.6 2D Steady Delft
r/R=0.3
2
3D Steady r/R=0.3
1.4
3D Steady
1.2
Cl
1.5
Cdp
1
0.8
1 kmax = 0.21
kmin = 0.12 0.6
kmax = 0.21
kmin = 0.12
0.5 0.4
0.2
0
0
0 10 20 30 40 50
(deg) 0 10 20 30 40 50
3
(deg)
2
3D Unsteady
1.8 3D Unsteady
2.5 2D Delft kmax = 0.14
2D Steady Delft
r/R=0.47
3D Steady kmin = 0.09
r/R=0.47
1.6
3D Steady
2 1.4
1.2 kmax = 0.14
kmin = 0.09
Cl
1.5
Cdp
1
0.8
1
0.6
0.5 0.4
0.2
0 0
0 10 20 30 40 50
(deg)
0 10 20
(deg)
30 40 50
3
2
Cl
Cdp
1.5
1
1
0.8
0.6
0.4
0.5
0.2
0 0
0 10 20
(deg)
30 40 50 0 10 20 30 40 50
(deg)
3 2
2.5
3D Unsteady
kmax = 0.06
1.8 3D Unsteady
2D Steady Delft
2D Steady Delft kmin = 0.04 1.6
r/R=0.8
2
3D Steady r/R=0.8
1.4
3D Steady
Cdp
1.5 1
0.8
1
0.6
0.4
0.5
0.2
0
0 10 20 30 40 50
0
0 10 20 30 40 50
(deg)
(deg)
3 2
3D Unsteady
1.8
3D Unsteady
2.5 kmax = 0.05
2D Steady Delft 1.6 2D Steady Delft
r/R=0.95 3D Steady
kmin = 0.03
r/R=0.95 1.4
3D Steady
2
1.2
kmax = 0.05
Cdp
Cl
0.8
1
0.6
0.5
0.4
0.2
0 0
0 10 20 30 40 50 0 10 20 30 40 50
(deg) (deg)
Fig.6.33(i):ClhysteresisloopsatU=15m/sFig.6.33(j):CdphysteresisloopsatU=15m/s
335
Chapter6AerodynamicAnalysisoftheNRELPhaseVIWindTurbine
U=5m/s U=13m/s
U=7m/s U=15m/s
U=10m/s
Figure6.34Variationofboundcirculationatbladeswithradiallocationandbladeazimuthangle.
336
Chapter6AerodynamicAnalysisoftheNRELPhaseVIWindTurbine
WakePlots
Figs. 6.35(a)(j) present the freewake plots derived from the measured aerodynamic
loadingontheblades.Colourcodingisusedtodenotethetrailingandshedcirculationin
the near wake. For each windspeed, two plots are presented: one for trailing circulation,
the other for shed circulation. From the wake geometry, it is possible to derive the
approximate pitch of the wake vortex sheets. In the study this was derived by taking a
crosssectionalplane through the freewake, with the plane being horizontal and passing
throughthe rotorhub. Fig.6.36illustratesthederivedvariationofthewakevortexsheet
pitchwiththewindspeedforyawangleof300.Duetowakeskewnessresultingfromrotor
yaw,thewakevortexsheetpitchontheupstreamsideofthewakeisdifferentfromthaton
thedownstreamside.InFig.6.36,thecorrespondingvortexsheetpitchvariationfornon
yawed conditions from Fig. 6.18(a) is also included. In both nonyawed and yawed
conditions, the vortex sheet pitch increases approximately linearly with wind speed. For
theyawedcase,thepitchesontheupstreamanddownstreamsideofthewaketendtobe
veryclose.Howeveritmaybeobservedthat,forallwindspeeds,therotoryawdecreases
the vortex sheet pitch below that for zero yaw. It was very difficult to derive the wake
expansionfromthewakegeometrysincethiswasnotedtobeverysmall.
ThecolourcodinginFigs.6.35isveryhelpfulinunderstandingbetterhowtheunsteady
bound circulation at the blades (refer to Fig. 6.34) eventually diffuses into the wake
downstreamoftherotorintheformoftrailingandshedcirculation.Thepresenceofroll
up of the vortex sheets is apparent. Since in a yawed rotor, the bound circulation is a
function of rotor azimuth angle, then the wake trailing and shed circulation will be time
dependent. Looking at Figs. 6.35 will reveal that, at low windspeeds (U=5, 7m/s), the
trailing circulation in the wake tends to be concentrated towards the blade tip and root
regions.Thisisaresultofthefactthattheboundcirculationatthebladesishighestatthe
middlesections,butthendecreasessteadilytozerotowardsthebladetipandroot(referto
Figs. 6.34). The trailing circulation at the tip region is positive, while it is negative at the
rootregion.Forhighwindspeedconditionshowever,theboundcirculationdistributionis
irregular and this yields considerable levels of horseshoe trailing circulation to be
released from the middle blade sections. This is especially observed at U=15m/s (Fig.
6.35(i)).Consequently,thetrailingcirculationismoredispersedthoughthewake.Theshed
circulation levels are small at low wind speeds (U=5, 7m/s) when compared with the
trailingcirculation.Athigherwindspeeds(U=10,13,15m/s),highershedcirculationlevels
areobservedatlocalspotsofthewakevortexsheets.Howevertheshedcirculationacross
most of the vortex sheets remains small in magnitude in comparison with the trailing
circulation.Atlowwindspeeds,thehelicalshapeofthevortexsheetsiswelldefined,butat
the higher windspeeds, the vortex sheets become distorted. This distortion is induced by
the unsteady separated air flow formed around the blades which then ends up into the
wake.
337
Chapter6AerodynamicAnalysisoftheNRELPhaseVIWindTurbine
Figure6.35(a)Trailingcirculation(m2/s)distributioninwakeatU=5m/s.
Figure6.35(b)Shedcirculation(m2/s)distributioninwakeatU=5m/s.
338
Chapter6AerodynamicAnalysisoftheNRELPhaseVIWindTurbine
Figure6.35(c)Trailingcirculation(m2/s)distributioninwakeatU=7m/s.
Figure6.35(d)Shedcirculation(m2/s)distributioninwakeatU=7m/s.
339
Chapter6AerodynamicAnalysisoftheNRELPhaseVIWindTurbine
Figure6.35(e)Trailingcirculation(m2/s)distributioninwakeatU=10m/s.
Figure6.35(f)Shedcirculation(m2/s)distributioninwakeatU=10m/s.
340
Chapter6AerodynamicAnalysisoftheNRELPhaseVIWindTurbine
Figure6.35(g)Trailingcirculation(m2/s)distributioninwakeatU=13m/s.
Figure6.35(h)Shedcirculation(m2/s)distributioninwakeatU=13m/s.
341
Chapter6AerodynamicAnalysisoftheNRELPhaseVIWindTurbine
Figure6.35(i)Trailingcirculation(m2/s)distributioninwakeatU=15m/s.
Figure6.35(j)Shedcirculation(m2/s)distributioninwakeatU=15m/s.
342
Chapter6AerodynamicAnalysisoftheNRELPhaseVIWindTurbine
Yaw 0 deg Yaw 30 deg Upstream side Yaw 30 deg Downstream side
14
10
2
5 6 7 8 9 10 11 12 13 14 15
U (m/s)
Figure6.36Vortexsheetpitchvariationasestimatedfromfreewakeplotsat=00and=300.
InducedVelocitiesatRotorplane
HAWT_FWC calculated the 3D unsteady induced velocity fields across the whole
rotorplanefordifferentbladeazimuthanglesfromthetimedependentdistributionsofCn
andCtobtainedfromthebladepressuremeasurements.Thecontourplotsdisplayingthe
induced velocity fields for U=5 and 15m/s at =300 for =00 are plotted in Figs. 6.37. The
asymmetry in the flow field at the rotorplane resulting from the skewed wake may be
easilynotedintheseplots.
Figs. 6.38 and 6.39 illustrate the unsteady axial induction factors at the rotorplane as
predictedbythefreewakeplotinaccordancewiththebladepressuremeasurementstaken
inthewindtunnel.Figs.6.38(a)(e)showthevariationoftheazimuthallyaveragedaxial
induction factor (i.e. annular averaged axial induction factor, a1) at the rotorplane with
rotorazimuthanglefordifferentradiallocations.Figs.6.39(a)(e)showthevariationofthe
axial induction factor at the blade lifting line (a1,c) with blade azimuth angle for different
radial locations. The corresponding tangential and radial induction factors were also
calculated but these are not included in this paper as they were found to be very small
comparedtotherotorangularspeed.
When the windspeed is low (U=5, 7m/s), the free wake model predicts an azimuthally
averagedaxialinductionfactorthatisalmostconstantwithrotorazimuthangle.Increasing
thewindspeedfurtherincreasesthecyclicvariationofa1.Thecycliccomponentishighest
at the inboard blade sections. The cyclic variation has frequency equal to 2p and this is
becausetherotorhastwoblades.
343
Chapter6AerodynamicAnalysisoftheNRELPhaseVIWindTurbine
= 3450 = 450
Tangential induced
velocity field (ux)
= 225 0
= 135 0
= 345 0
= 45 0
Axial induced
velocity field (uy)
= 225 0
= 135 0
= 345 0 = 45 0
Radial induced
velocity field (uz)
= 225 0
= 135 0
Figure6.37(a)3DinducedvelocityfieldatrotorplaneatU=5m/s, =300.Bladeisatanazimuth
angleof0/1800.Viewislookingupstreamfrombehindtherotorplane.
344
Chapter6AerodynamicAnalysisoftheNRELPhaseVIWindTurbine
= 3450 = 450
Tangential induced
velocity field (ux)
= 225 0
= 135 0
= 345 0
= 45 0
Axial induced
velocity field (uy)
= 225 0
= 135 0
= 345 0 = 45 0
Radial induced
velocity field (uz)
= 225 0
= 135 0
Figure6.37(b)3DinducedvelocityfieldatrotorplaneatU=15m/s, =300.Bladeisatanazimuth
angleof0/1800.Viewislookingupstreamfrombehindtherotorplane.
345
Chapter6AerodynamicAnalysisoftheNRELPhaseVIWindTurbine
0 0
0 60 120 180 240 300 360 0 60 120 180 240 300 360
-0.02 -0.02
-0.04
-0.04
-0.06
-0.06
-0.08
a1
a1
-0.08
-0.1
-0.1
-0.12 r/R = 0.3
r/R = 0.47 -0.12
r/R = 0.3
-0.14 r/R = 0.47
r/R = 0.63
-0.14
r/R = 0.63
-0.16 r/R = 0.8 r/R = 0.8
-0.18
r/R = 0.95
-0.16
r/R = 0.95
(deg) (deg)
Fig.(a):U=5m/s Fig.(b):U=7m/s
0 0
0 60 120 180 240 300 360 0 60 120 180 240 300 360
-0.02 -0.01
-0.02
-0.04 -0.03
-0.04
-0.06
a1
a1
-0.05
-0.08 -0.06
r/R = 0.3 r/R = 0.3
r/R = 0.47 -0.07 r/R = 0.47
-0.1 r/R = 0.63 r/R = 0.63
-0.08
r/R = 0.8 r/R = 0.8
-0.12 r/R = 0.95
-0.09
r/R = 0.95
(deg) (deg)
Fig.(c):U=10m/s Fig.(d):U=13m/s
0
0 60 120 180 240 300 360
-0.01
-0.02
-0.03
-0.04
a1
-0.05
-0.06
r/R = 0.3
-0.07
r/R = 0.47
r/R = 0.63
-0.08
r/R = 0.8
r/R = 0.95
-0.09
(deg)
Fig.(e):U=15m/s
Figure6.38Variationofazimuthallyaveragedaxialinductionfactora1withrotorazimuthangle
()at=300andU=5,7,10,13and15m/s.
346
Chapter6AerodynamicAnalysisoftheNRELPhaseVIWindTurbine
Thevariationoftheaxialinducedvelocityatthebladeliftinglinewiththebladeazimuth
angle is periodic and regular at low wind speeds (U=5, 7m/s), especially at the outboard
regions of the blades (see Figs. 6.39(a) and (b)) but then becomes very irregular at higher
windspeeds(U=10,13,15m/s;seeFigs.6.39(c),(d)and(e)).Atlowwindspeeds,theperiodic
variationa1,cwith hasaphaseanglethatchangesdependingontheradialposition(r/R)of
thebladesection.Thisisduetothefactthatinaskewedwake,theproximityoftheblades
tothewakecirculationisafunctionofbothbladeazimuthangleandradiallocation.Recall
from Fig. 6.35 (a) and (c) that at the low windspeeds, the trailing circulation in the wake
tends to be concentrated at the tip and root regions. The outboard blade sections are in
general closer to the wake circulation that is originating from the blade tips at blade
azimuth position of 900 than at 2700. Consequently the induced velocity at the outboard
bladesectionsishigherat900thanat2700,asinfactdepictedinFigs.6.39(a)and(b).Asone
movesinboard,theproximityofthebladesectionstothewakecirculationoriginatingfrom
the tip decreases, but this is followed by an increased proximity to the wake circulation
originatingfromthebladeroots.Thiscausesthemaximuminducedvelocitytooccurata
differentbladeazimuthangle,thuschangingthephaseangleofthevariationofa1,cwith.
At higher windspeeds (U=10, 13 and 15m/s), the presence of horseshoe trailing
circulationlevelsatthemiddlebladesectionsremovetheregularvariationofa1,cwith at
allradiallocations,makingitveryirregular.
Fromtheresultsofthisstudy,onecaneasilyremarkthefollowingdeficienciesofGlauerts
Eqt.3.23(andsimilarmodelsdescribedinChapter3,section3.5):
Glauerts model assumes that the azimuthallyaveraged axial induced velocity (a1)
doesnotvarywithrotorazimuthangle.Althoughthisassumptionmayberealisticat
lowwindspeeds,itdoesnotnecessarilyapplyforhighwindspeedconditionswhere
theflowisseparated.
Glauertsmodeldoesnotcaterforthephaseanglechangewithradiallocationofthe
variationa1,cwith ,resultingfromrootcirculation.Alsoasnotedinreferences[83,
84] root circulation effects create an induced velocity distribution that has a higher
harmoniccontentthanonly1p.
Glauertsmodelisinvalidforhighwindspeedswithstalledflowswerethevariation
a1,cwithbecomesveryirregular.
AsalreadydescribedinChapter5,inthefreevortexcomputationswithHAWT_FWC,the
induction at the lifting line of the blades is due to the freewake trailing and shed
circulation and due to the trailing circulation of the prescribed far wake model. The
inductionattheliftinglineduetoboundcirculationfromthetwobladesiszero.Thisisa
consequence of the fact that a lifting line model is used to represent the blades and also
because both lifting lines are inline with one another (no blade coning). As already
mentionedinsectionA.(page320),inthefreewakecalculations,thenearwakeparameter
nRevwasselectedlargeenoughsothattheinductioncontributedbythefarwakeisvery
small.Thusitmaybeassumedthatthetotalinductionisduetotrailingandshed
347
Chapter6AerodynamicAnalysisoftheNRELPhaseVIWindTurbine
0.05 0
0 60 120 180 240 300 360
0 -0.02
-0.04
0 60 120 180 240 300 360
-0.05
-0.06
-0.08
-0.1
a1,c
a1,c
-0.1
-0.12
-0.15
r/R = 0.3
r/R = 0.47
-0.14 r/R = 0.3
r/R = 0.47
-0.16
-0.2 r/R = 0.63 r/R = 0.63
r/R = 0.8 -0.18 r/R = 0.8
-0.25 r/R = 0.95
-0.2
r/R = 0.95
(deg) (deg)
Fig.(a):U=5m/s Fig.(b):U=7m/s
0 0.05
0 60 120 180 240 300 360
0
a1,c
-0.15 -0.15
Fig.(c):U=10m/s Fig.(d):U=13m/s
0.1
0.05
0
-0.05
0 60 120 180 240 300 360
a1,c
-0.1
-0.15
r/R = 0.3
-0.2
r/R = 0.47
rR = 0.63
-0.25
r/R = 0.8
r/R = 0.95
-0.3
(deg)
Fig.(e):U=15m/s
Figure6.39Variationofaxialinductionfactora1,cwithbladeazimuthangle()at=300and
U=5,7,10,13and15m/s.
348
Chapter6AerodynamicAnalysisoftheNRELPhaseVIWindTurbine
circulationofthenearwakeonly.Usingthisfreewakemodel,itispossibletocalculatethe
induced velocity components resulting from the near wake trailing and shed circulation
separately. In this way it would be possible to determine which circulation type has the
greaterinfluenceontheflowattheblades.TheinducedvelocitiesofFig.6.39areshown
again in Figs. 6.40 for U=5 and 15m/s, but including the individual induced velocity
components due to trailing and shed circulation. The induced velocity by the shed
circulation is in general very small. This is observed even at U=7, 10 and 13m/s. This
provesthatthetrailingcirculationisbyfarmoredominantthantheshedcounterpart.The
inductioncomponentduetothefarwakealoneisalsoincludedinFig.6.40andasitmay
beobserveditisverysmallcomparedtothetotalinduction.
GlobalLoads
The unsteady lift and drag coefficients and induced velocities derived by the freewake
vortex model were used to calculate the lowspeed shaft torque and the blade
flap/edgewise bending moments using the bladeelement theory equations 3.16. These
resultsareshowninFigs.6.41,6.42and6.43weretheyarenotedbyplotsFreewake.Asa
crosscheck for these calculations, the same loads were computed directly from the
experimental data by using a linear interpolation to obtain a distribution for Cn, Ct and
QNORM across the whole blade span and using Eqts. 6.4, 6.5 and 6.6. In the linear
interpolation, Cn and Ct at the blade tips were set equal to zero. The trapezium rule was
adopted for integrating numerically across the whole blade span. These results are
includedinFigs.6.41,6.42and6.43andarereferredtoExp.Theloadsmeasureddirectly
using straingauges are also included in these figures. The results calculated by the free
wakemodelandthosederiveddirectlyfromthepressuremeasurements(Exp)agreevery
well.Howeveragreementwiththeloadsmeasuredbythestraingaugesisnotasgood.The
reasonsforthediscrepanciesarethesameasthoseexplainedforaxialconditions(seepage
306).Gravitationalloadsarenotincludedandthusthecomputededgewisemomentdoes
not include the cyclic component induced by the blade weight. This leads to a large
discrepancyinFigs.6.43.
349
Chapter6AerodynamicAnalysisoftheNRELPhaseVIWindTurbine
total trail shed far wake total trail shed far wake
0.04 0.1
0.02 0.05
0
0
0 60 120 180 240 300 360
0 60 120 180 240 300 360
r/R=0.3 a1,c -0.02 r/R=0.3 -0.05
a1,c
-0.1
-0.04
-0.15
-0.06
-0.2
-0.08
-0.25
-0.1 -0.3
(deg)
(deg)
0.04
total trail shed far wake
0.1
total trail shed far wake
0.02 0.05
0
r/R=0.47 -0.02
0 60 120 180 240 300 360
r/R=0.47
0
0 60 120 180 240 300 360
-0.05
-0.04
a1,c
a1,c
-0.06 -0.1
-0.08 -0.15
-0.1
-0.12
-0.2
-0.25
-0.14
-0.16 -0.3
(deg) (deg)
0
0.05
0
r/R=0.63
0 60 120 180 240 300 360
r/R=0.63
-0.05
0 60 120 180 240 300 360
-0.05
a1,c
a1,c
-0.1
-0.1 -0.15
-0.2
-0.15
-0.25
-0.2 -0.3
(deg) (deg)
total trail shed far wake total trail shed far wake
0.05 0.1
0.05
0
0 60 120 180 240 300 360 0
r/R=0.8 r/R=0.8
-0.05
-0.05
0 60 120 180 240 300 360
a1,c
a1,c
-0.1 -0.1
-0.15
-0.15
-0.2
-0.2
-0.25
-0.25
(deg)
-0.3
(deg)
total trail shed far wake total trail shed far wake
0.05 0.1
0.05
r/R=0.95
0
0 60 120 180 240 300 360 0
r/R=0.95 0 60 120 180 240 300 360
-0.05
-0.05
a1,c
a1,c
-0.1
-0.1 -0.15
-0.2
-0.15
-0.25
-0.2
(deg)
-0.3
(deg)
Fig.6.40(a):contributionofcirculationcomponents Fig.6.40(b):contributionofcirculation
fromnearandfarwaketoinductionfactoratblade componentsfromnearandfarwaketoinduction
liftinglineatU=5m/s factoratbladeliftinglineatU=15m/s
350
Chapter6AerodynamicAnalysisoftheNRELPhaseVIWindTurbine
1200
700 U = 7 m/s Free-wake
U = 5m/s Free-wake 1000 U = 7 m/s Exp
U = 5 m/s Exp
U = 7 m/s Strain gauge
500 U = 5 m/s Strain gauge 800
LSSTQ (Nm)
LSSTQ (Nm)
600
300
400
100
200
-100 0 30 60 90 120 150 180 210 240 270 300 330 360 0
Rotor azimuth angle (deg) 0 30 60 90 120 150 180 210 240 270 300 330 360
Rotor azimuth angle (deg)
Fig.(a):U=5m/s Fig.(b):U=7m/s
2100 2100
U = 10 m/s Free-wake
1900 U = 10 m/s Exp 1900
U = 10 m/s Strain gauge
1700 1700
1500 1500
LSSTQ (Nm)
1300
LSSTQ (Nm)
1300
1100 1100
900 900
700 U = 13 m/s Free-wake
700
U = 13 m/s Exp
500 500
U = 13 m/s Strain gauge
300 300
100 100
0 30 60 90 120 150 180 210 240 270 300 330 360 0 30 60 90 120 150 180 210 240 270 300 330 360
Rotor azimuth angle (deg) Rotor azimuth angle (deg)
Fig.(c):U=10m/s Fig.(d):U=13m/s
2400
2200
2000
1800
1600
LSSTQ (Nm)
1400
1200
1000
800
600
U = 15 m/s Free-wake
U = 15 m/s Exp
400
U = 15 m/s Strain gauge
200
0
0 30 60 90 120 150 180 210 240 270 300 330 360
Rotor azimuth angle (deg)
Fig.(e):U=15m/s
Fig.6.41Variationoflowspeedshafttorquewithrotorazimuthangle()atU=5,7,10,13and
15m/s.
