Beruflich Dokumente
Kultur Dokumente
edings of the
e 1st Global P
Power and Propulsion
P Fo
orum
GPPF 2017
Jan 16-118, 2014, Zu
urich, Switzerrland
www.pps.global
GPPF
F-2017--51
A CO
OMBINED EXPERIM
MENTAL A
AND NUME ERICAL AN
NALYSIS OF THE WAKE
W
STRUCCTURE AN
ND PERFO
ORMANCEE OF A H-S
SHAPED DARRIEUS
D S WIND TU
URBINE
T work is licen
This nsed under a Creaative Commons Attribution
A 4.0 Intternational Licensse
an extended experimental reseaarch program on VAWTs w was h a swept arrea of 1.5 m2 (H x D, 1.55 m x
The rotor has
carriied out in colllaboration witth the Universsity of Trentoo, in 1.0
028 m), and itt is equipped with dihedrall blades compposed
the frame
f of a N
National foundded project, whose
w results are by un-staggered d NACA00211 profiles (c = 0.086 m). Full
partiially documennted in Dosseena et al [7] anda Persico eet a. dettails on the lab
boratory modeel are reported
d in [7].
[8]. Throughout tthis project, tw wo rotor archiitectures (clossed-
tropooskien-blade and open-H-bblade) were tested in seveeral EXXPERIMENT TAL CAMPAIIGN
w conditions, with the aim to give insights into the
flow In this papeer, CFD simullations are thooroughly comppared
diffeerent aerodynnamics of thee two config gurations andd to agaainst wind tu unnel measureements perforrmed some of o the
makke results avaailable as an experimentall benchmark for autthors in a previous, but reccent, experimeental campaiggn. In
futurre projects onn similar maachines. In th he present stuudy, thee frame of this experimentaal activity, aerodynamic ass well
than
nks to the availlability of succh a unique set of experimenntal as performance measurementss were carried d out to investtigate
data, a systematicc cross-compaarison was carrried out betw ween thee turbine wak ke and perforrmance. A brrief review of o the
the experimental results of thhe H-shaped rotor r at the mmid expperimental laayout is repoorted in the following secction,
plan
ne and the numerical ones o comingg from a tw wo- whhile full detailss on the test caampaign can be
b found in [77].
dimeensional U-R RANS computtational modeel developed by
some of the authoors in the last few years (e..g. Balduzzi eet al Wiind Tunnel
[9]). The main aeerodynamic phhenomena aree highlighted and The measu urements repported in this t paper were
the prediction
p cappability of the numerical moodel is discusssed. perrformed in thee large-scale w wind tunnel off the Politecniico di
One may nootice that thee use of two-dimensional U- Miilano (Italy). It consists of a closed-loop facility, whose test
RAN NS simulationns is probabbly not the most m approprriate secction (4.00 m wide, 3.84 m high and 6.00m long) alllows
apprroach to this kind of analyyses since 2D D simulations are thee aerodynamicc characterizaation of real-scale wind turb rbines
deemmed to producce vortices thhat are more coherent
c and lless forr micro-generration, up too a maximum m wind speeed of
pronne to breakdow wns than real 3D ones. Thee present reseaarch 55 m/s. A comb bination of ho
honeycombs and a anti-turbulence
on Darrieus
D VAW WTs is indeed trying to move the C CFD scrreens providess high flow qu quality, resultiing in a freesttream
simuulations towarrds fully threee-dimensionall analyses, whhich turrbulence intensity at the inleet lower than 1%.
are however stilll hampered by b the enorm mous calculattion Since the soolid blockage of the rotor amounts
a to 100% of
resouurces needed.. On the otherr hand, recent studies (e.g. [6]) thee wind tunnel chamber, a relevant blo ockage effectt was
showwed that the appproach here proposed can n give an accurrate observed in conventional w wind tunnel teests performeed in
mation of the aerodynamiccs of these machines,
estim m evenn in con nfined (or cllosed chambeer) configuraation. To limiit the
term
ms of instantanneous torque profile. On theset bases, iit is bloockage, furtheer tests were ccarried out using a ''free jett'' (or
heree thought to bbe also exploiitable to assesss the main fl flow op pen chamber) configuratio ion, by remov ving the test room
strucctures attendedd in the wake of a H-Darrieeus rotor. upstream of the diffuser leadin ing to the fanss, and installinng the
rottor directly facing
f the uppstream tunneel. Applicatioon of
VAW WT TEST MO ODEL claassical correlaations for freee-jet blockagee suggests thaat the
The turbine pprototype conssidered in thiss study is a thrree- bloockage is neg gligible in thee present casee (below a few w %)
bladde, H-shaped D Darrieus VAW WT. The blad des are conneccted tunnel results are believed to be
[7]]; therefore, frree-jet wind tu
to th
he support steeel pole by twoo aluminium spokes,
s as shoown reppresentative of the open fie ield configuraation, and are here
in Fiig. 1. con nsidered for comparison wiith CFD simullations.