351
Chapter6AerodynamicAnalysisoftheNRELPhaseVIWindTurbine
1000
1400
900
800 1200
700
1000
600
RFM (Nm)
RFM (Nm)
800
500
400 600
Fig.(a):U=5m/s Fig.(b):U=7m/s
3000 4000
3600
2600
3200
2200 2800
RFM (Nm)
RFM (Nm)
1800 2400
2000
1400
1600
1000 U = 10 m/s Free-wake 1200
U = 13 m/s Free-wake
600 U = 10 m/s Exp 800 U = 13 m/s Exp
U = 10 m/s Strain gauge 400 U = 13 m/s Strain gauge
200 0
0 30 60 90 120 150 180 210 240 270 300 330 360 0 30 60 90 120 150 180 210 240 270 300 330 360
Rotor azimuth angle (deg) Rotor azimuth angle (deg)
Fig.(c):U=10m/s Fig.(d):U=13m/s
4400
4000
3600
3200
2800
RFM (Nm)
2400
2000
1600
1200
800
U = 15 m/s Free-wake
U = 15 m/s Exp
400 U = 15 m/s Strain gauge
0
0 30 60 90 120 150 180 210 240 270 300 330 360
Fig.(e):U=15m/s
Fig.6.42Variationofbladerootflapbendingmomentwithrotorazimuthangle()atU=5,7,10,
13and15m/s.
352
Chapter6AerodynamicAnalysisoftheNRELPhaseVIWindTurbine
1500 1600
1200 1400
U = 5m/s Free-wake
U = 5 m/s Exp
1200 U = 7 m/s Free-wake
900 U = 7 m/s Exp
1000
600 U = 5 m/s Strain gauge
U = 7 m/s Strain gauge
800
300 600
REM (Nm)
REM (Nm)
400
0 200
0 30 60 90 120 150 180 210 240 270 300 330 360
-300 0
-600 -200 0 30 60 90 120 150 180 210 240 270 300 330 360
-400
-900
-1200 -600
-800
-1500 -1000
Rotor azimuth angle (deg) Rotor azimuth angle (deg)
Fig.(a):U=5m/s Fig.(b):U=7m/s
2600 2600
2200 U = 10 m/s Free-wake 2200 U = 13 m/s Free-wake
U = 10 m/s Exp U = 13 m/s Exp
1800 U = 10 m/s Strain gauge 1800
U = 13 m/s Strain gauge
1400 1400
REM (Nm)
REM (Nm)
1000 1000
600 600
200 200
-200 0 30 60 90 120 150 180 210 240 270 300 330 360 -200 0 30 60 90 120 150 180 210 240 270 300 330 360
-600
-600
-1000 -1000
Rotor azimuth angle (deg) Rotor azimuth angle (deg)
Fig.(c):U=10m/s Fig.(d):U=13m/s
2600
2200 U = 15 m/s Free-wake
1800
U = 15 m/s Exp
U = 15 m/s Strain gauge
1400
REM (Nm)
1000
600
200
-200 0 30 60 90 120 150 180 210 240 270 300 330 360
-600
-1000
Rotor azimuth angle (deg)
Fig.(e):U=15m/s
Fig.6.43Variationofbladerootedgebendingmomentwithrotorazimuthangle()atU=5,7,10,
13and15m/s.
353
Chapter6AerodynamicAnalysisoftheNRELPhaseVIWindTurbine
ComparisonofInflowAngle
Figs.6.446.48comparethelocalinflowangles(LFA)predictedbyHAWT_FWCwiththose
measuredbytheflowdirectionprobesduringthewindtunnelexperiments(seeFig.6.4)at
=300 and windspeeds U=5, 7, 10, 13 and 15m/s. The angle of attack derived by
HAWT_FWCfromthebladepressuremeasurementsisalsoincluded.Intheidealsituation
theLFAvaluescomputedbyHAWT_FWCandthosemeasuredbytheprobesinthetunnel
should be equal. But the correlation is not always good due to the various reasons
explained already when describing the results for axial conditions (see page 309). The
influenceofbladedeflectionissmall,evenatyawedconditionsandcouldnotbeconsidered
asamajorsourceoferror.AcalculationontheECNaeroelasticcodePHATASconfirmthat
blade deflection is small at =300 (maximum blade tip deflection<10cm), (ECN, Gerard
Schepers, personal communication). Also, as it may be noted from Figs. 6.44 6.48, good
correlationintheLFAwasachievedattheoutboardbladesections(r/R>0.67)whereblade
deflection is normally largest. Surprisingly, this good correlation in the outboard blade
sectionsisobservedevenatthehighwindspeedsatwhichtheangleofattackexceedsthe
2Dstaticstallingangle(whichisaround100).
Attheinboardsections,thediscrepancybetweentheHAWT_FWCandexperimentalvalues
oftheLFAissometimeslargeatthewindspeeds(U=10,13and15m/s),reachingavalueof
about100.Themostprobablereasonsforthisdiscrepancyaretwo:(1)thedeficiencyofthe
lifting line model implemented in HAWT_FWC to model 3D effects on the blades, as
alreadynotedalreadywhen treatingaxialconditions;(2)thefactat =300,theprobewas
subjected to highly unsteady flow situations which are most prominent at the inboard
bladesections.AtU=10m/s,HAWT_FWCestimatesthatthetimerateofchangeofangleof
attack reaches a peak of about 78deg/s at r/R=0.3. At U=15m/s, this value exceeds
180deg/s(seeFigs.6.32(c),(d)and(e)).Furtherworkisrequiredtoestablishtheaccuracy
with which fivehole pressure probes manage to measure the LFA at such high levels of
unsteadiness.
354
Chapter6AerodynamicAnalysisoftheNRELPhaseVIWindTurbine
8 12
Free-wake, angle of attack Free-wake, angle of attack
6 Free-wake, LFA 10
Free-wake, LFA
Exp, LFA Exp, LFA
or LFA (deg)
4
or LFA (deg)
2 6
0 4
0 30 60 90 120 150 180 210 240 270 300 330 360
2
-2
0
-4
0 30 60 90 120 150 180 210 240 270 300 330 360
(deg)
(deg)
Fig.(a):r/R=0.34 Fig.(b):r/R=0.51
6 6
Free-wake, angle of attack
Free-wake, LFA
Free-wake, angle of attack
Free-wake, LFA
5 5
Exp, LFA Exp, LFA
4 4
or LFA (deg)
or LFA (deg)
3 3
2 2
1 1
0 0
0 30 60 90 120 150 180 210 240 270 300 330 360 0 30 60 90 120 150 180 210 240 270 300 330 360
(deg) (deg)
Fig.(c):r/R=0.67 Fig.(d):r/R=0.84
6
Free-wake, angle of attack
Free-wake, LFA
5
Exp, LFA
4
or LFA (deg)
0
0 30 60 90 120 150 180 210 240 270 300 330 360
(deg)
Fig.(e):r/R=0.91
Fig.6.44VariationofLFAand withbladeazimuthangle()atU=5m/sand=300.
355
Chapter6AerodynamicAnalysisoftheNRELPhaseVIWindTurbine
25 20
Free-wake, angle of attack Free-wake, angle of attack
18
Free-wake, LFA Free-wake, LFA
20 Exp, LFA 16 Exp, LFA
14
or LFA (deg)
or LFA (deg)
15 12
10
10 8
6
5 4
2
0 0
0 30 60 90 120 150 180 210 240 270 300 330 360 0 30 60 90 120 150 180 210 240 270 300 330 360
(deg) (deg)
Fig.(a):r/R=0.34 Fig.(b):r/R=0.51
12 12
Free-wake, angle of attack Free-wake, angle of attack
Free-wake, LFA Free-wake, LFA
10 Exp, LFA 10
Exp, LFA
8 8
or LFA (deg)
or LFA (deg)
6 6
4 4
2 2
0 0
0 30 60 90 120 150 (deg) 180 210 240 270 300 330 360 0 30 60 90 120 150 180 210 240 270 300 330
(deg)
360
Fig.(c):r/R=0.67 Fig.(d):r/R=0.84
12
Free-wake, angle of attack
10
Free-wake, LFA
Exp, LFA
8
or LFA (deg)
2
0
0 30 60 90 120 150 180 210 240 270 300 330 360
(deg)
Fig.(e):r/R=0.91
Fig.6.45VariationofLFAand withbladeazimuthangle()atU=7m/sand=300.
356
Chapter6AerodynamicAnalysisoftheNRELPhaseVIWindTurbine
45 35
Free-wake, angle of attack Free-wake, angle of attack
40
Free-wake, LFA 30 Free-wake, LFA
Exp, LFA
35 Exp, LFA
25
or LFA (deg)
30
or LFA (deg)
25 20
20 15
15
10
10
5
5
0 0
0 30 60 90 120 150 180 210 240 270 300 330 360 0 30 60 90 120 150 180 210 240 270 300 330 360
(deg) (deg)
Fig.(a):r/R=0.34 Fig.(b):r/R=0.51
25 18
Free-wake, angle of attack
16
Free-wake, LFA
20 Exp, LFA 14
or LFA (deg)
12
or LFA (deg)
15
10
8
10
6
Free-wake, angle of attack
5 4 Free-wake, LFA
Exp, LFA
2
0 0
0 30 60 90 120 150 (deg) 180 210 240 270 300 330 360 0 30 60 90 120 150 180 210 240 270 300 330 360
(deg)
Fig.(c):r/R=0.67 Fig.(d):r/R=0.84
18
16
14
12
or LFA (deg)
10
8
6
Free-wake, angle of attack
4
Free-wake, LFA
Exp, LFA
2
0
0 30 60 90 120 150 180 210 240 270 300 330 360
(deg)
Fig.(e):r/R=0.91
Fig.6.46VariationofLFAand withbladeazimuthangle()atU=10m/sand=300.
357
Chapter6AerodynamicAnalysisoftheNRELPhaseVIWindTurbine
70 50
Free-wake, angle of attack Free-wake, angle of attack
45
60 Free-wake, LFA Free-wake, LFA
Exp, LFA 40 Exp, LFA
50 35
or LFA (deg)
or LFA (deg)
40 30
25
30
20
20 15
10
10 5
0 0
0 30 60 90 120 150 180 210 240 270 300 330 360 0 30 60 90 120 150 180 210 240 270 300 330 360
(deg) (deg)
Fig.(a):r/R=0.34 Fig.(b):r/R=0.51
35 25
Free-wake, angle of attack
30 Free-wake, LFA
Exp, LFA 20
25
or LFA (deg)
or LFA (deg)
15
20
15 10
25
20
or LFA (deg)
15
10
Free-wake, angle of attack
5 Free-wake, LFA
Exp, LFA
0
0 30 60 90 120 150 180 210 240 270 300 330 360
(deg)
Fig.(e):r/R=0.91
Fig.6.47VariationofLFAand withbladeazimuthangle()atU=13m/sand=300.
358
Chapter6AerodynamicAnalysisoftheNRELPhaseVIWindTurbine
80 60
Free-wake, angle of attack Free-wake, angle of attack
70 Free-wake, LFA
50
Free-wake, LFA
Exp, LFA Exp, LFA
60
or LFA (deg)
or LFA (deg)
40
50
40 30
30
20
20
10
10
0 0
0 30 60 90 120 150 180 210 240 270 300 330 360 0 30 60 90 120 150 180 210 240 270 300 330 360
(deg) (deg)
Fig.(a):r/R=0.34 Fig.(b):r/R=0.51
40 30
Free-wake, angle of attack Free-wake, angle of attack
35 Free-wake, LFA Free-wake, LFA
25
30 Exp, LFA Exp, LFA
25
or LFA (deg)
20
or LFA (deg)
20 15
15
10
10
5 5
0 0
0 30 60 90 120 150 180 210 240 270 300 330 360 0 30 60 90 120 150 180 210 240 270 300 330 360
(deg) (deg)
Fig.(c):r/R=0.67 Fig.(d):r/R=0.84
30
Free-wake, angle of attack
25
Free-wake, LFA
Exp, LFA
or LFA (deg)
20
15
10
5
0
0 30 60 90 120 150 180 210 240 270 300 330 360
(deg)
Fig.(e):r/R=0.91
Fig.6.48VariationofLFAand withbladeazimuthangle()atU=15m/sand=300.
359
Chapter6AerodynamicAnalysisoftheNRELPhaseVIWindTurbine
ComparisonofResultswiththoseofFreeWakeVortexModelAWSM
TofurtherevaluatethereliabilityoftheresultsfromHAWT_FWC,ablindcomparisonwas
performedwiththeresultspredictedbyanotherfreewakevortexmodel,AWSMdeveloped
byvanGarrelatECN[26].AWSMisalsoafreewakevortexmodelthatmodelstheblades
as lifting lines and the wake as sheets, each consisting of a mesh of straightline vortex
filaments to account for both trailing and shed circulation. The main difference from
HAWT_FWCismainlythatAWSMmakesuseofaerofoildata(ClandCdasfunctionof)to
determineiterativelytheaerodynamicloadsandinductionattherotorplane.Recallthatin
HAWT_FWC, no aerofoil data is used, but the wind tunnel measurements of Cn and Ct.
FurtherdetailsonAWSMmaybefoundin[26].ThecomparisonwaslimitedtoU=5m/sand
=300only,forwhichthestatic2DS809aerofoildatausedinAWSMisreasonablycloseto
that observed in the NASA Ames tunnel experiments. The results for the axial induction
factor at the blade lifting line (a1,c) from the two codes are compared in Fig. 6.49. It was
found that the results agree very well. This is very encouraging and this gives more
confidenceintheresultspresentearlierderivedusingHAWT_FWC.
360
Chapter6AerodynamicAnalysisoftheNRELPhaseVIWindTurbine
0.06 0
0 60 120 180 240 300 360
0.04
-0.02
0.02
0 -0.04
AWSM
-0.02 0 60 120 180 240 300 360 -0.06
HAWT_FWC
-0.04 -0.08
a1,c
a1,c
-0.06 -0.1
-0.08 -0.12
AWSM
-0.1 HAWT_FWC -0.14
-0.12
-0.14 -0.16
(deg) (deg)
Fig.(a):r/R=0.3 Fig.(b):r/R=0.47
0 0
0 60 120 180 240 300 360 0 60 120 180 240 300 360
-0.02 -0.02
-0.04 -0.04
-0.06 -0.06 AWSM
AWSM
-0.08 HAWT_FWC -0.08
HAWT_FWC
-0.1 -0.1
a1,c
a1,c
-0.12 -0.12
-0.14 -0.14
-0.16 -0.16
-0.18 -0.18
-0.2 -0.2
(deg) (deg)
Fig.(c):r/R=0.63 Fig.(d):r/R=0.8
0
0 60 120 180 240 300 360
-0.02
-0.04 AWSM
HAWT_FWC
-0.06
-0.08
-0.1
a1,c
-0.12
-0.14
-0.16
-0.18
-0.2
(deg)
Fig.(e):r/R=0.95
Fig.6.49Comparisonoftheaxialinductionfactordistributionatthebladeliftingline(a1,c)
computedbyHAWT_FWCwiththosefromfreewakevortexmodelAWSMatU=5m/sand=
300.AWSMresultsfromECN(GerardSchepers,personalcommunication).
361
Chapter6AerodynamicAnalysisoftheNRELPhaseVIWindTurbine
6.4BEMPredictionsfortheNRELPhaseVIRotorwithNew
AerofoilDataandInflowCorrections
Asalreadyoutlinedinsection6.1,themainobjectivesforthisanalysisontheNRELrotor
was to improve BEMbased design codes. The new aerofoil data derived from the blade
pressuremeasurementsinconjunctionwithHAWT_FWC(see Figs.6.19for =00and Figs.
6.33 for =300) were used in the BladeElementMomentum (BEM) theory to model the
NREL Phase VI rotor in both axial and yawed conditions. For yawed conditions, inflow
corrections for BEM codes were derived from the inflow results of HAWT_FWC (see Figs.
6.38and6.39).Inthisanalysis,theBEMmodeldevelopedinthisproject(HAWT_BEM)was
used for the computations. The results computed by this BEM code were repeated with
differentaerofoildata(with2Dstaticdata/withnew3DdatafromHAWT_FWC)andinthe
case of yawed conditions also with different inflow corrections (with/without inflow
correctionsfromHAWT_FWC).Inthiswayitwaspossibletoassessbothquantitativelyand
qualitatively the improvements in BEM predictions if the new aerofoil data and inflow
correctionsareused.
An important point to make is that in the following BEM computations with the new
aerofoil data derived from the NREL blade pressure measurements and HAWT_FWC, the
drag coefficient is only due to blade pressure (i.e. Cdp). Hence the drag coefficient due to
skinfrictionisbeingignored.
This analysis is divided into two parts: section 6.4.1 will consider axial conditions only
(=00)whilesection6.4.2issolelydedicatedtoyawedconditions(=300).
362
Chapter6AerodynamicAnalysisoftheNRELPhaseVIWindTurbine
6.4.1AxialConditions
The3DaerofoildataderivedfromtheNRELbladepressuremeasurementsinconjunction
with HAWT_FWC for =00 (Figs. 6.19 (a) and (b)) were double interpolated to yield the
aerofoildataasafunctionoftwovariables,i.e.theaerofoildataatdifferentradiallocations
andanglesofattack(i.e.intheformCl,Cd(r/R,)).Acubicsplineinterpolationasdescribed
inAppendixDwasusedtoobtaintheaerofoildataasafunctionofr/Rwiththeboundary
conditionsthatbothClandCdarezeroatthebladetipandroot.Asplineinterpolationwas
selectedsinceunlikelinearinterpolation,thiswasfoundtoyieldmorerealisticallysmooth
variationofloadingatthebladetipandroot.
To be able to assess the improvement in the BEM predictions when using the new 3D
aerofoildata,eachoperatingconditionontheNRELrotorwasmodelledwithHAWT_BEM
using both the standard 2D static aerofoil data from the Delft wind tunnel and the new
aerofoildata.Table6.2liststhedifferentcomputationsperformed.
Table6.2:ComputationsperformedonNRELrotorfor=00usingHAWT_BEM
Computation BEMCodeConfiguration
Reference
BEM1 2D static aerofoil data (from Delft wind tunnel) extrapolated
usingViternaCorriganmodel(Fig.6.50)
Prandtltip/rootcorrectionfactorincluded
BEM2 New3DstaticaerofoildataobtainedfromNRELbladepressure
measurementsandHAWT_FWC(Fig.6.19)
Prandtltip/rootcorrectionfactorincluded
BEM3 New3DstaticaerofoildataobtainedfromNRELbladepressure
measurementsandHAWT_FWC(Fig.6.19)
Prandtltip/rootcorrectionfactorNOTincluded
WhethertoincludeaPrandtltip/rootlosscorrectionornotwiththenewlyderivedaerofoil
data is contradicting and subject to dispute. On one hand in the freewake model a 2D
definitionfortheangleofattackwasusedsincetheeffectofthewakewasremoved.This
suggests that the tip/root loss should be included. On theother hand it should be keptin
mindthatthemeasuredvaluesofCnandCtfromwhichthenewaerofoildatawasderived
are3Dvalueswhichalreadyincludetip/rooteffects.Forthisreasonitcouldbearguedthat
thetip/rootlossfactorshouldnotbeintroduced.Inthisstudy,theBEMcalculationswere
thereforeperformedbothwithandwithoutPrandtltip/rootlosscorrectionfactor.
ForthecomputationswiththeBEM1configuration(seetable6.2),the2Dstaticwindtunnel
aerofoil data was extrapolated for high angles of attack using the ViternaCorrigan
empiricalmodelwhichwasdescribedearlierinChapter3,section3.5.The2DDelftWind
363
Chapter6AerodynamicAnalysisoftheNRELPhaseVIWindTurbine
1.4
Viterna Corrigan Cd
1
Cl or Cd
0.8
0.6
0.4
0.2
0
0 5 10 15 20 25 30 35 40 45
(deg)
Figure6.502DstaticaerofoildatafromDelftwindtunnel(forRe=1000000)extrapolatedforhigh
anglesofattackusingtheViternaCorriganmodel.
tunnelaerofoildataextrapolatedusingtheaboveViternaCorriganmodelisplottedinFig.
6.50.
A. Comparisonofspanwisedistributionsofaerodynamicloadingandinductionfactors
TheresultsfromthethreedifferentBEMcomputations(BEM1,BEM2andBEM3,seetable
6.2)fordifferentwindspeedsoftheNRELrotor(U=5,10,15,20and25m/s)arepresentedin
Figs. 6.51, 6.52, 6.53, 6.54 and 6.55, respectively. For each windspeed, the spanwise
variationsofnormalandtangentialaerodynamicloading(dAanddA)fromHAWT_BEM
are compared with the corresponding values obtained by NREL using pressure
measurementsintheNASAAmeswindtunnel.ItiseasilynotedfromFigs.6.516.55(a)and
(b) that the BEM results for dA and dA improve significantly when the new 3D aerofoil
data is used instead of the 2D aerofoil data with the ViternaCorrigan extrapolation
(compare results of BEM1 and BEM2). This improvement is observed at all wind speeds,
even at the extreme windspeed of 25m/s. This proves that given the appropriate aerofoil
data, the BEM theory can model aerodynamic loading augmentation at the inboard blade
sectionsreallywell.AnimportantcommenttomakeconcernsthePrandtltip/rootloss:the
comparisonobtainedfromBEM2andBEM3resultsrevealsthattheeffectofincludingthe
Prandtl tip/root loss correction is very small. This consolidates the reason that this
correctionisnotveryimportantanymorebecausethenewaerofoildatacanhandletiploss
correction. This is a consequence of the fact that the loading parameters Cn and Ct from
whichthenewaerofoildatawasoriginallyderivedalreadyaccountsfortipandrootloss.
364
Chapter6AerodynamicAnalysisoftheNRELPhaseVIWindTurbine
Figs. 6.516.55(c) and (d) display the spanwise variations of the azimuthally (annular)
averaged axial induction factor (a1) and the axial induction factor at the blade lifting line
(a1,c) obtained by BEM. These are compared with those obtained by the freewake model
(HAWT_FWC) (which were already shown earlier in Figs. 6.21). Despite the fact that the
aerodynamicloadingdistributionsimprovedconsiderablyatallwindspeedswhenthenew
aerofoildatawasused,noimprovementwasobservedintheBEM2andBEM3predictions
fora1anda1,cathighwindspeeds(U=10,15,20and25m/s)atradiallocationswherethelocal
angleofattackisknowntobelarge.Thisislikelytooriginatefromtheinherentlimitation
oftheliftinglinemodelinHAWT_FWCusedtoderivetheaerofoildataandinductionfrom
thebladepressuremeasurementsatlargeanglesofattackandnotduetotheBEMtheory
itself. In fact, in the work of Johansen et al. [40] where new aerofoil data was derived for
BEM from CFD computations (see page 312), a good agreement was achieved when
comparingtheBEMpredictionsfora1withthosefromCFD.