2
direction, thus exposing the wire normal to the wind to To simulate the rotation of the turbine, the sliding mesh
measure directly the velocity magnitude. The second probe technique was employed, i.e. the computed domain was split
was mounted in the vertical direction and was rotated during into a circular zone containing the turbine, rotating with the
operation around its own axis to measure the transversal same angular velocity of the rotor, and an outer rectangular
velocity components, by exploiting the angular sensitivity of fixed zone, determining the overall domain extent. The final
wire. The so-called triple decomposition was performed on dimensions of both domains were defined, according to the
hot wire data, by extracting the time-averaged, phase- sensitivity analysis reported by Balduzzi et al [9], so to allow
resolved, and turbulent components of the velocity, as a full and unconstrained development of the turbine wake. In
discussed by Persico et al [8]. The phase-resolved component this way, CFD results can be properly compared to
was obtained by ensemble averaging the velocity signals, experiments obtained in unconfined configuration. A
using the encoder as a key-phasor. The streamwise velocity-inlet boundary condition was imposed at the inlet
turbulence intensity was determined by extracting the section (40D upwind from the revolution axis). The ambient
unresolved unsteady fluctuations of the velocity signal. A pressure condition was imposed at the outlet boundary (100D
pneumatic five-hole probe was also applied throughout the downwind), while a symmetry condition was defined on
tests to measure the pressure level and the 3D flow direction. lateral boundaries (30D).
Measured pressure values confirmed the absence of In the present study, the finest mesh defined during the
blockage-induced over-speeds outside the wake region; flow sensitivity analyses reported by Balduzzi et al [12] was used
angle measurements with the vertical hot wire and the for all the three considered tip-speed ratios. Even if
pneumatic probe also indicated negligible transversal and redundant at TSR=3.3, this refinement level indeed ensured a
vertical velocity components in the wake at midspan. very accurate discretization of the whole area around the
rotor, greatly limiting any bias error due to the numerical
NUMERICAL SIMULATIONS discretization. The mesh is of unstructured and hybrid type,
The two-dimensional simulations presented in this paper with triangular elements in the core flow region, and an O-
were carried out using the commercial code ANSYS grid made of quads in the boundary layer. As discussed, the
FLUENT [10]. The time-dependent unsteady Reynolds- first element height was chosen to guarantee that the
averaged NavierStokes (U-RANS) approach was adopted in dimensionless wall distance at the grid nodes of the first
a pressure-based formulation. The fluid was air, which was layer above the blade wall does not exceed y+ ~ 1. The
modeled as an ideal compressible fluid following the expansion ratio for the growth of elements starting from the
suggestion of Balduzzi et al [9]. Inlet air conditions were the surface was kept below 1.1 to achieve good mesh quality.
same monitored during the experimental tests, i.e. a pressure The airfoil surface was discretized with 1400 nodes, resulting
of 1.01105 Pa, a temperature of 293 K and an inlet in 1.2106 elements for the rotating region, while the
turbulence level of 1%. Some of the authors have recently stationary region was discretized with 2.0105 elements. The
presented (Balduzzi et al [9]) the assessment and validation mesh characteristics fully accomplish the criteria based on
of the main settings for a proper CFD simulation of Darrieus dimensionless thresholds proposed by Balduzzi et al [16].