Asregardsthespanwisevariationsofa1,c,theBEM1predictionsyieldanegativeriseinthis
parameter at the blade tip, which is far different from that predicted by the freewake
vortexmodel.ThisisalimitationofthePrandtlcorrectionandwasalreadydiscussedwhen
modellingtheTUDelftrotor(section4.4.2).Withtheinclusionofthenewaerofoildatathis
negativeriseatthebladetipisnolongerpresent.
Figs. 6.516.55 show that the Prandtl tip/root correction has a notable influence on
distributions for a1,c, especially at high windspeeds and at the blade tip and root regions.
This influence on the aerodynamic loads is however only marginal because the induction
factorsaregenerallysmall.
365
Chapter6AerodynamicAnalysisoftheNRELPhaseVIWindTurbine
160
140
120
100
dA (N/m)
80
60
NREL
40 BEM1
Fig. (a) BEM2
20
BEM3
0
0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
r/R
6
dA (N/m)
2 NREL
BEM1
1 BEM2
BEM3
Fig. (b)
0
0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
-1
r/R
0.05
0
0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
-0.05 Free-Wake
BEM1
BEM2
a1
-0.1 BEM3
-0.15
Fig. (c)
-0.2
-0.25
r/R
0.05
0
0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
-0.05
Free-Wake
BEM1
-0.1 BEM2
BEM3
a1,c
-0.15
-0.2
-0.25
Fig. (d)
-0.3
-0.35
r/R
Figure6.51ComparisonofBEMpredictionsforU=5m/s,=00.
366
Chapter6AerodynamicAnalysisoftheNRELPhaseVIWindTurbine
350
300
250
dA (N/m)
200
150
100
NREL
BEM1
Fig. (a) BEM2
50
BEM3
0
0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
r/R
50
40
30
dA (N/m)
20
NREL
10
BEM1
BEM2
BEM3
Fig. (b)
0
0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
-10
r/R
0.05
0
0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
-0.05
a1
-0.1
Fig. (c)
Free-Wake
-0.15 BEM1
BEM2
BEM3
-0.2
r/R
0.05
0
0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
-0.05
-0.1
a1,c
-0.15
-0.2
Free-Wake
-0.25
Fig. (d)
BEM1
BEM2
-0.3 BEM3
-0.35
r/R
Figure6.52ComparisonofBEMpredictionsforU=10m/s,=00.
367
Chapter6AerodynamicAnalysisoftheNRELPhaseVIWindTurbine
400
350
300
250
dA (N/m)
200
150
NREL
100 BEM1
Fig. (a) BEM2
50
BEM3
0
0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
r/R
70
NREL
60
50
BEM1
BEM2
40 BEM3
dA (N/m)
30
20
10
0
Fig. (b) 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
-10
-20
r/R
0.05
0
0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
-0.05
a1
-0.1
-0.2
r/R
0.05
0
0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
-0.05
-0.1
a1,c
-0.15
Free-Wake
-0.2
Fig. (d) BEM1
BEM2
-0.25 BEM3
-0.3
r/R
Figure6.53ComparisonofBEMpredictionsforU=15m/s,=00.
368
Chapter6AerodynamicAnalysisoftheNRELPhaseVIWindTurbine
600
500
400
dA (N/m)
300
NREL
200
BEM1
Fig. (a) 100 BEM2
BEM3
0
0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
r/R
70
60
50
40
30
NREL
dA (N/m)
BEM1
20
BEM2
10 BEM3
0
0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
Fig. (b) -10
-20
-30
r/R
0.05
0
0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
-0.05
a1
-0.1
Fig. (c)
Free-Wake
BEM1
-0.15
BEM2
BEM3
-0.2
r/R
0.05
0
0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
-0.05
a1,c
-0.1
-0.15 Free-Wake
BEM1
Fig. (d) BEM2
-0.2
BEM3
-0.25
r/R
Figure6.54ComparisonofBEMpredictionsforU=20m/s,=00.
369
Chapter6AerodynamicAnalysisoftheNRELPhaseVIWindTurbine
900
800
NREL
BEM1
700
BEM2
600 BEM3
dA (N/m)
500
400
300
200
Fig. (a)
100
0
0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
r/R
80
60
40 NREL
dA (N/m)
BEM1
20 BEM2
BEM3
0
0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
Fig. (b)
-20
-40
r/R
0.05
0
0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
-0.05
a1
-0.1
Fig. (c)
Free-Wake
-0.15
BEM1
BEM2
BEM3
-0.2
r/R
0.05
0
0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
-0.05
a1,c
-0.1
-0.15 Free-Wake
BEM1
Fig. (d) BEM2
-0.2 BEM3
-0.25
r/R
Figure6.55ComparisonofBEMpredictionsforU=25m/s,=00.
370
Chapter6AerodynamicAnalysisoftheNRELPhaseVIWindTurbine
B.ComparisonofGloballoads
Figs.6.56(a),(b)and(c)comparetheaerodynamicgloballoads(lowspeedshafttorqueand
thebladerootflapandedgemoments)predictedbythethreedifferentBEMcomputations
(BEM1,BEM2andBEM3)forwindspeedsU=5,7,10,13,15,20and25m/s.Theseplotsalso
show the results shown in Fig. 6.22, i.e. the global loads derived directly from the
measurements using Eqts. 6.4, 6.5 and 6.6 (denoted by Exp) and those predicted by
HAWT_FWC (denoted by Freewake). At low windspeeds (U= 5, 7, 10m/s), the BEM
predictions with 2D static aerofoil data (BEM1 computations) are within acceptable
accuracy. However this is not the case at high windspeeds and this is due to the
inconsistencies of the ViternaCorrigan model. The radical improvement in the BEM
predictions when using the new aerofoil data (BEM2 and BEM3 computations) is easily
notedinFigs.6.56(a),(b)and(c).WhethertoincludethePrandtltip/rootlosscorrectionor
not has negligible influence on the calculation of global loads since only very small
differencesareseenbetweentheBEM2andBEM3results.
2100
Torque
1800
1500
LSSTQ (Nm)
1200
900
600 Free-Wake
Exp
BEM1
300
BEM2
BEM3
0
4 6 8 10 12 14 16 18 20 22 24 26
U (m/s)
Figure6.56(a)Variationlowspeedshafttorquewithwindspeed.
371
Chapter6AerodynamicAnalysisoftheNRELPhaseVIWindTurbine
5000
4500 Flap
Moment
4000
3500
RFM (Nm)
3000
2500
2000
Free-Wake
1500
Exp
BEM1
1000
BEM2
BEM3
500
4 6 8 10 12 14 16 18 20 22 24 26
U (m/s)
Figure6.56(b)Variationofbladerootflapmomentwithwindspeed.
800
700
Edge
600 Moment
500
400
REM (Nm)
300
200
100
Free-Wake
0 Exp
4 6 8 10 12 14 16 18 20 22 24 26 BEM1
-100 BEM2
BEM3
-200
U (m/s)
Figure6.56(c)Variationofbladerootedgemomentwithwindspeed.
372
Chapter6AerodynamicAnalysisoftheNRELPhaseVIWindTurbine
6.4.2YawedConditions
Asforaxialconditions,thenewlyderivedaerofoildataforunsteadyconditionsinyawwas
also interpolated to be able to use it for the BEM calculations. Strictly speaking, unsteady
aerofoil data for 3D rotor applications should be interpolated with variables r/R, ,
{ }
, V andk,i.e. Cl , Cd r , , , V , k .Butsincetheonlyscopeforthisanalysiswasto
R
assesstheimprovementinBEMpredictionsonlyratherthandevelopinganewengineering
model,thenitwaspossibletosimplyuseadoubleinterpolationusingonlytwovariables
r/R and , i..e., Cl , Cd { r R , } . A double interpolation was performed separately at each
windspeed. The variation of Cl and Cd with r/R at any was fitted with a cubic spline
interpolation(seeAppendixD)withtheboundaryconditionsthatatthebladetipandroot
bothClandCdarezero.
To derive the inflow correction factor (denoted by Fsa) to account for the uneven induced
velocityattherotordiskinyawthefollowingmethodwasused:recallfromEqt.3.27(see
section3.5)thatFsacorrectsonlytheaxialinducedvelocityatthebladeliftinglines.Making
FsainEqt.3.27subjectoftheformulayields:
u
Fsa = a ,c f
ua
(6.8)
For each wind speed (U= 5, 7, 10, 13 and 15m/s at =300), parameters ua,c and ua at each
blade azimuth angle () and radial location (r/R) were derived from the induced velocity
distributions obtained from HAWT_FWC (Figs. 6.38 and 6.39). The Prandtl tip/root loss
factor f could also be derived from HAWT_FWC, from knowledge of the inflow angle, ,
andusingEqts.3.10.TheresultsforFsaatU=5,10and15m/sareplottedinFig.6.57.Acubic
double spline interpolation was then done to obtain the inflow correction factor Fsa as a
functionofeachbladeazimuthangleandradiallocation,i.e. Fsa ( r R , ) .
The BEM computations were performed using different aerofoil data and with/without
inflow corrections. The different BEM configurations are listed in table 6.3. They were
selected in such a way that it was possible to determine quantitatively whether
improvementisdueto:
a.theuseof3Dstaticaerofoildataalone(bycomparingBEM1andBEM2computations)
b.theeffectofincludingunsteadyeffects(includingdynamicstall)byhaving3Dunsteady
aerofoildatainsteadof3Dstaticdata(bycomparingBEM2andBEM3computations)
c.theColemanscorrection(seeEqts.3.23,3.24andtable3.1)whileatthesametimehaving
3Dunsteadyaerofoildata(comparingBEM3andBEM4)
d.replacingColemanscorrectionwiththenewlyderivedcorrections(Figs.6.57)obtained
fromHAWT_FWC(comparingBEM4andBEM5)
373
Chapter6AerodynamicAnalysisoftheNRELPhaseVIWindTurbine
2.5
2.0
1.5
Fig.(a):U=5m/s
1.0
Fsa
0.5
r/R = 0.3
0.0 r/R = 0.47
0 30 60 90 120 150 180 210 240 270 300 330 360
r/R = 0.63
-0.5
r/R = 0.8
r/R = 0.95
-1.0
(deg)
2.5
2.0
Fig.(b):U=10m/s
1.5
Fsa
r/R = 0.95
0.0
0 30 60 90 120 150 180 210 240 270 300 330 360
(deg)
3.0
2.5
2.0
Fig.(c):U=15m/s
1.5
Fsa
1.0
Figure 6.57 Variations of the inflow correction factor Fsa derived by HAWT_FWC with blade
azimuthangleandradiallocation.
374
Chapter6AerodynamicAnalysisoftheNRELPhaseVIWindTurbine
InallthedifferentBEMcomputations,thePrandtltip/rootlosscorrectionwasused.When
usingthenewinflowcorrections(intheBEM5configuration),thislosscorrectionhadtobe
includedbecauseitseffectwasaccountedforwhenderivingthevaluesforFsa(seeEqt.6.8).
Table6.3:ComputationsperformedonNRELrotorfor=300usingHAWT_BEM
Computation BEMCodeConfiguration
Reference
BEM1 2D static aerofoil data (from Delft wind tunnel) extrapolated
usingViternaCorriganmodel(Fig.6.50)
CorrectionforinflowinyawNOTincluded
Prandtltip/rootcorrectionfactorincluded
BEM2 New3DstaticaerofoildataobtainedfromNRELbladepressure
measurementandHAWT_FWCforaxialconditions(Figs.6.19)
CorrectionforinflowinyawNOTincluded
Prandtltip/rootcorrectionfactorincluded
BEM3 New 3D unsteady aerofoil data obtained from NREL blade
pressure measurement and HAWT_FWC for yawed conditions
(Figs.6.33)
CorrectionforinflowinyawNOTincluded
Prandtltip/rootcorrectionfactorincluded
BEM4 New 3D unsteady aerofoil data obtained from NREL blade
pressure measurement and HAWT_FWC for yawed conditions
(Figs.6.33)
Colemancorrectionforinflowinyawincluded
Prandtltip/rootcorrectionfactorincluded
BEM5 New 3D unsteady aerofoil data obtained from NREL blade
pressure measurement and HAWT_FWC for yawed conditions
(Figs.6.33)
New inflow corrections for inflow in yaw obtained from NREL
blade pressure measurement and HAWT_FWC for yawed
conditions(Figs.6.57)
Prandtltip/rootcorrectionfactorincluded
375
Chapter6AerodynamicAnalysisoftheNRELPhaseVIWindTurbine
A. Comparisonofinductionfactordistributions
Figs. 6.58, 6.59 and 6.60 compare the distributions for the azimuthallyaveraged axial
induction factor (a1) for the different BEM computations (BEM1, BEM2, BEM3, BEM4 and
BEM5) with those from the freewake model HAWT_FWC (from Fig. 6.38) at windspeeds
U=5,10and15m/s.Ineachplot,thevariationofa1againstrotorazimuthangle()isplotted
at radial locations r/R=0.3, 0.47, 0.63, 0.8 and 0.95. In Figs. 6.61, 6.62 and 6.63, the axial
induction factor at the blade lifting line (a1,c) is shown. The predictions from HAWT_FWC
for a1,c from Figs. 6.39 are included. Figs. 6.58, 6.59 and 6.60 show that, in most cases, the
results for a1 from all five BEM computations are in close agreement with those from the
freewake model. Nonetheless the results for a1,c from the BEM computations differ
considerablyfromeachother,asnotedinFigs.6.61,6.62and6.63.Recallfromtable6.3that
the first three BEM computations (BEM1, BEM2 and BEM3) have no inflow correction to
accountfortheuneveninductionattherotorplanecreatedbytheskewedhelicalwake.The
results for a1,c from these three BEM computations are very much different from those
predictedbytheHAWT_FWC.Thisprovidesnecessaryevidencethattheinflowcorrection
in BEM computations is necessary to obtain a more realistic induction distribution at the
blades.TheBEM4computationincludestheColemancorrectionfortheinflowbutinmost
cases the predictions for a1,c are still not in good agreement with those of the freewake
model(seeFigs.6.61,6.62and6.63).TheColemancorrectionyieldsgoodagreementonlyat
U=5m/sandattheoutboardbladesectionsr/R=0.8and0.95.Thebestpredictionsfora1,care
those from the BEM5 computations since the inflow corrections used here were directly
derivedfromthefreewakevortexmodelitself.
B. Comparisonofaerodynamicloadingdistributions
The BEM results for the normal and tangential loading on the blades are now compared.
The results for the normal loading (dA) are plotted against blade azimuth angle () for
windpseedsU=5,10and15m/sinFigs.6.64,6.65and6.66,respectively.Thecorresponding
results for the tangential loading (dA) are given in Figs. 6.67, 6.68 and 6.69. The
experimental results obtained from the blade pressure measurements are included. It is
shownthatlittleimprovementintheloadingpredictionsisachievedbyBEMwhenusing
the new 3D static aerofoil data obtained for axial conditions instead of the 2D static data
(compareBEM1andBEM2results).Howeveraradicalimprovementinthecorrelationwith
the experimental data is achieved when introducing the new unsteady 3D aerofoil data
(compare BEM2 and BEM3 results). This provides evidence that with the use of reliable
unsteady 3D aerofoil data, BEM predictions in yawed conditions can be significantly
improved. The BEM4 and BEM5 results for loading differ only marginally from those of
BEM3. This explains the fact that the influence of the inflow correction is very small and
that the reason for this is that for the operating conditions being studied, the induction
factors are small. It is expected that for other operating conditions, especially those for
376
Chapter6AerodynamicAnalysisoftheNRELPhaseVIWindTurbine
which the tip speed ratio is high, the axial induction factors are larger and in such
circumstancestheinflowcorrectionbecomesmoreimportant.
C. Comparisonofgloballoads
Figs. 6.70, 6.71 and 6.72 compare the lowspeed shaft torque and the blade root flap and
edgemomentsfromthefivedifferentBEMcomputationswiththosederivedfromtheblade
pressure measurements using Eqts. 6.4 6.6. It can be noted that the use of the new 3D
unsteadyaerofoildatayieldsbestcorrelationwiththeexperimentaldata.Theinclusionof
the inflow correction due to yaw leaves only a small influence on the calculated global
loads,asalreadyobservedintheloadingresultsofFigs.6.646.69.
377
Chapter6AerodynamicAnalysisoftheNRELPhaseVIWindTurbine
0 0
0 60 120 180 240 300 360 0 60 120 180 240 300 360
-0.02
-0.01
-0.04
-0.02
a1
-0.06
-0.03
a1
-0.08
0 0
0 60 120 180 240 300 360 0 60 120 180 240 300 360
-0.02 -0.02
-0.04 -0.04
-0.06 -0.06
-0.08 -0.08
a1
a1
-0.1 -0.1
-0.12 HAWT_FWC -0.12 HAWT_FWC
-0.14 BEM1
BEM2 -0.14
BEM1
BEM2
-0.16 BEM3
BEM4 -0.16
BEM3
BEM4
-0.18 BEM5
-0.18
BEM5
(deg) (deg)
0
0 60 120 180 240 300 360
-0.02
-0.04
-0.06
a1
-0.08
HAWT_FWC
-0.1
BEM1
BEM2
-0.12 BEM3
BEM4
-0.14
BEM5
(deg)
Figure6.58Variationsoftheazimuthallyaveragedaxialinductionfactorwithbladeazimuthangle
atU=5m/s,=300.
378
Chapter6AerodynamicAnalysisoftheNRELPhaseVIWindTurbine
0 0
0 60 120 180 240 300 360 0 60 120 180 240 300 360
-0.02 -0.02
-0.04
-0.04
-0.06
a1
a1
-0.06
-0.08
HAWT_FWC -0.08 HAWT_FWC
-0.1 BEM1
BEM2
BEM1
BEM2
BEM3 -0.1 BEM3
-0.12 BEM4 BEM4
-0.14 BEM5
-0.12
BEM5
(deg) (deg)
a1
-0.06 -0.05
-0.06
-0.08 HAWT_FWC HAWT_FWC
-0.07
BEM1 BEM1
BEM2 -0.08 BEM2
-0.1 BEM3 BEM3
BEM4 -0.09 BEM4
BEM5 BEM5
-0.12 -0.1
(deg) (deg)
Fig. (c): r/R=0.63 Fig. (d): r/R=0.8
0
0 60 120 180 240 300 360
-0.01
-0.02
-0.03
-0.04
a1
-0.05
-0.06 HAWT_FWC
BEM1
-0.07 BEM2
BEM3
-0.08 BEM4
BEM5
-0.09
(deg)
Figure6.59Variationsoftheazimuthallyaveragedaxialinductionfactorwithbladeazimuthangle
atU=10m/s,=300.
379
Chapter6AerodynamicAnalysisoftheNRELPhaseVIWindTurbine
0 0
0 60 120 180 240 300 360 0 60 120 180 240 300 360
-0.02 -0.02
-0.04
-0.04
-0.06
a1
-0.08
a1
-0.06
-0.1
HAWT_FWC -0.08 HAWT_FWC
-0.12 BEM1 BEM1
BEM2 BEM2
-0.14 BEM3 -0.1 BEM3
BEM4 BEM4
-0.16 BEM5
-0.12
BEM5
(deg) (deg)
0 0
0 60 120 180 240 300 360 0 60 120 180 240 300 360
-0.01 -0.01
-0.02
-0.02
-0.03
-0.04
a1
a1
-0.03
0
0 60 120 180 240 300 360
-0.005
-0.01
-0.015
-0.02
a1
-0.025
-0.03
HAWT_FWC
-0.035 BEM1
BEM2
-0.04
BEM3
-0.045 BEM4
BEM5
-0.05
(deg)
Figure6.60Variationsoftheazimuthallyaveragedaxialinductionfactorwithbladeazimuthangle
atU=15m/s,=300.
380
Chapter6AerodynamicAnalysisoftheNRELPhaseVIWindTurbine
0.04 -0.06
0 60 120 180 240 300 360
0.02
-0.08
0
0 60 120 180 240 300 360 -0.1
-0.02
-0.04
a1,c
a1,c
-0.12
-0.06
HAWT_FWC -0.14 HAWT_FWC
-0.08 BEM1 BEM1
BEM2 BEM2
-0.1 BEM3 -0.16 BEM3
BEM4 BEM4
-0.12 BEM5
-0.18
BEM5
(deg) (deg)
-0.1 -0.1
0 60 120 180 240 300 360 0 60 120 180 240 300 360
-0.12 -0.12
-0.14 -0.14
a1,c
a1,c
-0.16 -0.16
0
0 60 120 180 240 300 360
-0.05
-0.1
a1,c
-0.15
HAWT_FWC
BEM1
-0.2 BEM2
BEM3
BEM4
BEM5
-0.25
(deg)
Figure 6.61 Variations of the axial induction factor at the blade lifting line with blade azimuth
angleatU=5m/s,=300.
381
Chapter6AerodynamicAnalysisoftheNRELPhaseVIWindTurbine
0 0
0 60 120 180 240 300 360 0 60 120 180 240 300 360
-0.02
-0.05
-0.04
-0.06
-0.1
-0.08
a1,c
a1,c
-0.1
-0.15
HAWT_FWC -0.12 HAWT_FWC
BEM1 BEM1
-0.2 BEM2 -0.14 BEM2
BEM3
-0.16
BEM3
BEM4 BEM4
-0.25 BEM5
-0.18
BEM5
(deg) (deg)
0 0
0 60 120 180 240 300 360 0 60 120 180 240 300 360
-0.02
-0.02
-0.04
-0.04
-0.06
-0.06
-0.08
a1,c
a1,c
-0.08
-0.1
-0.1 HAWT_FWC -0.12 HAWT_FWC
-0.12 BEM1
BEM2 -0.14
BEM1
BEM2
-0.14 BEM3
BEM4 -0.16
BEM3
BEM4
-0.16 BEM5
-0.18
BEM5
(deg) (deg)
0
0 60 120 180 240 300 360
-0.05
-0.1
a1,c
-0.15
HAWT_FWC
BEM1
-0.2 BEM2
BEM3
BEM4
-0.25
BEM5
(deg)
Figure 6.62 Variations of the axial induction factor at the blade lifting line with blade azimuth
angleatU=10m/s,=300.
382
Chapter6AerodynamicAnalysisoftheNRELPhaseVIWindTurbine
0 0.1
0 60 120 180 240 300 360
-0.05 0.05
-0.1 0
0 60 120 180 240 300 360
-0.15 -0.05
a1,c
a1,c
-0.2 -0.1
HAWT_FWC HAWT_FWC
-0.25 BEM1 -0.15 BEM1
BEM2 BEM2
-0.3 BEM3 -0.2 BEM3
BEM4 BEM4
-0.35 BEM5
-0.25
BEM5
(deg) (deg)
Fig. (a): r/R=0.3 Fig. (b): r/R=0.47
0 0.04
0 60 120 180 240 300 360
0.02
-0.02
0
-0.02
0 60 120 180 240 300 360
-0.04
-0.04
a1,c
a1,c
-0.06 -0.06
-0.08
-0.08 HAWT_FWC HAWT_FWC
BEM1 -0.1
BEM2
BEM1
BEM2
-0.12
-0.1
BEM3
BEM4 -0.14
BEM3
BEM4
-0.12 BEM5
-0.16
BEM5
(deg) (deg)
Fig. (c): r/R=0.63 Fig. (d): r/R=0.8
0
0 60 120 180 240 300 360
-0.02
-0.04
-0.06
a1,c
-0.08
-0.1
HAWT_FWC
-0.12 BEM1
BEM2
-0.14
BEM3
BEM4
-0.16
BEM5
(deg)
Fig. (e): r/R=0.95
Figure 6.63 Variations of the axial induction factor at the blade lifting line with blade azimuth
angleatU=15m/s,=300.
383
Chapter6AerodynamicAnalysisoftheNRELPhaseVIWindTurbine
40 65
30
60
20
dA (N/m)
55
dA (N/m)
10
NREL
NREL 50
BEM1
BEM1
0
BEM2
0 60 120 180 240 300 360 BEM2
45 BEM3
BEM3
-10
BEM4
BEM4
BEM5
40
-20 BEM5
0 60 120 180 240 300 360
(deg) (deg)
100 120
95 115
90 110
85 105
dA (N/m)
dA (N/m)
80 100
NREL NREL
75 95
BEM1 BEM1
BEM2 BEM2
70 90
BEM3 BEM3
65 BEM4 85 BEM4
BEM5 BEM5
60 80
0 60 120 180 240 300 360 0 60 120 180 240 300 360
(deg) (deg)
120
110
100
dA (N/m)
90
NREL
80 BEM1
BEM2
70
BEM3
BEM4
BEM5
60
0 60 120 180 240 300 360
(deg)
Figure6.64VariationsofthenormalloadingatthebladeliftinglinewithbladeazimuthangleatU
=5m/s,=300.
384
Chapter6AerodynamicAnalysisoftheNRELPhaseVIWindTurbine
160 240
220
140
120 200
180
100
dA (N/m)
160
dA (N/m)
80 140
NREL NREL
60 BEM1
120
BEM1
40 BEM2 100 BEM2
BEM3 80 BEM3
20 BEM4 60 BEM4
0 BEM5
40
BEM5
0 60 120 180 240 300 360 0 60 120 180 240 300 360
(deg) (deg)
Fig. (a): r/R=0.3 Fig. (b): r/R=0.47
350 390
300 340
250
290
dA (N/m)
dA (N/m)
200
240
NREL NREL
150 BEM1 BEM1
190
100 BEM2 BEM2
BEM3 BEM3
140
50 BEM4 BEM4
BEM5 BEM5
0 90
0 60 120 180 240 300 360 0 60 120 180 240 300 360
(deg) (deg)
450
400
350
dA (N/m)
300
NREL
250
BEM1
200 BEM2
BEM3
150
BEM4
100
BEM5
Figure6.65VariationsofthenormalloadingatthebladeliftinglinewithbladeazimuthangleatU
=10m/s,=300.
385
Chapter6AerodynamicAnalysisoftheNRELPhaseVIWindTurbine
350 440
300 390
340
250 290
dA (N/m)
dA (N/m)
200 240
NREL NREL
150 BEM1 190 BEM1
100 BEM2
140
BEM2
BEM3 BEM3
50 BEM4 90 BEM4
0
BEM5
40
BEM5
0 60 120 180 240 300 360 0 60 120 180 240 300 360
(deg) (deg)
Fig. (a): r/R=0.3 Fig. (b): r/R=0.47
490
450
400 440
350 390
300 340
dA (N/m)
dA (N/m)
250 290
200 NREL NREL
BEM1 240 BEM1
150
BEM2
190
BEM2
100 BEM3 BEM3
50 BEM4 140 BEM4
BEM5
0
BEM5
90
0 60 120 180 240 300 360 0 60 120 180 240 300 360
(deg) (deg)
400
350
dA (N/m)
300
250
NREL
200
BEM1
150
BEM2
100 BEM3
50 BEM4
BEM5
0
0 60 120 180 240 300 360
(deg)
Figure6.66VariationsofthenormalloadingatthebladeliftinglinewithbladeazimuthangleatU
=15m/s,=300.
386
Chapter6AerodynamicAnalysisoftheNRELPhaseVIWindTurbine
4 4
3 3
2 2
dA (N/m)
dA (N/m)
1 1
NREL NREL
0 BEM1 0 BEM1
0 60 120 180 240 300 360 BEM2 0 60 120 180 240 300 360 BEM2
-1 BEM3
-1
BEM3
BEM4 BEM4
BEM5 BEM5
-2 -2
(deg) (deg)
Fig. (a): r/R=0.3 Fig. (b): r/R=0.47
5 5
4 4
3 3
dA (N/m)
dA (N/m)
2 2
NREL NREL
1 BEM1 1 BEM1
BEM2 BEM2
0 BEM3
0
BEM3
0 60 120 180 240 300 360 BEM4 0 60 120 180 240 300 360 BEM4
BEM5 BEM5
-1 -1
(deg) (deg)
1.5
0.5
dA (N/m)
0
0 60 120 180 240 300 360 NREL
-0.5
BEM1
-1 BEM2
BEM3
-1.5 BEM4
BEM5
-2
(deg)
Fig. (e): r/R=0.95
Figure6.67Variationsofthetangentialloadingatthebladeliftinglinewithbladeazimuthangleat
U=5m/s,=300.
387
Chapter6AerodynamicAnalysisoftheNRELPhaseVIWindTurbine
40 50
35
30 40
25 30
20
dA (N/m)
dA (N/m)
15 20
NREL NREL
10
5 BEM1 10 BEM1
BEM2 BEM2
0 BEM3
0
BEM3
0 60 120 180 240 300 360 BEM4
-5 0 60 120 180 240 300 360 BEM4
BEM5 BEM5
-10 (deg)
-10
(deg)
Fig. (a): r/R=0.3 Fig. (b): r/R=0.47
50 50
45
45
40
35
40
dA (N/m)
dA (N/m)
30
25 35
NREL NREL
20 BEM1 BEM1
30
15 BEM2 BEM2
10 BEM3 BEM3
25
5 BEM4 BEM4
BEM5 BEM5
0 20
0 60 120 180 240 300 360 0 60 120 180 240 300 360
(deg) (deg)
25
20 NREL
BEM1
15 BEM2
BEM3
10
BEM4
BEM5
5
0 60 120 180 240 300 360
(deg)
388
Chapter6AerodynamicAnalysisoftheNRELPhaseVIWindTurbine
100
90
80
70 60
50
dA (N/m)
dA (N/m)
40
20
NREL NREL
30 BEM1 BEM1
BEM2 0 BEM2
BEM3
0 60 120 180 240 300 360
BEM3
10
-20
BEM4 BEM4
0 60 120 180 240 300 360 BEM5 BEM5
-10 -40
(deg) (deg)
Fig. (a): r/R=0.3 Fig. (b): r/R=0.47
100
80
80
60
60
dA (N/m)
dA (N/m)
40
40
NREL NREL
20 BEM1 20 BEM1
BEM2 BEM2
0 BEM3
0
BEM3
0 60 120 180 240 300 360 BEM4
BEM4 0 60 120 180 240 300 360
BEM5 BEM5
-20 -20
(deg) (deg)
Fig. (c): r/R=0.63 Fig. (d): r/R=0.8
70
60
50
40
dA (N/m)
30
NREL
20
BEM1
10 BEM2
BEM3
0 BEM4
0 60 120 180 240 300 360 BEM5
-10
(deg)
Fig. (e): r/R=0.95
Figure6.69Variationsofthetangentialloadingatthebladeliftinglinewithbladeazimuthangleat
U=15m/s,=300.
389
Chapter6AerodynamicAnalysisoftheNRELPhaseVIWindTurbine
215
210
205
LSSTQ (Nm)
Fig.(a):U=5m/s 200 NREL
195 BEM1
190 BEM2
BEM3
185
BEM4
180
BEM5
175
0 60 120 180 240 300 360
(deg)
1300
1200
Fig.(b):U=10m/s
LSSTQ (Nm)
1100 NREL
BEM1
1000 BEM2
BEM3
900
BEM4
BEM5
800
0 60 120 180 240 300 360
(deg)
1800
1600
Fig.(c):U=15m/s
LSSTQ (Nm)
1400 NREL
BEM1
1200 BEM2
BEM3
1000
BEM4
BEM5
800
0 60 120 180 240 300 360
(deg)
Figure6.70VariationofthelowspeedshafttorquewithbladeazimuthangleatU=5,10,15m/s,
=300.
390
Chapter6AerodynamicAnalysisoftheNRELPhaseVIWindTurbine
900
850
800
Fig.(a):U=5m/s NREL
BEM2
700
BEM3
650 BEM4
BEM5
600
0 60 120 180 240 300 360
(deg)
3000
2600
2200
Fig.(b):U=10m/s NREL
RFM (Nm)
1800 BEM1
BEM2
1400
BEM3
1000 BEM4
BEM5
600
0 60 120 180 240 300 360
(deg)
4100
3600
3100
NREL
RFM (Nm)
Fig.(c):U=15m/s 2600
BEM1
2100
BEM2
1600 BEM3
BEM4
1100
BEM5
600
0 60 120 180 240 300 360
(deg)
Figure6.71VariationofthebladerootflapmomentwithbladeazimuthangleatU=5,10,15m/s,
=300.
391
Chapter6AerodynamicAnalysisoftheNRELPhaseVIWindTurbine
70
65
60
55
Fig.(a):U=5m/s NREL
50
40 BEM2
35 BEM3
30
BEM4
25
BEM5
20
0 60 120 180 240 300 360
(deg)
600
550
500
400 BEM1
350 BEM2
300 BEM3
BEM4
250
BEM5
200
0 60 120 180 240 300 360
(deg)
1000
900
800
700
Fig.(c):U=15m/s NREL
600
REM (Nm)
500 BEM1
400 BEM2
300
BEM3
200
BEM4
100
BEM5
0
0 60 120 180 240 300 360
(deg)
Figure6.72VariationofthebladerootedgewithbladeazimuthangleatU=5,10,15m/s,=300.
392
Chapter6AerodynamicAnalysisoftheNRELPhaseVIWindTurbine
6.5Conclusions
Inthischapter,anovelapproachwaspresentedforusingafreewakevortexmodelfor
finding the angle of attack variations, inflow distribution at the rotorplane and wake
geometry for the cases were the experimental data consists only of blade pressure
measurements. This made it possible to carry out a deeper investigation of rotor
aerodynamic phenomena and derive new aerofoil data and inflow corrections that would
beveryusefulindevelopingbetterengineeringmodelsforBEMcodes.TheNRELPhaseVI
UAEexperimentaldatawasusedforthiscasestudyandbothaxialandyawedconditions
(=00 and 300) were studied. The wind tunnel measurements were readily available from
NREL and no further measurements were required. New 3D aerofoil data were derived
from the blade pressure measurements and the angle of attack distributions predicted by
the freewake vortex model. This data was compared with 2D static wind tunnel data for
the S809 aerofoil. For axial conditions in which the local angle of attack is steady, two
conclusionscouldbemade:
(1) forsmallanglesofattack(lessthanthe2Dstaticstallingangle),thenew3Daerofoil
dataisveryclosetothe2Daerofoildataexceptatthetipandrootregion.Inthelatter
regions,thenew3Dliftcoefficientwasfoundtobelowerthanthe2Daerofoildata.
(2) for larger angles of attack, the new 3D aerofoil data was considerably different than
the 2D data and the presence of stalldelay was evident, especially at the inboard
sectionsoftheblades.Itwasalsoobservedthattheincreaseintheliftcoefficientdue
tostalldelayisaccompaniedbyacorrespondingincreaseinthedragcoefficient.
For yawed conditions, the local angle of attack was unsteady and a function of the blade
azimuthangle.Thereforethenew3Dderivedaerofoildataforyawwasalsounsteady.This
data was compared with the 2D static wind tunnel aerofoil data as well as with the new
steady3Ddataobtainedfromaxialconditions.Considerablehysterisisinboththeliftand
dragwasfound,especiallyathighanglesofattackwheredynamicstalltakesplace.Itwas
foundthatthe3Dunsteadycoefficientscouldwellexceedthecorrespondingsteadyvalues
derived for axial conditions. It could also be concluded that the unsteady Cl and Cdp
hysterises loops were in general closer to the 3D steady values than to the corresponding
2Dsteadydata.Thiswasobservedatbothattachedandstallconditionsanditimpliesthat
3Deffects(inparticularstalldelay)haveasignificantroleintheunsteadyflowbehaviour
ofwindturbineblades.
The computed freewake geometries were plotted together with the corresponding wake
circulation distributions. These provided a pictorial explanation of how the bound
circulation at the blades is diffused into the fluid stream under the action of complex 3D
flows. In axial conditions, this bound circulation is diffused into the wake in the form of
trailing circulation only, while in yawed conditions the bound circulation will transform
393
Chapter6AerodynamicAnalysisoftheNRELPhaseVIWindTurbine
intowakecirculationconsistingofbothtrailingandshedcirculation.Itwasnotedthatfor
theNRELrotoratlowwindspeeds,atwhichtheanglesofattackatthebladesaresmall,the
trailingcirculationinthewaketendstobeconcentratedatthebladetipandrootregions.
Forhigherwindspeeds,thetrailingcirculationismoredispersedradially,withthepresence
of horseshoe vortices emerging from the middle blade sections. This behaviour was
observed in both axial and yawed conditions. At =300, the shed circulation levels in the
wakewerefoundtobegenerallysmallatallthewindspeedsconsidered.
The induced velocity distributions predicted by the freewake model for the given
aerodynamic loading acquired from the blade surface pressure measurements were also
investigated. These were found to vary considerably with windspeed and yaw. Both the
distributions of the azimuthally averaged axial induction factor (a1) and those of the axial
induced factor at the lifting line of the blades (a1,c) were plotted. For yawed conditions, a1
was found to vary periodically with the rotor azimuth angle () at frequency 2p. At low
windspeeds, this cyclic component was negligibly small, but it then became considerably
largerathigherwindspeeds.Thevariationofa1,cwiththebladeazimuthangleinyawwas
also periodic. At low windspeeds, this periodic variation was found to be regular with a
phaseanglethatchangeswiththeradiallocation(r/R).Athigherwindspeedshowever,the
periodicvariationdoesnotremainregularandthisresultsfromthepresenceofhorseshoe
vorticesinthemiddlebladesections.
From the freewake plots, it was possible to derive the approximate pitch of the helical
vortex sheets in the wake. It was found that while the rotor speed was fixed, this pitch
increases almost linearly with windspeed in both axial and yawed conditions. The wake
expansionwasfoundtobeminimalandthisisduetothefactthattheturbinewasgenerally
operatingatlowthrustcoefficients(CT<0.6,seetable6.1).
The angle of attack values derived by this novel approach were compared with the local
flow angles (LFA) measured in the wind tunnel using the fivehole probes. Large
differenceswerenotedandthisconsolidatesthefactthattheLFAvaluescannotbeassumed
tobeequaltotheangleofattackwhenderivingnewaerofoildatafromrotorexperiments.
Sincewithafreewakevortexmodelitispossibletodeterminetheinducedvelocitiesatany
point in the flow field, this novel approach provided the possibility to calculate the LFA
values at the probe tip locations. These calculated LFA values were compared with the
experimentalresultsobtainedwiththefiveholeprobes.Thiscomparisonwasameansfor
evaluatingthisapproachforcouplingthefreewakevortexmodelwiththebladepressure
measurements to find the angle of attack. It was found that in both axial and yawed
conditions,thefreewakevortexmodelresultsfortheLFAwereincloseagreementwiththe
measuredvaluesatlowwindspeeds.Athighwindspeeds,theagreementwasonlygoodfor
radial locations r/R>0.67. The agreement was not good for the inboard blade regions
(r/R<0.67),withthemaximumdiscrepancyreaching100.Itislikelythatthisdisagreementis
originating from the inherent limitations of the lifting line model embedded in the free
394
Chapter6AerodynamicAnalysisoftheNRELPhaseVIWindTurbine
wakevortexcode.However,inthecaseofyawedconditions,theaccuracywithwhichthe
fivehole probes measure the LFA is also being questioned because of the relatively large
cyclic pitching rates created by the rotor yaw. Further work is required to evaluate the
accuracywithwhichfiveholeprobesmeasuretheLFAatlargepitchrates.
Recommendationsforfuturework
Itisimportanttopointoutthefactthat,althoughthisstudyhassuccessfullydemonstrated
the effectiveness of using a freewake model in conjunction with blade pressure
measurements to obtained further insight of wind turbine aerodynamics, the results
obtainedaresubjecttothefollowinguncertainties:
The accuracy of the results is dependent on the accuracy with which the freewake
vortexmodelmodelsthebladesandwake.Inthisstudy,aliftinglinerepresentationof
thebladeswasusedandfurtherworkisrequiredtoinvestigatetheaccuracyoflifting
line models in representing the blades. Lifting surface and panel methods are more
accurateincapturingcertain3Dfloweffectsaroundtheblades.Furtherresearchisalso
required to validate the viscous modelling methodology adopted by the subject free
wakecode.
This study did not cater for the cycletocycle variations observed in the measured
valuesofCnandCt.Thesevariationswereconsiderablyhighinyawedconditions(refer
to Fig. 6.5). In this study, only the mean values of Cn and Ct were used in the
calculations.Futureworkwillincludeastochasticanalysistoaccountforsuchcycleto
cyclevariations.
InvestigatingtheimprovementinBEMpredictionswithnew3Daerofoildataandinflow
corrections
The new 3D aerofoil data derived from the blade pressure measurements in conjunction
withthefreewakevortexmodelwereusedinatypicalBEMcodetomodeltheNRELrotor
in both axial/yawed and attached/stalled conditions. In this way it was possible to assess
the improvement in BEM predictions were reliable 3D aerofoil data is used. For yawed
conditions, new inflow corrections to account for skewed wake effects in BEM codes (Fsa)
werealsoderived.Itwaspossibletoquantifytherelativeimportanceforsuchcorrections.
Various conclusions could be drawn from this study, and these provide more insight on
howBEMmodelscouldbeimproved.Actuallysomeofthefindingsconsolidatewhatwas
found previously when analysing the TUDelft rotor. A detailed discussion on the
limitationsofBEMcodesandthepossibilitiesforimprovingsuchcodeswillbedescribedin
thenextChapter,(Chapter7).
395
Chapter6AerodynamicAnalysisoftheNRELPhaseVIWindTurbine
396
Chapter7GuidelinesforImprovingtheReliabilityofBEMbasedDesignCodes
7.GuidelinesforImprovingtheReliabilityof
BEMbasedDesignCodes
The research work on the TUDelft andNREL wind tunnel model turbines presented in
this thesis introduced new methodologies for taking more benefit of limited experimental
data to be able to obtain a deeper knowledge of HAWT aerodynamics. A major outcome
was a deeper understanding of the limitations of BEMbased design codes in general and
whythesedonotalwayspredictaccurateresultsinparticularfortheconditionsofyawand
stall.Thisunderstandingprovidedfurtherinsightonhowtheaccuracyofsuchcodescanbe
improved through the development of better engineering models. In this chapter, the
various limitations of typical BEM codes identified during the indepth analysis of the
TUDelftandNRELrotorsinChapters4and6respectivelyaresummarisedanddiscussed.
Various guidelines on how these codes can be improved are also presented. These
guidelines are applicable for modelling conditions in which the turbine is operating in a
uniformandsteadywindspeedwiththerotoryawangle,tipspeedratioandbladepitch
anglemaintainedconstantwithtime.
7.1ModellingofAerofoilData
It was evident through the analysis with the NREL rotor that the accuracy of BEM
predictions for the blade load distributions is very much subject to what aerofoil data is
used.Theflowphenomenaoverarotatingwindturbinebladearefarmorecomplexthan
thoseobservedin2Dstaticwindtunnelexperimentsandconsequentlytheaerofoildatais
alsodifferent.Thedifferencesaremainlynotedinthebladetip/rootregion,athighangles
of attack and in unsteady conditions (as for example in yawed rotors). With the novel
approach presented in this thesis for finding the angle of attack using a freewake vortex
model,itwaspossibletoderivenewreliableaerofoildataforwindturbinesfromtheblade
pressure measurements taken during the NREL Phase VI experiments. A very interesting
findingwasthattheBEMpredictionsimprovedconsiderablywhenthisnewaerofoildata
was used. The improvement was noted even for situations that are known to be very
difficult to predict accurately, mainly in highly stalled and yawed conditions. It can
therefore be concluded that in many situations, BEMbased codes are deficient not
necessarily because the BEM theory is in itself inadequate but because the aerofoil data
usedisnotcorrect.Moreaccurateengineeringmodelsthatmodelaerofoildataneedtobe
developedandmadeavailabletothewindturbinedesigner.
This study has also shown that the application of the standard Prandtl tip/root loss
correctionwiththeuseof2Dstaticwindtunnelaerofoildataisinaccurate.Itoverpredicts
397
Chapter7GuidelinesforImprovingtheReliabilityofBEMbasedDesignCodes
the loading at the blade tip in both the TUDelft and NREL rotors (see Fig. 4.109 and Figs
6.51(a)). This Prandtl correction works by artificially increasing the induction at the blade
tipandroottoreducetheangleofattackandhencetheloading.Thistechniquewasfound
to overpredict the induction at the blade tip and root (see Fig. 4.103). A better way of
applyingthetip/rootlosscorrectionfortheBEMtheoryshouldbebasedonmodifyingthe
2D aerofoil data instead of the induction. However, from the newly derived 3D aerofoil
data from the NREL blade measurements it was found that tip/root loss effects were
already included and that the tip/root loss correction in the BEM calculations was not
required. This suggests that if 3D aerofoil data is derived from blade pressure
measurements or CFD, the tip/root loss correction is not required any longer in the BEM
calculations.