wind turbines using the FLUENT code, whose accuracy has Based on the same criteria, in order to limit the Courant
been also successfully verified by means of experimental Number in proximity of the blades, an angular timestep of
data (Balduzzi et al [11]). Moreover, in recent works 0.21 was used, i.e. the finest one identified by Bianchini et
(Balduzzi et al [12] and Bianchini et al [13]) the accuracy of al in [13]. With the described settings, the calculation time to
the proposed numerical approach has been verified exactly achieve a complete revolution of the rotor is around 24 hours
on the same case study considered here. For the sake of in a 16 CPUs (2.8 MHz each) calculation center. The
completeness, the main numerical settings are briefly required number of revolutions to achieve a periodic solution
summarized below. The turbulence closure was achieved by is dependent on the TSR, varying from 10 to 20 revolutions.
means of the k- shear stress transport (SST) model (Menter Figure 2 finally shows some visual details of the mesh in
[14]), coupled with the Enhanced Wall Treatment embedded the rotating region (a) and near the leading edge (b).
in the FLUENT code for the computation of the boundary
layer in the near-wall regions; since the y+ was constantly RESULTS AND DISCUSSION
kept < 1 in the boundary layer, the linear law of the wall is To assess the prediction capability of the CFD approach,
actually used. Fig. 3 compares the curve of experimental power coefficient
The Coupled algorithm was employed to handle the cP, defined according to Eq. 1 and measured by Dossena et al
pressure-velocity coupling. The second order upwind scheme [7], with the one obtained with present simulations.
was used for the spatial discretization of the whole set of
P
RANS and turbulence equations, as well as the bounded cP = . (1)
1 3
second order for time differencing to obtain a good AV0
resolution (Amet et al [15]). The global convergence of each 2
simulation was monitored by considering the difference It is worth remarking that numerical results were
between the mean values of the torque coefficient over two corrected to account for the parasitic torque of the supporting
subsequent revolutions: according to Balduzzi et al [11], the struts of the rotor, which were of course not simulated in the
periodicity error threshold was set to 0.1%. two-dimensional approach.
3
Daarrieus VAWT Ts, can be nooted at low tip-speed ratioos; in
this region of the curve, the airfoils work in sttalled
connditions for a considerable fraction of th he revolution, with
sevvere flow sep paration phenoomena and laarge-scale vorrtices
dettaching even from the leaading edges of the airfoills. In
theese condition ns, additionaal elements like the suurface
rouughness, the trailing-edge
t refinement in
n the experimmental
moodel, etc. can n play a funndamental ro ole in settingg the
chaaracteristics of such sepparations, wh hich are how wever
piv
votal for the final
f torque prroduction. Ovverall, by virtuue of
theese results, th
he CFD approoach can be considered
c relliable
forr the purposes of the presennt study.
To analyse the wake charracteristics, th hree relevant TSRs
T
weere selected for analysiis, for whicch detailed flow
meeasurements werew availablee. Based on the indicationns of
[7]], the tip-speeed ratios of 1..5, 2.4 and 3.3 were considdered
forr comparison n, correspondding to a wind w speed V0 of
14.2 m/s, 9.0 m/s and 6.5 m//s, respectively y, having assiigned
a constant
c angular speed of thhe turbine equal to 400 rpm..
Figure 4 compares the CFD-based torque coeffiicient
Figu
ure 2 - Mesh
h refinement in the rota
ating region (a) pro hree investigated conditions.
ofiles, cT (Eq.. 2), for the thr
and near the lea
ading edge (b).
T
cT = . (2)
D 2
A V 0
4
4
Finally, TSR R=3.3 repreesents a staable functionning Upon exammination of thee figure, one can c notice thaat the
conddition, with noo evidence of abrupt stall, even
e if the torrque firsst two ripples correspond to the onset an nd the propaggation
extraaction is stronngly unbalancced towards the upwind zoone, of strong maccro vortices connected to o the deep stall
as ty
ypically foundd for three-blaade rotors of similar
s solidityy. exp perienced by the
t airfoils, allready discusssed in Fig. 4.