From the new aerofoil data derived in this study for the NREL rotor in axial conditions
(Figs. 6.19(a) and (b)), a new engineering model could be formulated. The model is still
based on input from the 2D aerofoil data (Cl,2D and Cd,2D) for both low and high angles of
attack.Thebasicformulationissimilartothatofpresentengineeringmodels(e.g.thoseof
Du and Selig [22] and Chaviaropoulos et al. [17] (Eqts. 3.31 and 3.33), but a tip/root loss
correctionfactorisincluded.The3Dliftanddragcoefficientsareexpressedas
Cl ,3 D (
= f * Cl ,2 D + fl ( Cl ,lin Cl ,2 D ) ) (7.1)
Cd ,3 D (
= f * Cd ,2 D + f d ( Cd ,2 D Cd ,2 D MIN ) ) (7.2)
f is a tip/root loss correction that corrects the aerofoil data and replaces the Prandtl
correction.Thetip/rootlosscorrectionisincludedbecausethe2Daerofoildataisstillused
asabasisforthemodel.FromtheanalysisoftheloadingdistributionsontheNRELrotorin
axialconditions,itwasfoundthatthefollowingmodelforthefactorfmodelsthetipand
rootlossonthisrotoradequatelyatallthetipspeedratiosconsidered:
r Rr
2
2
1
f = cos exp 18
1 r
R ( ) * cos 1
R R
exp 3.8 (7.3)
r R
r
R
R
Parametersflandfdareusedtomodeltheinfluencesofbladerotation(stalldelay)onthelift
and drag coefficients respectively. Values for these parameters were derived for the Cl,3D
and Cd,3D values of Figs. 6.19(a) and (b). The results are plotted in Figs. 7.1(a) and (b). For
<s, the influences of blade rotation were found to be small and Cl,2D and Cd,2D are very
closetoCl,linandCd,2DMIN,respectively.Thereforefor<s,flandfdcouldbetakenaszeroin
themodelling.For>s,flandfdwerefoundtovarybothwiththeradiallocationandangle
ofattack.
398
Chapter7GuidelinesforImprovingtheReliabilityofBEMbasedDesignCodes
1.2
r/R=0.25
1
r/R=0.3
r/R=0.4
0.8
r/R=0.47
0.6 r/R=0.63
r/R=0.8
0.4 r/R=0.95
fl
0.2
0
10 15 20 25 30 35 40 45 50
-0.2
-0.4
-0.6
(deg)
Figure 7.1(a) Variation of fl with angle of attack at different radial locations as derived from the
resultsofFig.6.19(a)andusingEqts.7.1and7.3.
3.2 r/R=0.25
r/R=0.3
2.8 r/R=0.4
r/R=0.47
2.4
r/R=0.63
2 r/R=0.8
r/R=0.95
1.6
fd
1.2
0.8
0.4
0
10 15 20 25 30 35 40 45 50
-0.4
-0.8
(deg)
Figure 7.1(b) Variation of fd with angle of attackat different radial locations as derived from the
resultsofFig.6.19(b)andusingEqts.7.2and7.3.
399
Chapter7GuidelinesforImprovingtheReliabilityofBEMbasedDesignCodes
In developing the new engineering model,the variationsof fl and fd were modelled using
theformulas:
fl = 0 if s
( )
fl = g ( m0 m ) e
a ( s ) 0.003( s )
3
* 1 e if > s
(7.4a)
fd = 0 if s
( )
f d = h ( n0 n ) e
b ( s ) 0.003( s )
3
* 1 e if > s
(7.4b)
whereswastakentobeequalto100.ThevaluesofFigs.7.1(a)and(b)wereusedtomodel
parametersa,g,m0,m,b,h,n0andnasfunctionofthelocalsolidity(c/r).Wheneachofthe
lattersevenparameterswasplottedagainstc/r,itwasfoundthatitisimpossibletomodel
the variations with sufficient accuracy using a simple linear or quadratic fit. A third or
fourthorder polynomial fit was required to model each of parameters a, g, m0, m, b, h, n0
andnasfunctionofc/r,asfollows:
4 3 2
c c c c
a = 180.34 + 236.02 109.19 + 20.01 0.927 (7.5a)
r r r r
3 2
c c c
g = 81.06 70.19 + 20.4 0.9 (7.5b)
r r r
3 2
c c c
m0 = 10.75 + 0.99 + 2.45 0.247 (7.5c)
r r r
3 2
c c c
m = 0.2934 + 0.1461 0.005 + 0.0006 (7.5d)
r r r
4 3 2
c c c c
b = 2.26 3.06 + 1.346 0.19 + 0.01 (7.6a)
r r r r
3 2
c c c
h = 0.17 + 19.4 5.49 + 1.84 (7.6b)
r r r
3 2
c c c
n0 = 88.85 + 75.21 15.1 + 0.72 (7.6c)
r r r
400
Chapter7GuidelinesforImprovingtheReliabilityofBEMbasedDesignCodes
3 2
c c c
n = 2.12 + 2.29 0.655 + 0.071 (7.6d)
r r r
Althoughtheabovenewmodelseemscomplex,itcanstillbehandledveryefficientlywith
personal computers. Figure 7.2 compares the 3D lift and drag coefficients (Cl,3D and Cd,3D)
predictedbythisnewengineeringmodel(Eqts.7.17.6)withthosederivedfromtheblade
pressuremeasurementsontheNRELrotorinconjunctionwiththefreewakevortexmodel
(i.e.thoseplottedinFigs.6.19(a)and(b)).Itisshownthatthenewmodelcanreproducethe
results quite accurately. The 3D lift coefficient predicted by Snels model (Eqts. 3.30 with
parameters a and b taken as 3 and 2, respectively) is also being included. It can be easily
notedthatthenewengineeringmodelprovidesmuchmoreaccuratepredictionsthanSnels
model at the blade root where stalldelay is predominant. Another shortcoming of Snels
modelisthatthedragcoefficientremainsuncorrected.FromFigures7.2,itisevidentthat
the drag coefficient also needs to be corrected, as in fact is being carried out by the new
model.
A limitation of the new engineering model (Eqts. 7.17.6) is that it has been developed
basedonalimitedamountofdatafromasinglerotorgeometry,i.e.theNRELrotorwith
theS809aerofoilshapefortheblades.Furtherworkisnecessarytoassessthevalidityofthis
modelforotherrotorsanditislikelythatfurthertuningwouldthenberequired.Also,this
model is only valid when the rotor is operating in axial and steady conditions. The
unsteady aerofoil data for yawed rotors may differ considerably from the corresponding
steadydataforaxialconditions,especiallyathighanglesofattackatwhichdynamicstall
takesplace.Thesedifferenceswerealreadydiscussedwhenderiving3Dunsteadyaerofoil
data from blade pressure measurements on the NREL rotor in yaw with the freewake
vortex model (see Figs. 6.33(a)(j)). This derived data is quite unique since it includes the
combined influences of unsteady effects (including dynamic stall) and 3D effects (stall
delay and tip/root loss). Developing a new improved engineering model for BEMbased
codestoproduceunsteadyaerofoildatarequiresanelaborateanalysiswhichisbeyondthe
scope of this thesis. Yet the results of Figs. 6.33(a)(j) should be very useful in developing
such a model. A major finding from this study is that the unsteady 3D Cl and Cdp
hysterisis loops are in general closer to the 3D steady aerofoil data than to the
corresponding2Dsteadydata.Thiswasobservedatbothattachedandstallconditionsand
itimpliesthat3Deffects(inparticularstalldelay)haveasignificantroleinunsteadyflow
behaviourofwindturbineblades.
401
Chapter7GuidelinesforImprovingtheReliabilityofBEMbasedDesignCodes
3.5 2.0
3.0 1.8
2.5 1.5
r/R=0.3 r/R=0.3 1.3
Cl 2.0
Cd
1.0
1.5
2D Delft
0.8
2D Delft
1.0
Exp Exp
0.5
0.5 New Model New Model
0.3
Snel
0.0
0 5 10 15 20 25 30 35 40 45
0.0
0 5 10 15 20 25 30 35 40 45
(deg) (deg)
2.0 1.5
1.8
1.3
1.5
Cl
1.0
Cd
0.8
0.8 2D Delft
Exp 0.5 2D Delft
0.5
New Model Exp
0.3 Snel 0.3
New Model
0.0 0.0
0 5 10 15 20 25 30 35 40 45 0 5 10 15 20 25 30 35 40 45
(deg) (deg)
1.5 1.5
1.3 1.3
Cd
0.8
Cl
0.8
2D Delft
0.5 0.5 2D Delft
Exp
New Model
Exp
0.3 0.3
New Model
Snel
0.0
0.0
0 5 10 15 20 25 30 35 40 45 0 5 10 15 20 25 30 35 40 45
(deg) (deg)
1.0
1.3
0.9
0.8
1.0
r/R=0.8 r/R=0.8
0.7
0.8 0.6
Cd
Cl
0.5
0.5
2D Delft
0.4
2D Delft
0.3
Exp
0.3 New Model
0.2 Exp
Exp 2D Delft
New Model
0.7
Exp
1.0 0.6
Snel
r/R=0.95 r/R=0.95 0.5 New Model
0.8
Cd
Cl
0.4
0.5 0.3
0.2
0.3
0.1
0.0 0.0
0 5 10 15 20 25 30 35 40 45 0 5 10 15 20 25 30 35 40 45
(deg) (deg)
Figure7.2Comparisonoftheliftanddragcoefficientspredictionsbythenewengineeringmodel
withthosefromFigs.6.19(a)and(b).Snelspredictionoftheliftcoefficientisalsoincluded.
402
Chapter7GuidelinesforImprovingtheReliabilityofBEMbasedDesignCodes
7.2Correctingtheinductionforskewedwakeeffectsinyaw
The BEM theory can produce quite accurate results for the aerodynamic load
distributionsinaxialconditionsprovidedreliable3Daerofoildataisused.Howeverthisis
notthecaseinyawedconditionssincetheavailabilityofreliable3Dalonemaystillnotbe
sufficient to provide accurate results. This is because the BEM theory lacks the physics to
model the complex and skewed wake created by a yawed turbine. Consequently, apart
fromengineeringmodelsthatprovidereliableaerofoildata,otherengineeringmodelsare
additionallyrequiredtocorrectforskewedwakeeffectsinyawasalreadyoutlinedbriefly
inChapter3,section3.5.
InmanystateoftheartBEMbaseddesigncodes,thecorrectionforskewedwakeeffectsis
carried out by introducing a correction factor Fsa as shown in Eqt. 3.27. In effect, this
correction factor corrects the ratio of the local induction factor at the blades (a1,c) to the
azimuthally averaged value (a1) at different radial locations and azimuth angles of the
blade.VariousengineeringmodelsareusedtomodelFsa,someofwhichwerepresentedin
Chapter3(Eqts.3.233.26).Inthisproject,theinducedvelocitydistributionsfortheTUDelft
andNRELrotorsinyawwhereinvestigatedindetail.Inthiswayitwaspossibletostudy
the variations of a1,c vs and thus be able to understand better the limitations of the
engineering models currently available to model correction parameter Fsa in BEMbased
designcodes.
IntheanalysisoftheTUDelftrotor,detailedhotfilmmeasurementsinthenearwakewere
taken with the rotor operating at yaw angles =300 and 450 with the tip speed ratio and
bladetippitchanglemaintainedfixedat8and20respectively.IntheanalysisoftheNREL
rotor,thenovelapproachforcouplingthebladepressuremeasurementswiththefreewake
vortexmodelmadeitpossibletoestimatetheunsteadyinducedvelocitydistributionsatthe
rotorplane.Thesedistributionswerederivedfordifferentwindspeedswiththeyawangle,
rotorspeedandbladepitchanglemaintainedfixedat300,72rpmand30,respectively.Both
the hotfilm measurements on the TUDelft rotor and the freewake calculations on the
NRELrotorconfirmthattheperiodicvariationofa1,cvs onthebladeliftinglinesresulting
fromskewedwakeeffectsinyawedrotorsisfarmoredifferentthanthesinusoidalvariation
predictedbyGlauert(Eqt.3.23).Thisismainlyduetorootvorticityeffectswhichtendtobe
predominant in wind turbines, as opposed to what is usually seen in helicopter rotors in
forwardflight.ActuallythisfindingisnotnewasthiswasalreadydiscoveredintheJoule
projects[83,84]andintheworkofSchepers[69].Itwasfoundinthepresentstudythatfor
hightipspeedratios( =5.48)atwhichthetimedependentanglesofattackattheblades
are small, the periodic variation of a1,c vs on both the TUDelft and NREL rotor was
regular with a phase angle changing with radial location (see Figs. 4.89, 4.90 and Figs.
6.39(a),(b)).Forsuchconditions,anengineeringmodeloftheformasthatdevelopedbythe
ECN(Eqt.3.26)tendstosuitable.However,itwasdiscoveredthatforlowtipspeedratios
wherethetimedependentanglesofattackatthebladesarelarge,theperiodicvariationsof
403
Chapter7GuidelinesforImprovingtheReliabilityofBEMbasedDesignCodes
a1,c vs on the NREL rotor as predicted by the freewake code were found to be quite
complex and irregular (see Figs. 6.39(c), (d) and (e)). These variations have a higher
harmonic content than that catered for by Eqt. 3.26. Yet it was found that a third order
Fourierseriesofthefollowingformwouldbesufficient:
A0
Fsa = + A1 cos ( + 1 ) + A2 cos ( 2 + 2 ) + A3 cos ( 3 + 3 ) (7.7)
2
whereA0,A1,A2andA3aretheamplitudesoftherespectiveharmonicsandareafunctionof
theyawangle(),theradiallocation(r/R)andthetipspeedratio(). 1, 2and 3arethe
phaseanglesofeachharmonicindegreesandarealsoafunctionof , r/Rand.
7.2.1Developmentofengineeringmodelusingaheuristicapproach
basedonFourierseries
InanattempttoderiveanewimprovedengineeringmodelforFsa,theHAWT_FWCresults
of Figs. 6.38 and 6.39 were used to determine appropriate values for the amplitude and
phaseangleparametersofEqt.7.7tobeabletomodeltheNRELrotorat300yaw.Todoso,
Eqt.6.8wasappliedtocalculateFsaateachbladeazimuthangle()andradiallocation(r/R)
and at each windspeed (U =5, 7, 10, 13 and 15m/s). Eqt. 6.8 was applied with the Prandtl
tip/rootlossfactorfsettounity(sincethisfactorwouldnotberequiredifthetip/rootloss
correction is applied to the aerofoil data instead to the induced velocities, as discussed
earlierinsection7.1).Curvefittingnumericalmethodsdescribedin[27]werethenapplied
to the third order Fourier series to derive the values of A0, A1, A2, A3, 1, 2 and 3. The
resulting values are plotted in Figs. 7.3 and 7.4. It is observed from these plots that the
requiredparametersareverysensitivetoboththeradiallocation(r/R)andtheoperatingtip
speed ratio (). The variations are quite irregular and this makes it difficult to derive an
improvedengineeringmodelbasedonEqt.7.7thatisvalidforawiderangeofoperating
conditions.OneshouldalsorememberthatthevaluesshowninFigs.7.3and7.4areonly
valid for =300. For other yaw angles, new sets of values would be required. This makes
matters even more complicated. Moreover, parameters A0, A1, A2, A3, 1, 2 and 3 also
dependonthegeometryoftherotoranditsblades.Infactwhencomparingthedistribution
for Fsa obtained from HAWT_FWC on the TUDelft rotor at =300 and =8 with those
obtained for the NREL rotor at the same yaw angle and approximately same tip speed
(=7.54), considerable differences were noted. Such difficulties indicate the complexity of
theproblemwhendealingwithyawaerodynamicsandthatapplyingaheuristicapproach
for deriving a new improved engineering model for Fsa (i.e. deriving the model by only
observing numerical results from measurements or advanced models without involving
physical models) is quite restricted in providing more accurate predictions. A major
limitation is that the derived engineering model will only be valid for the rotor and
operatingconditionsforwhichitwasoriginallyderived.
404
Chapter7GuidelinesforImprovingtheReliabilityofBEMbasedDesignCodes
6 2
r/R=0.3
r/R=0.3 1.8
5 r/R=0.47
r/R=0.47
1.6 r/R=0.63
r/R=0.63
1.4 r/R=0.8
4 r/R=0.8
r/R=0.95 1.2 r/R=0.95
A1
A0
0.8
2
0.6
0.4
1
0.2
0 0
2 3 4 5 6 7 8 2 3 4 5 6 7 8
1 0.6
0.9 r/R=0.3 r/R=0.3
r/R=0.47 0.5 r/R=0.47
0.8 r/R=0.63
r/R=0.63
0.4
0.7 r/R=0.8 r/R=0.8
r/R=0.95
r/R=0.95
0.6 0.3
A3
A2
0.5
0.4 0.2
0.3 0.1
0.2
0
0.1 2 3 4 5 6 7 8
0 -0.1
2 3 4 5 6 7 8
Figure 7.3 Values of A0, A1, A2 and A3 thatare required for improved engineering model forFsa
(Eqt.7.7)inordertoyieldaninducedvelocitydistributionaspredictedbyHAWT_FWCforNREL
rotorat=300.
7.2.2Alternativeapproachusingprescribedwakevortexmodel
An important requirement for engineering modelsis thatthey should be valid for a wide
range of rotor/blade geometries and operating conditions. Knowing the limited physical
background of the BEM theory and keeping in mind the complexity of yawed rotor
aerodynamics,itisvitalfornewimprovedengineeringmodelsforBEMbaseddesigncodes
to implicitly make use of more comprehensive aerodynamic theory rather than simply be
based on a heuristic approach. One possibility is to integrate a rigid or prescribed vortex
model using the BiotSavart law directly into the BEM code. This would not affect
significantly the computational cost of the solution considering the fact that rigid and
prescribed vortex models are by far much more computationally efficient than freewake
vortexmodels.Butindoingso,itisnecessarytocheckwhetherthevortexmodeladoptedis
sufficiently reliable to modelthe circulationin the wake. In this project,the computations
with the freewake vortex model HAWT_FWC on the NREL rotor have provided a better
understanding of how the circulation is distributed in the rotor wake and how such a
distributioninfluencesthevariationofa1,cvs (andhencealsoparameterFsa).Itwasfound
405
Chapter7GuidelinesforImprovingtheReliabilityofBEMbasedDesignCodes
2 (deg)
0 0
2 3 4 5 6 7 8 -30 2 3 4 5 6 7 8
-30 r/R=0.3 -60
-60
-90 r/R=0.47 -90
r/R=0.63 -120
-120 r/R=0.8 -150
180
150
120
90
60
30 r/R=0.3
3 (deg)
0 r/R=0.47
-30 2 3 4 5 6 r/R=0.63
7 8
r/R=0.8
-60
r/R=0.95
-90
-120
-150
-180
-210
Figure7.4Valuesof 1, 2and 3thatarerequiredbyimprovedengineeringmodelforFsa(Eqt.
7.7)inordertoanyieldinducedvelocitydistributionaspredictedbyHAWT_FWCforNRELrotor
at=300.
thattheregularityofthevariationofa1,cvs at hightipspeedratiosonboththeTUDelft
and NREL rotors results from the fact that the trailing circulation emanating from the
bladesisconcentratedatthetipandroot(seeFigs.5.46,5.47andFigs.6.35(a),(c)).Itcould
therefore be concluded that for such conditions, a rigid or prescribed vortex model
consistingofonlyasingleskewedtipvortexhelixandasimilaroneattherootwouldlikely
be sufficiently accurate to model the wake provide that the pitch of each helix are
prescribed correctly. However, at low tip speed ratios on the NREL rotor, the unsteady
sheddingofhightrailingcirculationatthemiddlebladesectionscausedtheirregularityin
variationofa1,cvs .Forsuchconditionstherigidorprescribedvortexmodelshouldbetter
consistofhelicalskewedvortexsheetsinordertobeabletoincludetheeffectsofthehigh
trailingcirculationemanatingfromthemiddlesectionsoftheblade.
In this project it was noted that the developed prescribed vortex model HAWT_PVC can
produce induced velocity distributions that are reasonably close to those from the free
wakevortexmodelHAWT_FWC.ForinstancethiscouldbeobservedontheTUDelftrotor
predictionsforua,cat=300and450(compareHAWT_PVCresultsofFigs.4.91and4.92with
those of HAWT_FWC of Figs. 5.41 and 5.42). Recall that in HAWT_PVC, the wake is
406
Chapter7GuidelinesforImprovingtheReliabilityofBEMbasedDesignCodes
modelled as a vortex sheet and the BiotSavart law is applied to calculate the unsteady
inductionintheflowfieldoftherotorwiththeboundcirculationatthebladesrequiredas
input. This vortex model is described in detail in section 4.3.3. The wake skew angle and
pitchneedalsotobeprescribed.
The reasonably good predictions from HAWT_PVC suggest that better predictions for
yawedrotorscanbeachievedbyBEMcodesifthisvortexmodelisintegratedintheBEM
solution.Anapproachfordoingsoisnowproposedandisbasedontwosimplemodelsto
interfaceaBEMcodetoHAWT_PVC:
1. the first model is used to determine an approximate pitch (pw) for the helical vortex
sheets modelled by HAWT_PVC from an azimuthally averaged axial induction factor
computedbytheBEMequationEqt.3.21.Thismodelisgivenby:
pw p cos ( ) 2
= = ( cos ( ) + a ) (7.8)
1
R R
wherea1istakentobeequaltotheazimuthallyaveragedvalueat7080%R.Theabove
equation was found to be reasonably accurate in estimating the helical pitch when
comparing its predictions with those obtained on the TUDelft rotor using smoke
visualisation experiments (see section 4.2.3) and those obtained on the NREL rotor
usingthefreewakecalculations(seeFigs.6.18and6.36).Thecomparisonispresented
in Fig. 7.5. The values of a1 at 7080%R used in Eqt. 7.8 for the TUDelft rotor were
obtainedfromthehotfilmmeasurements(Figs.4.60,4.89and4.90.)whilethoseforthe
NREL rotor were obtained from HAWT_FWC (Figs. 6.21(d) and 6.38). The maximum
percentageerrorpwinthismodelwasfoundtobelessthan19%.
2. the second model is used to determine the wake skew angle (s) required by
HAWT_PVC. This is a new model that was derived from the smoke visualisation
experiments on the TUDelft rotor (section 4.2.3) and is based on input of the axial
thrustcoefficienttofindsataparticularyawangle.Thequasilinearrelationsbetween
s and CT noted in these experiments (see Fig. 4.32) were used to develop this model.
Themodelequationsaregivenby:
s = + s
CT 0 = 0.016 1X 10 4 2 3 X 106
3
s
= a ( CT CT 0 ) (7.9)
a=0 if CT 0 CT
407
Chapter7GuidelinesforImprovingtheReliabilityofBEMbasedDesignCodes
4.2
3.8
3.4
2.2
TUDelft, Approx, Yaw 30 deg
Figure 7.6 illustrates the proposed approach on how a typical BEM solution algorithm
could be modified to be able to interface the prescribed vortex model HAWT_PVC to the
BEMdesigncode.Theapproachconsistsofthefollowingsequenceofsteps:
Step 1: initially the whole rotor revolution is subdivided into a fixed amount of equally
spacedtimestepsasinFig.3.7.
Step2:initialvaluesforthelocalaxialandtangentialinductionfactorsattheblades(a1,cand
a2,c)areassumed.
Step3:therelativeflowvelocitycomponent(V,VandVr)andanglesofattack()atthe
bladesarefoundateachtimestepoverthewholerevolution.
Step 4: the 3D lift and drag coefficients (Cl,3D and Cd,3D) are then computed using an
engineering model similar to the one presented in Eqts. 7.1 7.6 in order to account for
tip/rootloss,stalldelayandunsteadyeffects(e.g.dynamicstall).
408
Chapter7GuidelinesforImprovingtheReliabilityofBEMbasedDesignCodes
Step5:theboundcirculationdistributionatthebladesforeachtimestepisthenestimated
usingtheKuttaJoukowskiequation:
B ,3 D = 1 cVr Cl ,3 D (7.10)
2
Step6:TheBEMEqt.3.21issolvediterativelytofinda1.
Step 7: The axial thrust loading distribution at the blades is determined using Eqt. 3.20a.
Numerical integration is then applied to find the azimuthally averaged rotor global axial
thrustcoefficientCT.
Step8:ParameterspwandsarecomputedfromEqts.7.8and7.9.
Step 9: The distribution for B,3D from Step 5 is then prescribed to the prescribed vortex
model HAWT_PVC, together with parameters pw and s from Step 8. HAWT_PVC then
calculatesanewdistributionfora1,canda2,cattheblades.
Step10:Thenewvaluesfora1,canda2,cfromStep9arecomparedwiththepreviousvalues
inputted to Step 3. Steps 3 to 9 are then repeated until convergence in a1,c and a2,c is
achieved.
TheinterfacingofaprescribedvortexmodelintotheBEMsolutionwillobviouslyincrease
thecomputationalcost.However,thisiswellaccommodatedbytheincreasedefficiencyof
moderncomputersthatarebecomingavailable.Alsothediscretizationofthevortexmodel
inHAWT_PVCneednotbesettoofinetoachieveverygoodnumericalaccuracy.Thehelical
vortical wake need not extend too downstream from the rotorplane. In this project it was
foundthatextendingonerotordiameterdownstreamonlyissufficient.Thecalculationof
correctingparameterFsaisnotrequiredsincethedistributionofa1,ciscalculateddirectlyby
thevortexmodel.
In this approach, the influence of wake expansion is being ignored by taking parameters
Rt,w1andRt,w2inHAWT_PVCaszero.Thisinfluenceissmallwhentherotorisoperatingat
lowthrustcoefficients(CT<0.6)butmaybecomeimportantathigherthrustvalues.
The proposed approach of Fig. 7.6 for integrating prescribed vortex methods in BEM
engineeringcodesisinfactverysimilartoprescribedwakevortexmethodsasthoseused
in references [19, 28, 90]. But instead of using empirical or semiempirical models for
prescribing the vortex wake pitch, the BEM equation (Eqt. 3.21) and Eqt. 7.8 are used
instead.Thismakesthemodellinglessdependentonexperimentalinputandsimplerwhich
inturnmightturnoutintoreducedcomputationalcosts.Yetfurtherworkisstillrequiredto
validateitoverawiderangeofrotorgeometriesandoperatingconditions.Furtherworkis
alsonecessarytoexamineanyadvantagesovercurrentlyavailableprescribedwakevortex
methodsasfarascomputationalefficiencyandreliabilityareconcerned.
409
Chapter7GuidelinesforImprovingtheReliabilityofBEMbasedDesignCodes
BEM Code
Start
Divideonewholerotorrevolutioninto
afixedamountofazimuthal timesteps
(asshowninFig.3.7)
Assumeinitialvaluesfora1,c anda2,c
atthebladesforeachrotortimestep
Computetherelativeflowvelocitiesand
theanglesofattackateachbladetimestep
(V,V, Vr and)
Computethe3Dliftanddragcoefficients
(Cl,3D andCd,3D)accountingfortip/root
loss/stalldelay/dynamicstall
HAWT_PVC
Solvefora1 byapplyingBEMEqt.3.21
foreachrotortimestep
Convergence YES
ina1,c anda2,c ?
Determinetheazimuthallyaveragedrotorglobal NO
axialthrustcoefficientCT byintegratingnumerically
loadingcomputedfromEqt.3.20a,andusingEqt.4.10
Prescribenewvaluesof
a1,c anda2,c attheblades
toBEMcode
Computeparameterspw ands usingEqts.7.8and7.9
End
Figure7.6ProposedapproachforinterfacingprescribedvortexcodeHAWT_PVCtoaBEMcode
toimproveBEMaerodynamicloadpredictionsforwindturbinesinyaw.
410
Chapter7GuidelinesforImprovingtheReliabilityofBEMbasedDesignCodes
7.2.3Correctiontotheaxialmomentumequationforyawedrotors
Values for ka for yawed rotors may be derived using experimental data and/or advanced
models using Eqt. 4.85. Note that a value of ka=1 would imply that no correction to the
momentumequationisrequired.Inthisproject,valuesforkawerederivedfortheTUDelft
rotorusingtheinflowmeasurementsinconjunctionwithunsteadyaerofoiltheoryfor=300
and 450. It was found that ka was less then unity and that the value decreases with yaw
anglewhenmaintaining and tipconstant(seeFig.4.102).Valuesforkawerealsoderived
for the NREL rotor at =00 and 300 for different tip speed ratios using the blade pressure
measurementsandthefreewakevortexmodelHAWT_FWC.Thevalueswerefoundtobe
veryclosetounityatalltipspeedratios,evenat=300.
The values of ka derived from the aerodynamic analysis on the TUDelft and NREL rotors
areplottedinFig.7.7.Theresultsareplottedagainsttheglobalaxialthrustcoefficient,CT.
The value of ka=1 is included for the case in which =00 up to CT =1. Beyond CT =1, the
momentumequation(Eqt.3.6a)isnolongervalidandanequationsuchasEqt.3.4bisused
instead. It is apparent that ka is dependent on both the yaw angle and the axial thrust
coefficient. It appears that, for a given yaw angle, ka will decrease with an increase in CT.
One can easily recognize the fact that the NREL rotor was operating at low values of CT
(refertotable6.1)andthisisthemainreasonwhykawasclosetounityupto =300.Onthe
otherhand,theTUDelftrotorhadlargeCTvaluesandconsequentlykawaswellbelowunity
atboth=300and450.
Sinceonlyalimitedsetofoperatingconditionsonbothrotorswereconsidered,itisdifficult
todevelopanengineeringmodelforkafromtheresultsinFig.7.7.Thelargeuncertaintiesin
411
Chapter7GuidelinesforImprovingtheReliabilityofBEMbasedDesignCodes
the ka values of the TUDelft rotor at =300 and 450 resulting from the uncertainties in the
inflow measurements from which they were originally derived, also make it difficult.
HoweveritislikelythatitispossiblethatthevariationofkawithCTatagivenyawangle
varies as depicted in Fig. 7.7. In this case, the engineering model could be expressed
mathematicallyasfollows:
k a (C , ) =
(1 aC ) e T
3
1CT b
(7.11)
T
2.718
where parameters a and b are variables depending on the yaw angle. However, more
analysisisrequestedtodeterminewhetherthisisatypicalvariationforwindturbinesby
considering a wider range of rotors and operating conditions. Furthermore, experiments
shouldbeconductedtodetermineappropriatekavaluesinyawforlargeCTvalues(CT>0.6).
1.8
1.6
1.4
NREL rotor
1.2 Momentum Eqt. 3.6a
for =00 fails
1
ka
0
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 1.1 1.2 1.3 1.4 1.5 1.6
CT
Figure7.7VariationofkawithCTandyawanglefortheTUDelftandNRELrotors.
412
Chapter8ProjectOutcome,ConclusionsandRecommendations
8.ProjectOutcome,Conclusionsand
Recommendations
8.1ProjectOutcome
1. In the first case, the TUDelft model turbine was tested in the openjet wind tunnel
facility at Delft University of Technology. Detailed hotfilm measurements were
performed in the near wake of the model turbine. Smoke visualisation experiments
werealsocarriedouttotracethetipvortexpathsandhenceobtaindetailsoftherotor
wakegeometry.Duetolimitationsoftheequipmentitwasimpossibletomeasurethe
blade aerodynamic loads distributions. A methodology was developed to derive the
timedependent aerodynamic load distributions at the rotor blades from the hotfilm
measurementsinconjunctionwithanadvancedunsteadyaerofoilmodel.Despitethe
factthattheanalysiswasonlyrestrictedtoattachedflowconditionsattheblades(for
which the unsteady aerofoil model is considerably accurate), this methodology was
foundtobequitechallengingtoapplyduetothefollowingtwodrawbacks:(a)First,
the derived aerodynamic loads are very sensitive to the errors in the hotfilm
measurements. In other words, a small percentage error in the inflow measurements
mayresultintoasignificantlylargererrorintheloading;(b)Second,itwasfoundthat
the bound circulation derived using the employed unsteady aerofoil model was
unrealisticallyhighatthebladetipandrootregions,eventhoughtheangleofattack
wasestimateddirectlyfromtheinflowmeasurements.Thiswasduetothefactthatthe
unsteadyaerofoilmodelusedisa2Dmodelandthusdoesnotcaterforthehighly3D
flowphenomenatakingplaceatthebladetipandroot.Thereforeacorrectionhadto
beadded.
413
Chapter8ProjectOutcome,ConclusionsandRecommendations
TheresultsderivedfromtheinflowmeasurementsontheTUDelftrotorwereusedto
carry out a detailed investigation of a typical BEM code when modelling axial and
yawedturbines.Despitethatattachedflowconditionswereonlybeingtreatedandalso
the fact that the results derived from the inflow measurements had a high level of
uncertainty, it was still possible to have a better understanding of the limitations of
BEMbaseddesigncodesandobtainfurtherintuitionofhowthesecouldbeimproved.
The study demonstrated that the application of the standard Prandtl tip/root loss
correctionwiththeuseof2Daerofoildataisinaccurateasitoverpredictstheinduced
velocityatthebladetipandrootregions.Abetterapproachtocorrectfortip/rootloss
in the BEM theory is to correct the 2D aerofoil data rather than the local induction.
When correcting for skewed wake effects in yaw, correcting the ratio of local blade
element induced velocity to the annular average induced velocity alone may not
alwaysbesufficient.Theannularaveragedinducedvelocityitselfmayalsoneedtobe
corrected.
2. Inthesecondcase,theNRELPhaseVIwindturbinethatwasextensivelytestedinthe
NASA Ames wind tunnel in 2000 was analysed. In the experiments, timeaccurate
blade pressure measurements were taken with the rotor operating in both axial and
yawedconditionstogetherwithstraingaugemeasurementsfortheoutputtorqueand
therootflap/leadmoments.However,detailedinflowmeasurementsattherotorplane
were not carried out. In this project, a novel and comprehensive methodology was
proposedforusingthebladepressuremeasurementsinconjunctionwithafreewake
vortex model to estimate the angle of attack distributions at the blades more
accurately, together with the inflow distributions at the rotorplane and wake
geometry.Thestudyconsidereddifferentoperatingtipspeedratiosinbothaxialand
yawconditionsthatyieldedbothattachedandseparatedflowsovertheblades.New
3Dliftanddragaerofoildatawerederived.Itcouldbeobservedthatthisnewdatais
considerablydifferentfromthecorresponding2Dwindtunnelaerofoildataduetothe
presenceofbladetip/rootlosseffects,stalldelayorelseunsteadyconditionsresulting
fromrotoryaw(mainlydynamicstall).Thederivedfreewakegeometrysolutionsfor
given blade loading conditions were plotted together with the corresponding wake
trailing and shed circulation distributions. These plots helped comprehend how the
unsteadyboundcirculationformedatthebladesiseventuallyconvectedintothewake
undertheactionofcomplexthreedimensionalflows.
The new 3D lift and drag aerofoil data were then used to assess to improvement in
BEM load predictions in axial and yawedconditions. The BEM predictions improved
considerablyevenwhendealingwithhighlystalledandyawedconditions.Foryawed
conditions, new inflow corrections to account for skewed wake effects in BEM codes
were also derived. It was possible to quantify the relative importance for such
correctionsinimprovingBEMresults.
414
Chapter8ProjectOutcome,ConclusionsandRecommendations
From this project, guidelines on how BEMbased models can be improved are presented
anddiscussed.Thesemainlyfallundertooclassifications:
2. Improvement of engineering models for skewed wake effects in yaw: Improved engineering
modelstocorrectforskewedwakeeffectsoverawiderangeofoperatingtipspeeds
are also necessary. An attempt was made to develop one such model from the
unsteadyinductiondistributionobtainedfortheNRELrotorusingFourierseries.The
modelwasintendedtocorrectsolelytheratioofthelocalinductionatthebladestothe
correspondingannularaveragedvalue.Althoughitwasconfirmedthatathirdorder
Fourierseriesissufficienttorepresenttherequiredcorrections,itwasfoundthatthe
Fouriercoefficientsarehighlydependentnotonlyontheradiallocationandtheyaw
anglebutalsoontheoperatingtipspeedratio.Thedependencyontheradiallocation
and tip speed ratio was found to be quite irregular and therefore quite complex to
model. Also comparison of the unsteady induction distributions of the TUDelft and
NREL rotors suggest that there is also a considerable dependency on the rotor
geometry. The analysis has shown that the use of this approach for developing an
accurate engineering model applicable over a wide range of rotor geometries and
operating conditions in yaw is difficult. This is a consequence of the fact that the
physicsbehindtheaerodynamicsofyawedturbinesiscomplex.Analternateapproach
isbeingproposedinwhichaBEMcodeisinterfacedtoaprescribedvortexmodelto
correctforskewedwakeeffects.Thisapproachismorephysicallycomprehensiveand
isthereforemorereliable.
Asalreadydescribedearlier,correctingforskewedwakeeffectsbyonlycorrectingthe
ratioofthelocalinductionatthebladestothecorrespondingannularaveragedvalue
without correcting the ratio of annular averaged induced velocity itself may be
insufficient. This is due to the inherent limitation of the axial momentum equation
whentreatingyawedconditions.Insomecases,anotherengineeringmodelisrequired
tocorrectthemomentumequation.Thiscorrectionissimilartothatusedtocorrectthe
same equation for high loading conditions. Whether this correction is required for
yawedcasesnotonlydependsontheyawanglebutalsoontheaxialthrustcoefficient.
FromtheanalysisontheTUDelftandNRELrotoritwasfoundthatsuchacorrection
isnotrequireduptoayawangleof300aslongastheaxialthrustcoefficientdoesnot
exceed 0.45. For higher axial thrust coefficients, an engineering model to correct the
415
Chapter8ProjectOutcome,ConclusionsandRecommendations
momentumequationbecomesanecessity.Suchamodelcouldbederivedfromwind
tunnelexperimentsoradvancedmodelssuchasfreewakevortexcodes.
8.2Conclusions
1. Usingafreewakevortexmodelitispossibletoderivetheangleofattackattheblades
from blade pressure measurements in both axial and yawed conditions. From the
derivedangleofattackandfromknowledgeofthemeasuredCnandCtvalues,new3D
steadyandunsteadyaerofoildatacanbeobtained.
2. With a freewake vortex model it is also possible to derive from the blade pressure
measurements the complex 3D induction at the rotorplane together with the wake
geometry and circulation distribution. These are vital in order to have a better
understanding of a wind turbines aerodynamic behaviour in both steady and
unsteadyenvironments.
3. Given that reliable 3D aerofoil data is available, BEM predictions for wind turbines
improve significantly, even for conditions of deep stall and rotor yaw. In many
situations,BEMresultsforthebladeaerodynamicloadingdistributionsareunreliable
notnecessarilybecausetheBEMtheoryisdeficientbutbecauseincorrectaerofoildata
isused.
4. TheapplicationofthestandardPrandtltip/rootlosscorrectionmodelwith2Daerofoil
data in BEM codes is inaccurate since it overpredicts the induction at the blade
tip/root.Amoreappropriatetip/rootlosscorrectionintheBEMtheoryistocorrectthe
2Daerofoildataratherthantheinduction.
5. When using 3D lift and drag aerofoil data derived directly from the blade pressure
measurements, the tip/root loss correction model is no longer required in the BEM
calculations.Thisisbecausethetip/rootlosseffectsarealreadyaccountedforinthe3D
aerofoildata.
6. To improve the BEM predictions in yawed wind turbines, the use of accurate 3D
unsteadyaerofoildataalonemaynotbesufficient.Twoadditionalmodelsneedtobe
incorporated to correct the induction computed by BEM for skewed wake effects in
yaw:
Model (a) to correct the ratio of the local axial induction at the blades to the
corresponding annular averaged value. Many models developed in the past years to
correctthisratioinBEMcodesarenotaccurateatallformanycircumstances.Dueto
thecomplexityofyawedrotoraerodynamics,developinganewimprovedengineering
model based on a heuristic approach isvery difficult.This is because the ratiois not
onlydependentontheradiallocation(r/R)andyawanglebutalsoontheoperatingtip
416
Chapter8ProjectOutcome,ConclusionsandRecommendations
speed ratio and rotor geometry. Therefore it is imperative to include more physical
modelsasforexampleprescribedwakevortexmodelsusingtheBiotSavartlaw.
Model (b) to correct the axial momentum equation for its deficiency when treating
yawedrotors.Yet,thiscorrectionisnotrequiredwhenconsideringsmallyawangles
(<300)andlowaxialthrustcoefficients(CT<0.45).
8.3Recommendations
This study has clearly demonstrated that, until the time will come when computer
capacity becomes powerful enough to practically integrate the more advanced freewake
vortexmethodsandCFDtechniquesfullyintowindturbineengineeringdesignsoftware,
there is still considerable potential to improve the accuracy of BEMbased aerodynamic
models.Thefollowingrecommendationsarebeingmadeforfuturework:
1. Emphasisismadeontheurgentneedformorereliableengineeringmodelsthatpredict
3Daerofoildataforwindturbinedesigners.Researcheffortsshouldfocusonproviding
a better understanding of the flow physics locally around the rotating blades. More
extensive blade pressure measurements are indispensably useful in reaching this aim
and to further validate CFD models. Despite that NavierStokes solvers are still too
computationally demanding to be fully encapsulated into wind turbine design
software, they have a vital role in supporting the development of improved
engineering models for aerofoil data to be used in BEMbased codes. Here it is
importanttopointoutthefactthatthelimitedavailabilityofaccurate3Daerofoildata
for wind turbine applications is not only restricting BEMbased models from being
moreaccurate,butalsothemorecomprehensivemodelssuchasliftinglinefreewake
vortexmodelsandliftingline/actuatordiscCFDmodels.
2. The new engineering model for aerofoil data for axial conditions developed in this
projectusingtheNRELmeasurementsneedstobevalidatedforotherrotorgeometries.
A similar engineering model for aerofoil data in unsteady conditions as in yawed
turbinescouldbedevelopedusingthedatabeingpresentedinthisthesis.
3. As regards improving BEM predictions for the inflow distributions in yawed rotors,
furtherworkisrequiredtovalidatetheapproachproposedforinterfacingprescribed
wake vortex models to BEM codes. The work should investigate the improvement in
predicting the unsteady induced velocity at the blades as well as the increased
computational costs incurred by introducing the vortex model at different levels of
wake discretization. As this proposed method still relies on the use of the axial
momentum equation, further work is required to develop engineering models for
parameter ka in yaw. Experimental campaigns consisting of both blade pressure
measurements and detailed near wake inflow measurements over a wide range of
417
Chapter8ProjectOutcome,ConclusionsandRecommendations
operating conditions (varying yaw angle and including both low and high operating
thrustcoefficients)arenecessaryindevelopingsuchmodels.
ThenewexperimentaldatathatwillshortlybecomeavailablefromtheEUfundedproject
MEXICOproject(shortforMeasurementsandExperimentsunderControlledConditions)
should be of great benefit for the work being recommended here. In this project, a three
bladed modelturbine with a geometry very similar to fullscale rotors will be extensively
testedintheDutchGermanDNWlargewindtunnel.Theturbinewillhaveadiameterof
about 4m and will be highly instrumented to be able to carry out both blade pressure
measurementsaswellasdetailedinflowmeasurementsattherotorplaneandinthewake.
Flowvisualisationmeasurementswillalsobeperformed.
418
References
References
1. Afjeh A.A. and Keith T.G. Jr., A Simplified Free Wake Method for Horizontal
AxisWindTurbinePerformancePrediction,JournalofFluidsEngineering,Vol.108,
pp.400406,1986.
2. Ananthan S., Leishman J.G., Ramasamy M., The Role of Filament Stretching In
The FreeVortex Modelling of Rotor Wakes, 58th Annual Forum and Technology
DisplayoftheAmericanHelicopterSocietyInternational,Montreal,Canada,2002.
3. AndersonJ.D.,FundamentalsofAerodynamics,McGrawHill,Thirdedition,USA,
2001,pp.331332.
4. AndersonM.B.,MilborrowD.J.,andRossN.J.,PerformanceandWake
Measurementsona3mDiameterHorizontalAxisWindTurbine,DOEContract
ReportNo.E/5A/CON/1090/177/020,UniversityofCambridge,England,1982.
5. BagaiA.andLeishmanJ.G.,RotorFreeWakeModellingusingaPseudoImplicit
Technique Including Comparisons with Experimental Data, Journal of the
AmericanHelicopterSociety,Vol.40,No.2,1995.
6. BareiR.andWagnerS.,TheFreeWake/HybridWakeCodeROLVMATool
for Aerodynamic Analysis of Wind Turbines, European Wind Energy Conference,
LbeckTravemnde,Germany,1993.
7. BarlowJ.B.,RaeW.H.andPopeA.,LowSpeedWindTunnelTesting,Wiley,New
York,ThirdEdition1999,Chapt.17.