To understannd the physicss underlying these trends, the The third ripple (locatted approxim mately behindd the
posssibility of having experimenntal measurem ments of the fl flow tow
wer) is insteaad connected to another seet of vorticess that
field
d becomes ccrucial In particular,
p forr all the thhree oriiginate from the airfoil ttrailing edgee as soon ass the
funcctioning pointts comparisoons were maade between the inccidence angle is decreasinng again towaards values which w
nummerical resultts and the experimentss in terms of alloow the flow to reattach tto the bladess. Both thesee two
dimeensionless vellocity (V/V0) and turbulencce intensity (IITU). phenomena havee a major imppact on the ov verall perform
mance
CFD D data have bbeen acquiredd in correspon ndence of virt rtual of the rotor. Fully comparablee trends were also found foor the
rakees positioned eexactly at thee same distancce from the shhaft farr traverse (Fiig. 7), with the main flo ow structures still
of th
he experimenttal traverses, i.e. near 0.7 75D downstreeam notticeable, even n though witth a general reduction of the
and far 1.5D ddownstream thhe shaft. At each e section, 100 flu
uctuations intensity.
equaally spaced pooints were acqquired at a freqquency of 10 kkHz Consideringg the conditioon corresponding to TSR R=2.4,
for at least 3 revolutions. Baased on thesee data, averaaged Figg. 8 reports thee results againn for the near traverse.
proffiles were alsoo calculated annd added to the comparisonss.
Figure 5 repports the resuults at TSR=1.5 for the nnear
traveerse, in the foorm of measurred and compu uted velocity and
turbuulence profilees. Super-impoosed to the com mputed averaaged
proffiles, the unssteady dispersion is also reported. Goood
agreeement is geenerally founnd between simulations and
expeeriments, in teerms of wake width and defficit as well aas in
ms of flow and turbulence morphology inside the waake.
term
Fromm an in-depth analysis of thhe trends, it is worth remarkking
that CFD data present a high dispersion, esspecially in thhree
posittions, i.e. arouund Y/D=-0.44, Y/D=-0.2 an nd 0.1 <Y/D<00.4.
In th
hese positions, both the velocity and the turbulence fieelds
present wide ripples, whose naature can be comprehended
c d by
lookking at the com ose distributionn is
mputed velociity field, who Figure
F 6 - Ins
stantaneouss velocity fie
eld at TSR=1
1.5.
repoorted in Fig. 6 for a given innstant.
Figure
F 7 - Far traverse: d
dimensionle ess velocity
y (a)
Figure 5 - Nearr traverse: dimensionle
d ss velocity (a) and turbbulence inteensity (b) att TSR=1.5.
and turbu ulence intennsity (b) at TSR=1.5.
T
5
In this case, tthe velocity prrofile is smooother and the ddata
dispersion is low wer. In furtheer detail, as testified by the
snappshot of the innstantaneous velocity field d reported in F Fig.
9, a more uniforrm wake struccture was preedicted by CF FD.
Even n if the overaall accuracy of o simulationss is still goodd in
term
ms of wake widdth, a discrepaancy can be heere noticed inn the
regioon behind thhe rotor, whhere experim ments revealedd a
velocity deficit hiigher than thaat predicted nu umerically. S uch
deviiation might be connectedd to a higheer impact of the
interraction betweeen the flow annd the tower. This evidencce is
in aggreement withh the slightlyy lower value of the prediccted
powwer coefficient previously diiscussed in Fig g. 3.
A general deecrease of thee turbulence level can be aalso Figure
F 9 - Ins
stantaneouss velocity fie
eld at TSR=2
2.4.
noticced in Fig.8((b). With resspect to the lower TSR, the
turbuulence intensiity profile preedicted by CF FD is almost flat
and of low magnnitude in the center c of the wake, with twtwo,
clearrly marked, peaks at the tw wo sides of thee wake itself; the
posittion and relaative intensityy of these peeaks is againn in
remaarkable agreeement withh experimen nts. In tunnnel
measurements, a further peak is however distinguishable
d e at
the center
c of the wwake, again suggesting a sttronger influeence
of th
he tower.
To better undderstand the nature
n of the tw
wo lateral peaaks,
Fig. 10 reports thhe instantaneoous distributio on of turbuleence
inten
nsity in the sam me instant in time of Fig. 9.
9
6
Fig
gure 11 - Farr traverse: dimensionles
d ss velocity ((a) Fiigure 13 - Fa
ar traverse: dimensionless velocity y (a)
and turbu ulence intennsity (b) at TSR=2.4.
T and turbbulence inteensity (b) att TSR=3.3.
CO ONCLUSION NS
In the present study, a comprehen nsive set of flow
meeasurements inside
i the waake of a Darrrieus verticall-axis
win nd turbine have
h been coompared with h two-dimenssional
CFFD simulations of the same rotor in the midm plane.