8. BeddoesT.S.,PracticalComputationofUnsteadyLift,Vertica,1984,8(1),pp.55
71.
9. BeniniE.andToffoloA.,OptimalDesignofHorizontalAxisWindTurbines
UsingBladeElementTheoryandEvolutionaryComputation,JournalofSolar
EnergyEngineering,Nov.2002,Vol.124,pp.357363.
10. BhagwatM.J.andLeishmanJ.G.,GeneralizedViscousVortexModelfor
ApplicationtoFreeWakeandAeroacousticCalculations,58thAnnualForumand
TechnologyDisplayoftheAmericanHelicopterSocietyInternational,Montreal,Canada,
2002.
11. BierboomsW.A.A.M.,AComparisonbetweenUnsteadyAerodynamicModels,
ProceedingsfromtheEuropeanWindEnergyConference,Amsterdam1991.
419
References
12. BjrckA.andThorS.E.,DynamicStalland3DEffects,ProceedingsfromThe
EuropeanWindEnergyConference,May1996,Gteborg,Sweden.
13. Bruining A., van Bussel G.J.W., Corten C.P. and Timmer W.A., Pressure
DistributionsfromaWindTurbinesBlade:FieldMeasurementsComparedtoa2
dimensional Wind Tunnel Data, DUTIVW93065R, Delft University of
Technology,1993.
14. BruunH.H.,HotWireAnemometry,OxfordSciencePublications,Oxford,England,
1995.
15. van Bussel G.J.W, The Aerodynamics of Horizontal Axis Wind Turbine Rotors
Explored with Asymptotic Expansion Methods, Phd thesis, Delft University of
Technology,theNetherlands,1995.
16. ChattotJ.J.,OptimizationofWindTurbinesUsingaHelicoidalVortexModel,
AIAA20030522,Proceedingsfrom22ndASMEWindEnergySymposium,2003,Reno,
USA.
17. Chaviaropoulos P.K. and Hansen M.O.L., Investigating ThreeDimensional and
RotationalEffectsonWindTurbineBladesbymeansofaQuasi3DNavierStokes
Solver,JournalofFluidsEngineering,Vol.122,June2000,pp.330336.
18. Conlisk A.T., Modern Helicopter Rotor Aerodynamics, Progress in Aerospace
Sciences,Vol.37,2001,pp.419476.
19. CotonF.N.,WangT.andMcD.GalraithR.A.,AnExaminationofKey
AerodynamicModellingIssuesRaisedbyTheNRELBlindComparison,Wind
Energy,2002;5:pp.199212.
20. deVriesE.,ThinkingBiggerAretherelimitstowindturbinesize?,Renewable
EnergyWorld,MayJune2005.
21. Diveux T., Sebastian P., Bernard D. and Puiggali J.R., Horizontal Axis Wind
Turbine Systems: Optimisation Using Genetic Algorithms, Wind Energy, 2001; 4:
pp.151171.
22. Du Z. and Selig M., A 3D StallDelay Model for Horizontal Axis Wind Turbine
Performance Predictions, AIAA980021, ASME Wind Energy Symposium, Reno,
1998.
23. EWEA, Wind Energy The Facts, European Wind Energy Association, Brussels,
2004.
420
References
24. EWEA and Greenpeace, Wind Force 12 A Blueprint to Achieve 12% of the
WorldsElectricityfromWindPowerby2020,EuropeanWindEnergyAssociation,
Brussels,andGreenpeace,Hamburg,2004.
25. EWEA, Prioritising Wind Energy Research Strategic Research Agenda of the
WindEnergySector,EuropeanWindEnergyAssociation,Brussels,2005.
26. van Garrel A., Development of a Wind Turbine Aerodynamics Simulation
Module,NationalEnergyResearchFoundation,ECNC03079,2003.
27. Gerald C.F. and Wheatley P.O., Applied Numerical Analysis, AddisonWesley
PublishingCompany,FifthEdition,USA,1994.
28. Gerber B.S, Tangler J.L, Duque E.P.N and Kocurek J.D, Peak and Post Power
AerodynamicsfromPhaseVINASAAmesWindTurbineData,Proceedingsofthe
SpecialTopicConference:theScienceofMakingTorquefrom Wind,EWEA,Delft,The
Netherlands,2004.
29. Giguere P., Selig M.S., Design of a Tapered and Twisted Blade for the NREL
combinedExperimentRotor,NREL/SR50026173,NREL,Golden,CO,1999.
30. GlauertH.,AGeneralTheoryfortheAutogiro,ARCR&M786,1926.
31. GrantI.,MoM.,PanX.,ParkinP.,PowellJ.,ReineckeH.,ShuagK.,CotonF.N.and
Lee D., An Experimental and Numerical Study of the Vortex Filaments in the
Wake of an Operational, Horizontalaxis, Wind Turbine, Journal of Wind
EngineeringandIndustrialAerodynamics,Vol.85,2000,pp.177189.
32. GrantI.,ParkinP.andWangX.,OpticalVortexTrackingStudiesofaHorizontal
AxisWindTurbineinYawusingLaserSheet,FlowVisualisation,Experimentsin
Fluids,Vol.23,1997,pp.513519.
33. GrantI.andParkinP.,ADPIVStudyoftheTrailingVortexElementsfromthe
BladesofaHorizontalAxisWindTurbineinYaw,ExperimentsinFluids,Vol.28,
2000,pp.368376.
34. Gupta S. and Leishman J.G, Accuracy of the Induced Velocity of Wind Turbine
WakesUsingVortexSegmentation,AIAA20040828,Proceedingsfrom23rdASME
WindEnergySymposium,Reno,2004.
35. HandM.M,Simms,FingershL.J.,JagerD.W.,CortellJ.R.,SchreckS.andLarwood
S.M., Unsteady Aerodynamics Experiments Phase VI: Wind Tunnel Test
ConfigurationsandAvailableDataCampaigns,NRELTechnicalReportNREL/TP
50029955,Golden,Colorado,Dec2001.
421
References
36. HaansW.,SantT.,vanKuikG.A.M.andvanBusselG.J.W.,MeasurementofTip
VortexPathsintheWakeof aHAWTunderYawedFlow Conditions,Journalof
Solar Energy Engineering, Special Issue on Wind Energy, Vol. 127, No. 4, Nov. 2005,
pp.437596.
37. Haans W., Sant T., van Kuik G.A.M. and van Bussel G.J.W., Velocity
MeasurementsintheNearWakeofaHorizontalAxisWindTurbine,Proceedings
fromthe31stEuropeanRotorcraftForum,Florence,Italy,2005.
38. Hendriks H.B. and Zaaijer M., Dutch Offshore Wind Energy Converter
Executive summary of the public research activities, Report by Energy Research
CentreandDelftUniversityofTechnology,theNetherlands,January2004.
39. Himmelskamp H., Profile Investigations on a Rotating Aircrew, Dissertation
Gottingen1945,ReportsandTranslationsNo.832,Sept.1947.
40. Johansen J. and Sresen N.N., Airfoil Characteristics from 3D CFD Rotor
Computations, Proceedings of the Special Topic Conference: the Science of Making
TorquefromWind,EWEA,Delft,TheNetherlands,2004.
41. JonesR.T.,OperationalTreatmentoftheNonUniformLiftTheoryinAirplane
Dynamics,NASATechnicalNote667,1938
42. Katz J. and Plotkin A., Low Speed Aerodynamics, Second Edition, Cambridge
UniversityPress,2001.
43. Kuhn M., Dynamics and Design Optimisation of Offshore Wind Energy
Conversion Systems, Phd Thesis, Delft University Wind Energy Institute, The
Netherlands,2001.
44. LainoD.J.,HansenA.CandMinnemaJ.E.,ValidationoftheAerodynSubroutines
UsingNRELUnsteadyAerodynamicsExperimentData,WindEnergy,Vol5,2002,
pp.227244.
45. Liano D.J. and Hansen A.C, Continued Validation of the Aerodyn Subroutines
using NREL Unsteady Aerodynamics Experiment Data, AIAA20020518,
Proceedingsfrom21stASMEWindEnergySymposium,Reno,2002.
46. Lee D.J and Na S.U., Numerical Simulations of Wake Structures Generated by
Rotating Blades using a Time Marching Free Vortex Blob Method, European
JournalofMechanicsB/Fluids.Vol.18(1),1999.
47. Leishman J.G. and Beddoes T.S., A SemiEmpirical Model for Dynamic Stall,
JournaloftheAmericanHelicopterSociety,Vol.34,No.3,pp.317,1989.
422
References
48. Leishman J.G., Modelling Sweep Effects on Dynamic Stall, Journal of the
American.HelicopterSociety,34(3),pp.1829,1989.
51. LeishmanJ.G.,BhagwatM.J.,andBagaiA.,FreeVortexFilamentMethodsforthe
AnalysisofHelicopterRotorWakes,JournalofAircraft,Vol.39,No.5,pp759775,
2002.
61. Rasmussen F., Hansen M.H., Thomsen K., Larsen T.J., Bertagnolio F., Johansen J.
andMadsenH.A.,ChristianB.andHansenA.M.,PresentStatusofAerolasticity
ofWindTurbines,WindEnergy,Vol.6,2003,pp.213228.
423
References
62. RonstenG.,StaticPressureMeasurementsinaRotatingandnonRotating2.375m
Wind Turbine Blade. Comparison with 2D Calculations, Proceedings from the
EuropeanWindEnergyConference,1991,pp.214220.
63. vanRooijR.P.J.O.M,ModificationoftheboundarylayercalculationinRFOILfor
improved airfoil stall prediction, Report IW96087R, Delft University of
Technology,1996.
64. van Rooij R.P.J.O.M., Description of the aerodynamic design process for a wind
turbinerotor,DelftUniversityofTechnology,SectionWindEnergy,Sept.2003.
65. van Rooij R.P.J.O.M., Bruining A. and Schepers J.G., Validation of Some Rotor
StallModelsbyAnalysesoftheIEAAnnexXVIIIfielddata,Proceedingsfromthe
EuropeanWindEnergyConference,Madrid2003.
66. van Rooij R.P.O.M., Timmer W.A. and Bruining A., Determination of The Local
Flow Angle on Rotating Blades, Proceedings of the World Wind Energy Conference
andExhibition,July2002,Berlin,Germany.
67. Sant T., van Kuik G.A.M. and van Bussel G.J.W. , Using Near Wake
MeasurementstoImproveBEMEngineeringModelsforYawedWindTurbines,
ProceedingsfromtheEuropeanWindEnergyConference,London2004.
68. SantT.,vanKuikG.A.M.andvanBusselG.J.W.,EstimatingtheAngleofAttack
from Blade Pressure Measurements on the NREL Phase VI Rotor using a Free
WakeVortexModel:AxialConditions,WindEnergy,DOI:10.1002/we.210.
69. SchepersJ.G.,AnEngineeringModelforYawedConditions,Developedonthe
BasisofWindTunnelMeasurements,AIAA990039,ASMEConference,1999,
Reno,USA.
70. SchepersJ.G,HeidraJ.,ThomsenK.,LarsenT.J.,FousseskisD,SmithR.R.,KraanI.,
VisserB.,yeS.,GananderH.,CarlenI.,BelessisM.,VoutsinasS.,DrostL.,
VerificationofEuropeanWindTurbineDesignCodes,Proceedingsfromthe
EuropeanWindEnergyConference,Copenhagen,2001;pp.581586.
71. Scherer R., Blade Design Aspects, Renewable Energy, 16 (1999) 12721277,
Pergamon.
72. Schmitz S., Chattot J.J., Application of a Parallelized Coupled Navier
Stokes/VortexPanel Solver to the NREL Phase VI Rotor, AIAA20050593,
Proceedingsfrom24thASMEWindEnergySymposium,Reno,2005.
73. Schreck S., The NREL FullScale Wind Tunnel Experiment Introduction to the
SpecialIssue,WindEnergy,Vol.5,2002,pp.7784.
424
References
74. Schreck S. and Robinson M., Structures and Interactions Underlying Rotational
Augmentation of Blade Aerodynamic Response, AIAA20030520, Proceedings
from22ndASMEWindEnergySymposium,Reno,2002.
75. Schreck S. and Robinson M., Competing Local and Global Influences on
Rotationally Augmented HAWT Blade Aerodynamics, Proceedings of the Special
Topic Conference: The Science of Making Torque from Wind, EWEA, Delft, The
Netherlands,2004.
76. SchreckS.,RobinsonM.,HandM.andSimmsD.,BladeDynamicStallVortex
KinematicsforaHorizontalAxisWindTurbineinYawedConditions,ASME
JournalofSolarEnergyEngineering,Vol.123,pp.272281,Nov.2001.
77. SharpeD.,JenkinsN.,BossanyiE.andBurtonT.,WindEnergyHandbook,John
Wiley&Sons,NewYork,2002.
78. SimmsD.,SchreckS.,HandM.andFingershL.J.,NRELUnsteadyAerodynamics
ExperimentintheNASAAmesWindTunnel:AComparisonofPredictionsto
Measurements,NRELTechReportNREL/TP50029494,Golden,Colorado,June
2001.
79. Snel H., Heuristic Modelling of Dynamic Stall Characteristics, Proceedings from
theEuropeanWindEnergyConference,Dublin1997.
80. SnelH.,ReviewofthePresentStatusofRotorAerodynamics,WindEnergy,Vol.
1,1998,pp.4669.
81. SnelH.,ReviewofAerodynamicsforWindTurbines,WindEnergy,Vol.6,2003,
pp.203211.
82. SnelH.,HouwinkR.,BosschersJ.,Piers,W.J.,vanBussel,G.J.W.andBruining,A.
Sectional Prediction 3D Effects for Stalled Flow on Rotating Blades and
Comparison with Measurements, Proceedings from the European Wind Energy
Conference,1991,pp.395399.
83. SnelH.andSchepersJ.G.(ed),JointInvestigationofDynamicInflowEffectsand
Implementation of an Engineering Method, Netherlands Energy Research
Foundation,ECNC94107,1994.
84. SnelH.andSchepersJ.G.(ed),DynamicInflow:YawedConditionsandPartial
SpanPitchControl,NetherlandsEnergyResearchFoundation,ECNC95056,1995.
85. SresenB.,RenewableEnergyItsphysics,engineering,environmentalimpacts,
economics&planning,ElsevierAcademicPress,ThirdEdition,2004.
425
References
86. SresenN.N.,AGeneralPurposeFlowSolverAppliedtoFlowoverHills,Ris
NationalLaboratory,Roskilde,Denmark,June1995.
87. SresenN.N,MichelsenJ.A.,SchreckS.,NavierStokesPredictionsoftheNREL
PhaseVIRotorintheNASA80ftX120ftWindTunnel,WindEnergy,Vol5,2002,
pp.151169.
88. SperaD.A.,WindTurbineTechnologyFundamentalConceptsofWindTurbine
Engineering,ASMEPress,NewYork,1994.
89. Strelets M., Detached Eddy Simulation of Massively Separated Flows, AIAA
20010879,31stAIAAAerospaceSciencesMeetingandExhibit,2001.
90. Tangler,J.L.,TheNeboulousArtofUsingWindTunnelAerofoilDatafor
PredictingRotorPerformance,WindEnergy,2002,Vol.5,pp.245257.
91. TanglerJ.L.,InsightintoWindTurbineStallandPoststallAerodynamics,Wind
Energy,Vol7,2004,pp.247260.
92. van der Tempel J., Design of Support Structures for Offshore Wind Turbines,
PhdThesis,DelftUniversityofTechnology,theNetherlands,2006.
93. TheodorsenT.,GeneralTheoryofAerodynamicInstabilityandtheMechanismof
Flutter,NACAReport496,1935.
94. Thor S.E. and Taylor P.W., Longterm Research Needs for Wind Energy for the
TimeFrame20002020,WindEnergy,2002;Vol.5,pp.7375.
95. Tongchitpakdee C., Benjanirat S, Sankar L.N., Numerical Simulation of the
Aerodynamics of Horizontal Wind Turbines Under Yawed Flow Conditions,
AIAA20050773,Proceedingsfrom24thASMEWindEnergySymposium,Reno,2005.
96. VermeerL.J.,WindTunnelExperimentsonaRotorModelinYaw.Proceedingsof
the12thsymposiumonAerodynamicsofWindTurbines,Lyngby,1998,pp.1120.
97. Vermeer N.J., A New Rotor Model for Aerodynamic Measurements in the
ImprovedOpenJetWindTunnel,NationalWindEnergyConference,Lunteren,The
Netherlands,1990(inDutch).
98. VermeerL.J.,AReviewofWindTurbineWakeResearchatTUDelft,AIAA2001
0030,Reno,2002.
99. VermeerL.J.,SrensenJ.N.andCrespoA.WindTurbineWakeAerodynamics,
ProgressinAerospaceSciences,2003,Vol39,pp.467510.
426
References
100. Viterna L.A., and Corrigan R.D, Fixed Pitch Rotor Performance of Large
HorizontalAxisWindTurbines,ProceedingsfromtheWorkshoponLargeHorizontal
Axis Wind Turbines, NASA CP2230, DOE Publication CONF810752, Cleveland,
OH,NASALewisResearchCentre,1981.
101. Voutsinas S.G., Belessis M.A. and Huberson S., Dynamic Inflow Effects and
Vortex Particle Methods, Proceedings from the European Wind Energy Conference,
LbeckTravemnde,Germany,1993.
102. WagnerH.,berdieEntstehungdesdynamischenAuftriebesvonTragflgeln,
ZeitschriftfrAngewandteMathematikundMechanick,5(1),1925,pp17.
103. Wood D.H., A ThreeDimensional Analysis of StallDelay on a HorizontalAxis
Wind Turbine, Journal of Wind Engineering and Industrial Aerodynamics, Vol. 37,
1991,pp.114.
104. ZhiquanY.,XiogL.,andYanC.,GlobalOptimumDesignMethodandSoftware
for Rotor Blades of Horizontal Axis Wind Turbines, Wind Engineering, Vol. 26,
No.4,2002,pp.257267.
427
References
428
Appendices
AppendixAMaximumPowerCoefficientforaYawed
TurbineDisc
Thesimpleactuatordisktheoryofsection3.1isusedheretoderivethemaximumpower
coefficient for a yawed turbine. From simple energy considerations (neglecting energy
lossesduefriction,afinitenumberofbladesandotherlosses),thepowerextractedfromthe
windisequaltotherateofchangeofkineticenergyofthefluidflowingthroughtherotor.
FromFig.3.1,thisisequalto
= 1 AU ' (UCos ( ) ) (UCos ( ) + ua1 ) (A.1)
2 2
P
2
SubstitutingEqt.3.1intheaboveequationandsimplifyingresultsin
Defining the power coefficient as C P = P ( 1 2 AU ) and using Eqt. A.2, the following
3
expressionresults
Eqt.A.3isplottedinFig.A.1fordifferentyawangles.Itcanbenotedthatastheyawangle
increases,thepowercoefficientatanyinductionfactordecreases.Theoptimumvaluefora1
occursatlowervaluesathigheryawangles.
The maximum power coefficient, CP , Max , at a particular yaw angle is found by partial
CP
differentiation, using = 0 . This yields the relation for the optimum axial induction
a1
factor a 1,m ,foragivenyawangle.Thisrelationis
429
Appendices
==00deg = =1515
deg =
30deg
= 30
==45
45deg = =
6060
deg
0.6
0.5
0.4
CP(a1,)
0.3
0.2
0.1
0
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
-a 1
FigureA.1Variationofthepowercoefficientwiththeaxialinductionfactorfordifferentyawangles
aspredictedbyEqt.A.3.
0.4
0.35
0.3
0.25
-a1,m
0.2
0.15
0.1
0.05
0
0 10 20 30 40 50 60 70 80 90 100
( deg )
Figure A.2 Variation of the optimum axial induction factor with yaw angle as predicted by Eqt.
A.4.
430
Appendices
SubstitutingEqt.A.4inA.3forthecaseinwhicha1=a1,mresultsintheequationforCP,max:
C P , Max (a
1, m
,) = (A.5)
Cos ( ) + 2a1, m
1
2
Puttinga1,m=1/3inthisequationforzeroyawresultsinCP,Max=16/27whichisinfacttheBetz
efficiency. Eqt. A.5 gives the ideal wind turbine efficiency that can be reached at a given
yawangle.
431
Appendices
AppendixBCalculationofAerodynamicLoads
InducedattheYawBearing
Thisappendixdescribesanumericalmethodtoobtainthe3Dforcesandmomentsatthe
yawbearingresultingfromtheaerodynamicforcesattheblades.Theloadsaredetermined
asafunctionoftherotorazimuthangle.
Theknownaerodynamicforcesatthelocal referenceframearefirsttransformedto
the XaYaZa reference frame. These are then used to find the resultant (total) forces and
moments at the yaw bearing O by numerical integration. In Fig. B.1, the blade element
similar to that shown in Fig. 3.3 is illustrated in 3D space. The position of its centre with
G
respecttoOisdefinedbyvector vo andisequalto
rSin( )
G
vo ( X Y Z ) = d a + r (B.1)
n n n
rCos ( )
zr
z
Zn
r Z
Z
a q,a
A
G Vzo
A vo
zo Y
r Myo O Ya,yr,Yn
da Vyo
Mxo
F Y
Vxo
X
Rotor axis
Xn
Windvelocity
Xa
FigureB.1AerodynamicforcesandmomentsinducedatyawbearingOduetoonebladeelement
432
Appendices
The local aerodynamic forces at each blade element in the XaYaZa reference frame are
denotedby Vxh, Vyhand Vzh.Thecorrespondingmomentsare Mxh, Myhand Mzh.These
areobtainedusingthematrixtransformationKsuchthat
Vxh A
Vyh = K A (B.2)
Vzh 0
and
M xh 0
M yh = K
0 (B.3)
M zh q , a
Vxo Vxh
Vyo = Vyh (B.4)
Vzo Vzh
M xo M xh Vxh
G
M yo = M yh + vo ( X Vyh (B.5)
n Yn Z n )
M zo M zh Vzh
Eqts.B.1..5aboveareappliedtoeachelementofallthebladesfordifferentrotorazimuth
angles. An index notation is adopted as for codes HAWT_BEM and HAWT_LFIM using
parameters ,banditodenotetherotortimestep,bladenumberandradiallocationofthe
bladerespectively.TheresultantforcesandmomentsattheyawbearingOduetoasingle
bladearedeterminedbynumericalintegration.Applyingthetrapeziumruleandusingthe
elementalvaluesofeachbladeelement,thentheforcesandmomentsatOduetoasingle
bladenandatrotorazimuthanglemaybeexpressedby
433
Appendices
(Vxo ) ,b = 1
2 {( V )
xo , b ,0
+ 2 ( Vxo ) ,b ,1 + ... + ( Vxo ) ,b ,i + ... + ( Vxo ) ,b , n 1 + ( Vxo ) ,b , n 1 }
(Vyo ) ,b = 1
2 {(V )
yo ,b ,0
+ 2 ( Vyo ) ,b ,1 + ... + ( Vyo ) ,b ,i + ... + ( Vyo ) ,b , n 1 + ( Vyo ) ,b , n 1
}
(Vzo ) ,b = 1
2 {( V ) zo , b ,0
+ 2 ( Vzo ) ,b ,1 + ... + ( Vzo ) ,b ,i + ... + ( Vzo ) ,b , n 1 + ( Vzo ) ,b , n 1 }
( M xo ) ,b = 1
2 {( M )
xo , b ,0
+ 2 ( M xo ) ,b ,1 + ... + ( M xo ) ,b ,i + ... + ( M xo ) ,b , n 1 + ( M xo ) ,b , n 1 }
( M xo ) ,b = 1
2 {( M )
yo ,b ,0
+ 2 ( M yo ) ,b ,1 + ... + ( M yo ) ,b ,i + ... + ( M yo ) ,b , n 1 + ( M yo ) ,b , n 1
}
( M xo ) ,b = 1
2 {( M )
zo , b ,0
+ 2 ( M zo ) ,b ,1 + ... + ( M zo ) ,b ,i + ... + ( M zo ) ,b , n 1 + ( M zo ) ,b , n 1 }
(B.6)
Theresultant(total)forcesandmomentsatOforarotorwithBbladesareevaluatedusing
equations B.6 above and summing up for each time step . This operation is expressed
mathematicallyas
b = B 1 b = B 1 b = B 1
( SVxo ) = (Vxo ) ,b (S ) Vyo = (Vyo ) ,b ( SVzo ) = (V zo ,b )
b=0 b=0 b =0
b = B 1 b = B 1 b = B 1
( SMxo ) = (M xo ,b) (S ) Myo = (M )
yo ,b ( SMzo ) = (M )
zo ,b
b=0 b=0 b=0
(B.7)
In fact SVyo and SMyo and are equal to the rotor axial thrust and rotor output torque
respectively(i.e.TandQ).Therotoroutputpowerissimplyfoundfrom
P = S Myo = Q (B.8)
NotethatSMzoisthemomentabouttheZnaxisandnotabouttheZaxis.Itisdefinedasthe
yawing moment about the Zn axis. However when the rotor axis tilt angle is zero, then
these would be the same. The loads of Eqts. B.7 may be expressed in nondimensionless
formasshownintableB.1.
434
Appendices
TableB.1:Dimensionlessformfortheresultantaerodynamicforcesandmomentsattheyawbearing
Oandtheoutputpower.
DimensionlessCoefficient Equation
SideThrustCoefficient SVxo
CVxo =
1 AU 2
2
AxialThrustCoefficient SVyo
CVyo = = CT
1 AU 2
2
UpwardThrust SVzo
Coefficient CVzo =
1 AU 2
2
PitchingMoment S Mxo
Coefficient CMxo =
1 AU 2 R
2
TorqueCoefficient S Myo
CMyo = = CQ
1 AU 2 R
2
YawingMoment S Mzo
Coefficient CMzo =
1 AU 2 R
2
PowerOutputCoefficient P
CP =
1 AU 3
2
435
Appendices
AppendixCCalculationoftheInducedVelocityfroma
VortexFilamentusingtheBiotSavartLaw
Assuming that the region outside the vortex sheet and tip vortex is potential and the
temperature is constant, the vortex systems in the rotor wake can be represented by the
incompressibleBiotSavartlaw.Foranincompressiblefluid,thecontinuityequationwhich
appliesforthevelocityfieldisgivenby
G
u = 0 (C.1)
G
where u isthevelocityand isthegradientoperatorwhichinCartesiancoordinatesis
= i + j + k (C.2)
x y z G
Wemayexpressthevelocityfieldasthecurlofavectorpotential B suchthat
G G
u = B (C.3)
G
Since the curl of a gradient vector is zero, G B is indeterminate to within the gradient of a
scalarfunctionofpositionandtime,and B canbeselectedsuchthat
G
B = 0 (C.4)
Thevorticity, ,isdefinedby
G
= u (C.5)
UsingEqts.C.3andC.5,thevorticitybecomesequalto
G G G
( ) ( )
= B = B B (C.6)
2
G
SubstitutingEqt.C.4inC.6resultsinPiossonsequationforthevectorpotential B given
by G
= B (C.7)
2
G
G
Theaboveequationmaybesolvedfor B usingGreenstheorem.Substitutingtheresulting
relationfor B inEqt.C.3yieldsthefollowingtheBiotSavartlaw:
G G
G ds r
du =
4 rG 3
(C.8)
436
Appendices
This relation gives the induced velocity at a point A due to a single vortex line having a
G G
constantcirculation . ds isanelementallengthofthevortexlineand r istheposition
G
vectorfromthenodetotheelement ds .ThesolutionforEqts.C.7andC8.maybefoundin
various advanced aerodynamic textbooks such as Marshall [54, page 86] and Katz and
Plotkin [42, page 173174]. Note that the BiotSavart law is only valid for incompressible
flows.
Fig.C.1showsanodeAandastraightlinevortexhavingendpointsBandC.Theinduced
G G
velocityatpointAduetovortexBCisequalto u andisperpendiculartotheplane r ds .
TheinducedvelocityisobtainedbyintegratingEqt.C.8overthelengthofthevortexline.
G
B (xB, yB, zB) G ds
c Vortex Line C (xC, yC, zC)
h
G
r
G
a G
b
FigureC.1Vortexlinegeometryforcalculationofthethreedimensionalinducedvelocityatapoint
A
Define h as the perpendicular distance from node A to the vortex line and is the angle
G
betweenhand r .Then,
h
sin = cos = (C.9)
r
h
ds = d ( h tan ) = d (C.10)
cos
2
NowusingrelationshipEqt.C.9,Eqt.C.8canbewrittenintermsofhand:
du = cos d (C.11)
4 h
437
Appendices
SubstitutinginEqt.C.11andintegratingfrom = /2 to/2
2
4h
u = cos d
2
or
u = ( cos + cos ) (C.12)
4 h
G G
G
Thevelocity u isinadirectionperpendiculartothe
G r ds planeandhenceperpendicular
G
totheplaneofthevectors a and b .Nowdefinepositivecirculation suchthat u points
G
intothepaperasshowninFig.C.1.Now u canbewrittenas
G G
G ab
u = u G G (C.13)
ab
G G G G
Toderivethethreedimensionalinducedvelocity u intermsof a and b ,define c as
G G G
c = b a (C.14)
UsingthedotandcrossproductrelationsandFig.C.1,
GG GG
a.c = a c cos (C.15)
GG G G
b .c = b c cos (C.16)
G G
G ac
h = a sin = G (C.17)
G c
andfromthedefinitionof c wehave
G G G G G G G
a c = a (b a ) = a b (C.18)
G G G G G G G
b c = b (b a ) = a b (C.19)
ThevectoridentitiesinEqts.C.15C.19canbeemployedtoprovethat
cos + cos
= G G G G (C.20)
h ab b a
438
Appendices
SubstitutingEqt.C.20inC.12andC.13weget
G G G GG
G ( a + b )( a b b .a ) G G
G
=
u
4 G G G G2 ( a b ) (C.21)
a b a b
To express the three dimensional components of the induced velocity in Cartesian co
ordinateform,weproceedasfollows:
G G
ThepositionsofpointsA,BandCshowninFig.C.1havecoordinates(x A,yA,zA),(xB,yB,
G
a = ( x B x A )i + ( y B y A ) j + ( z B z A ) k (C.22)
G
b = ( xc x A )i + ( y c y A ) j + ( z c z A )k (C.23)
Eqts.C.22andC.23aresubstitutedinEqt.C.21toget
u x = GX (C.24)
u y = GY (C.25)
u z = GZ (C.26)
GX,GYandGZarethegeometricinfluencecoefficientsgivenby
( r1 + r2 )
G X = [( z A zC )( yC yB ) ( y A yC )( zC zB )] (C.27)
2 r1r2 ( r1 + r2 ) L2
2
( r1 + r2 )
GY = [( xA xC )( zC zB ) ( z A zC )( xC xB )] (C.28)
2 r1r2 ( r1 + r2 ) L2
2
( r1 + r2 )
GZ = [( y A yC )( xC xB ) ( xA xC )( yC yB )] (C.29)
2 r1r2 ( r1 + r2 ) L2
2
where
1
= ( x A xB ) + ( y A y B ) + ( z A z B ) (C.30)
2 2 2 2
r1
439
Appendices
1
= ( x A xC ) + ( y A yC ) + ( z A zC )
2 2 2 2
r2 (C.31)
1
= ( xC xB ) + ( yC y B ) + ( zC z B ) (C.32)
2 2 2 2
L
Eqts.C.24C.32arethediscretizationequationsfortheBiotSavartlawtodeterminethe3D
velocities (ux, uy and uz) at any point A in the flow field induced by a straightline vortex
filament having a circulation . However a problem exists in calculating the induced
velocitiesclosetothevortexfilaments.Forpointsonthevortexfilamentstheselfinduced
velocityiszero,butforpointsmovingclosetoavortexfilamenttheinducedvelocitywill
tendtoinfinity,accordingtoBiotSavartslaw.Soforcalculatinginducedvelocitiesinsuch
pointssomeformofdesingularisationisrequired.Onewayofdealingwiththisproblemis
to apply a cutoff method: when the perpendicular distance h is less or equal to a cutoff
distance,theinducedvelocitiesaresettozero.Anotherapproachistoapplyavortexcore
modelasexplainedinsection5.2.2.
Thedistancehmaybefoundfromthevectorrelation
r1 r2
h = (C.33)
r2 r1
FromEqts.C.30C.32,
12 + 2 2 + 2 2
h = (C.34)
L
where
1 = ( yB y A )( zC z A ) ( yC y A )( zB z A ) (C.35)
2 = ( xB xA )( zC z A ) ( xC xA )( z B z A ) (C.36)
3 = ( xB xA )( yC y A ) ( xC xA )( yB y A ) (C.37)
440
Appendices
AppendixDLinearandSplineNumerical
Interpolation
Interpolation is a method of constructing new data points from a discrete set of known
data points. Given a sequence of n+1 distinct data points [xi,yi] that may not be evenly
spacedsuchthat
x0 < x1 < ... < xn 1 < xn
we can fit in a piecewise function between each of two successive data points to form an
interpolatingfunctionG(x)where
g 0 ( x ) x [ x0 , x1 ]
g1 ( x ) x [ x1 , x2 ]
G ( x ) = g i ( x ) x [ xi , xi +1 ]
(D.1)
# #
g n 1 ( x ) x [ xn 1 , xn ]
wheregi(x)isthefittedfunctionbetweenthetwosuccessivedatapoints(seeFig.D1).Oneor
moreconstraintsareappliedtothefittedfunctionssuchthatG(x)iscontinuousintherange
x0xn.
gi1(x) gi(x)
gi+1(x) gn1(x)
g 0 (x)
LinearInterpolation
The most simple form of interpolation is linear interpolation where the data points are
graphicallyconnectedbystraightlines.TheresultingfunctionG(x)isapolygon.
Algebraically,eachgi(x)isalinearfunctionconstructedas
441
Appendices
yi +1 yi
g i ( x ) = yi + ( x xi ) (D.2)
xi +1 xi
SinceG(x)mustbecontinuousateachdatapoint,then
g i ( xi ) = gi +1 ( xi ) i = 0,1,...n 2 (D.3)
Thisisthecaseforwhich
yi yi 1
g i 1 ( xi ) = yi 1 + ( xi xi 1 ) = yi
xi xi 1
yi +1 yi
g i ( xi ) = yi + ( xi xi ) = yi
xi +1 xi
CubicSplineInterpolation
Cubic spline interpolation refers to interpolation in which the fitted functions consist of
cubicpolynomialfunctionshavingtheform
( x ) = ai ( x xi ) + bi ( x xi ) + ci ( x xi ) + d i (D.4)
3 2
g i
Thefollowingconstraintsareapplied:
Eqts.D.5saythatthecubicsplinefitstoeachofthepoints(D.5a),iscontinuous(D.5b),and
iscontinuousinslopeandcurvature(D.5candD.5d)throughouttheregionspannedbythe
points.
Iftherearen+1points,thenumberofintervalsandthenumberof g i ( x ) saren.Thereare
thusfourtimesnunknowns,whicharethe {ai , bi , ci , di } for i = 0,1,..., n 1 .Eqt.(D.5a)
immediatelyyields
d i = yi , i = 0,1,..., n 1 (D.6)
442
Appendices
Eqt.D.5bgives
yi +1 = ai ( x xi ) + bi ( x xi ) + ci ( x xi ) + yi
3 2
(D.7)
= ai hi 3 + bi hi 2 + ci hi + yi , i = 0,1,..., n 1
where hi = xi +1 xi ,thewidthoftheithinterval.
To relate the slopes and curvatures of the joining splines, we differentiate Eqt. D.4 with
respecttox:
g 'i ( x ) = 3ai hi 2 + 2bi hi + ci (D.8)
g "i ( x ) = 6ai hi + 2bi , for i = 0,1,..., n 1 (D.9)
Denoting the second derivative by S such that S i = g "i ( xi ) for i = 0,1,..., n 1 and
S i = g "i ( xi ) ,applyEqt.D.9
Si = 6ai ( xi xi ) + 2bi
= 2bi
Si +1 = 6ai ( xi +1 xi ) + 2bi
= 6ai hi + 2bi
Thuswecanwrite
Si
bi = (D.10)
2
Si +1 Si
ai = (D.11)
6hi
Substitutetherelationsforai,bianddigivenbyEqts.D.6,D.10andD.11intoEqt.D.4and
thensolveforci:
Si +1 Si 3 Si 2
yi +1 = hi + hi + ci hi + yi
6hi 2
yi +1 yi 2hi Si + hi Si +1
ci =
hi 6
We now apply the constraints that the slopes of the two cubics that join at [xi,yi] are the
same.Fortheequationintheithinterval,Eqt.D.5cbecomeswithx=xi,
Inthepreviousinterval,fromxi1toxi,theslopeatitsrightendwillbe
443
Appendices
(D.13)
= 3ai 1hi 12 + 2bi 1hi 1 + ci 1
Equating D.12 and D.13 and substituting for a, b, c and d their relationships in terms of S
andy,thensimplifyingweget
yi +1 yi yi yi 1
hi 1 Si 1 + ( 2hi 1 + 2hi ) Si + hi Si +1 = 6
hi hi 1 (D.14)
= 6 ( f [ xi , xi +1 ] f [ xi 1 , xi ])
Eqt.D.14appliestoeachinternalpoint,fromi=1toi=n1.Thisgivesn1equationsrelating
n+1valuesofSi.WegettwoadditionalequationsinvolvingS0andSnbyapplyingspecific
conditionspertainingtotheendintervalsofthewholecurve.OneconditionistotakeS0as
alinearextrapolationfromS1andS2 andSnasalinearextrapolationfromSn1andSn2.We
usethefollowingrelations:
S1 S 0 S 2 S1 ( h0 + h1 ) S1 h0 S2
Attheleftend: = , S0 = (D.15)
h0 h1 h1
S n S n 1 S n 1 S n 2 ( hn 2 + hn1 ) Sn 1 hn1Sn 2
Attherightend: = , Sn =
hn 1 hn 2 hn 2
(D.16)
ApplyingEqt.D.14toeachoftheinternalpoints,fromi=1toi=n1resultsinasystemof
equationsthatmaybewritteninmatrixformasfollows:
S 0
h 2 ( h0 + h1 ) S f [x , x ] f [x , x ]
h1
0
1 1 2 0 1
h 2 ( h1 + h2 ) h2 S f [x , x ] f [x , x ]
S = 6 f [x , x ] f [x , x ]
2 2 3 1 2
2 ( h2 + h3 )
h2 h3
3 3 4 2 3
% # #
hn 2 2 ( hn 2 + hn 1 ) hn 1 S n 1 f [ x , x ] f [ x , x ]
n 1 n n2 n 1
Sn
(D.17)
In this matrix array, there are only n1 equations for n+1 unknowns. But S0 and Sn can be
eliminatedbyusingEqts.D.15andD.16.Thecoefficientmatrixthenbecomes
444
Appendices
( h0 + h1 )( h0 + 2h1 ) h1 h0
2 2
h1 h
h1 2 ( h1 + h2 ) h2
h2 2 ( h2 + h3 ) h3
%
hn 2 hn 1
2 2
( hn 1 + hn 2 )( hn 1 + 2hn 2 )
hn 2 hn 2
(D.18)
AftersolvingforS1,S2,S3,Sn1usingthecoefficientmatrixabove,S0andSnarecomputed
usingEqts.D.15andD.16.Thencoefficientsai,bi,cianddiforthecubicsineachintervalare
determinedfrom
Si +1 Si
ai =
6hi
Si
bi =
2
yi +1 yi 2hi Si + hi Si +1
ci =
hi 6
d i = yi
Fromthese,variouspointsontheinterpolatingcurvemaybecomputed.
445
Appendices
AppendixEVortexFilamentStretching
Becauseinafreewakesolutionthepointsinthefluidflowfieldareallowedtoconvect
freely, the vortex filaments will be strained as the adjacent nodes will move relative to
eachother.Thiscausesachangeinvorticityoftheindividualfilamentswhichinturnresults
in changes of the induced velocity field around the vortex core. This effect is illustrated
pictoriallyinFig.E.1.
Swirlvelocity Swirlvelocity
l+l
l
rc rcrc
Figure E.1 Schematic showing the effect of vortex filament stretching: The vorticity increases
resultinginahigherswirlvelocity.
Ananthanetal.[2]presentthefollowingcorrectionforfilamentsstretchingeffects:
Consideravortexfilamentwithitslocationinspacedefinedintimestep bytwoadjacent
G G
nodeshavingpositionvectors ra and ra 1 asshowninFig.E.2.
Time
G G
ra 1 ra
G G
ra 2 ra +1
l +
Time + G G
ra 1 ra
G
ra +1
G
ra 2
FigureE.2Schematicrepresentationofstraininginavortexfilament
446
Appendices
Becausethefilamentisastraightlinesegment,itslengthisgivenby
G G
l = ra ra 1 (E.1)
Therateofchangeofthelengthofthefilamentasitconvectsthroughthevelocityfieldis
givenby
G G
dl d ( ra ra 1 ) G G
= = Va Va 1
d d
Thenewlengthofthevortexfilamentatthenexttimestep( + ) isthengivenby
G G
l + = Va Va 1 * (E.2)
The filament strain experienced by the vortex filament over time interval can then be
evaluatedfrom
(l + l )
= (E.3)
l
BecauseHelmhotzsthirdlawrequiresthatthenetstrengthofanyvortexfilamentshould
remainconstant,theproductofthevorticityanditscrosssectionalareashouldalsoremain
constant. To find the effect of strain on vorticity, the vorticity in the filament can be
assumed to be concentrated inside a cylinder of length l, with an effective core radius rc.
Since the flow field considered has a constant density, the principle conservation of mass
canbeusedtodeterminetheeffectivechangeinthecoreradiusasaresultofthechangein
filamentlength,l.Theprincipleofmomentumconservedisalsoimplicitlyobeyed.Thus,
rc l = (rc rc ) 2 (l + l )
2
2
rc rc l
=
rc l + l
1
rc = rc 1 (E.4)
1+
Toincludetheeffectoffilamentstrainingintothewakesolution,aninitialcoreradiusmust
becalculatedbasedonviscousdiffusionbehaviour.Thenthechangeincoreradiusdueto
strainingeffectsiscomputedusing
1
rceff = rc (E.5)
1+
447
Appendices
448
CurriculumVitae
TonioSantwasborninMaltaon13thJanuaryin1977.Hecompletedhisprimaryeducation
atStellaMarisSchool,BalzanandhissecondaryeducationaSt.AloysuisCollege,BKara.
He started hisstudies inEngineering at theUniversityofMalta in 1995and graduated in
MechanicalEngineeringin1999withfirstclasshonours(summacumlaude).Histhesiswas
titled Investigating the Aerodynamic Behaviour of Yawed HorizontalAxis Wind Turbines and
wasawardedasthebestmechanicalengineeringprojectforthatyear.
In July 1999, Tonio was employed with STMicroelectronics (Malta) Ltd. as a process
engineerinoneoftheproductionlinesmanufacturingintegratedelectronicchips.InMarch
2000,hejoinedAirMaltaasanaircraftmaintenanceengineer.Onemployment,hefollowed
courses and handson training on aircraft maintenance atAir Maltasengineering base, at
Lufthansa Technik in Frankfurt and at the Boeing Training Centre, in Seattle. On
completingthetraining,hisworkfocusedontheinspectionandmaintenanceofairplanes
structuresofAirMaltasBoeingB737andAirbusA320fleet.Furthermore,Toniowasalso
responsible for organising courses for airplane mechanics related to airplane structural
inspectionandrepair.
In2002,ToniowasengagedtemporarilyinthenewlyformedcompanyLufthansaTechnik
Malta where he provided technical support in airplane maintenance planning and in
trainingnewtechnicalstaff.
During his Phd studies, Tonio also had a number of duties within the Faculty of
Engineering at the University of Malta. Hewas responsible for teaching heat transfer, fluid
mechanics and turbomachinery (including wind energy) to students following the B.Eng.
(Hons.) course. He also supervised four final year projects related to wind turbine
aerodynamics.Hewasinvolvedintheinstallationandperformancetestingofasmallscale
gridconnectedwindturbineattheUniversityofMalta.
Toniosmaininterestsarewindenergy,aerospaceengineering,theologyandnature.Heis
marriedtoMarisandtheyliveinMalta.
Email:tonio.sant@um.edu.mt
449
AuthorPublications
450
Experiments, Proceedings from the 25th Wind Energy Symposium and the 44th AIAA
AerospaceMeeting,Reno,NV,USA,Jan.2006.
11. HaansW.,SantT.,vanKuikG.A.M.andvanBusselG.J.W,StallinYawedFlow
Conditions a correlation of Blade Element Momentum Predictions to
Experiments, Journal of Solar Energy Engineering, Vol. 128, No. 4, Nov. 2006, pp.
472480.
451