Overall, ann interesting aagreement was found bettween
meeasurements and a simulatioons, even iff a 2D U-R RANS
app proach is usuaally thought to be not fullyy suitable for these
anaalyses. In parrticular, the CCFD data weree able to corrrectly
preedict the perfoormance of thhe turbine and d also to highhlight
som me main flo ow structurees whose efffects were fully
com mpatible with experimentall trends.
In particulaar, it has beenn shown thatt at low TSR Rs the
waake is charactterized by veery intense riipples, due too the
passsage of macro vortices dettaching from thet blades eithher in
thee upwind regiion after the sstall angle is reached, or in the
dow wnwind regio on as soon as tthe flow tendss to re-attach to
t the
blaade. For an inttermediate TSSR, intense turrbulence and some
Fig
gure 12 - Neear traverse: dimension nless velocitty rip
pples are still clearly distingguishable in reason
r of the wide
(a) and turrbulence inte
ensity (b) att TSR=3.3. rannge of inciden nce angles expperienced by the airfoils, which
w
maay lead to stall onset. Finallly, for high TSRs, typical ofo the
staable part of thee functioning curve, the waake is more reegular
and d smoother wiith less intensse ripples and turbulence, evven if
7
the velocity deficit is now increased due to the higher load validation with particle image velocimetry data, Wind
experienced by the turbine. Energy, 13, 2010, pp. 1-17
[6] Rainbird. J, Bianchini, A, Balduzzi, F, Peiro, J,
NOMENCLATURE Graham, JMR, Ferrara, G, and Ferrari, L, On the Influence
A turbine swept area [m2] of Virtual Camber Effect on Airfoil Polars for Use in
cP power coefficient [-] Simulations of Darrieus Wind Turbines, Energy Conversion
cT torque coefficient [-] and Management, 106(December 2015), 2015, pp. 373-384
CFD Computational Fluid Dynamics [7] Dossena, V, Persico, G, Paradiso, B, Battisti, L,
D, R turbine diameter, radius [m] DellAnna, S, Brighenti, A and Benini, E, An Experimental
ITU turbulence intensity [m] Study of the Aerodynamics and Performance of a Vertical
k turbulent kinetic energy [m2s-2] Axis Wind Turbine in a Confined and Unconfined
P power [W] Environment, 2015, ASME Journal of Energy Resources
RANS Reynolds-Averaged Navier-Stokes Technology, 137(5) 2015, paper code 051207 (12 pages)
SST Shear Stress Transport [8] Persico, G, Dossena, V, Paradiso, B, Battisti, L,
T torque [Nm] Brighenti, A and Benini, E, Three-Dimensional Character of
TSR Tip-Speed Ratio VAWT Wakes: an Experimental Investigation for H-Shaped
V wind speed [m/s] and Troposkien Architectures, Proc. of the ASME Turbo
VAWT Vertical Axis Wind Turbines Expo 2016, Seoul, South Korea, June 13-17, 2016.
y+ dimensionless wall distance [-] [9] Balduzzi, F, Bianchini, A, Maleci, R, Ferrara, G and
Ferrari, L, Critical issues in the CFD simulation of Darrieus
Greek letters
wind turbines, Renewable Energy, 85(01), 2016, pp. 419-
azimuthal angle [deg]
435
air density [kg/m3]
[10] ANSYS, Inc., 2015, FLUENT Theory Guide,
Specific turbulence dissipation rate [s-1]
release 16.0
revolution speed [rpm]
[11] Balduzzi, F, Bianchini, A, Maleci, R, Ferrara, G, and
Subscripts Ferrari, L, Blade Design Criteria to Compensate the Flow
0 value at inlet Curvature Effects in H-Darrieus Wind Turbines, J. of
Turbomachinery, 137(1), 201, pp. 011006-011006-10
ACKNOWLEDGMENTS [12] Balduzzi, F, Bianchini, A, Gigante, FA, Ferrara, G,
Thanks are due to Prof. Ennio Antonio Carnevale of the Campobasso, MS, and Ferrari, L, Parametric and
University of Florence for supporting this study, and to Dr. Comparative Assessment of Navier-Stokes CFD
Andrea Tanganelli for his support in the CFD simulations. Methodologies for Darrieus Wind Turbine Performance
The authors would like also to acknowledge the Analysis, Proc. of the ASME Turbo Expo 2015, Montreal,
Company Tozzi-Nord Wind Turbines for the rotor model and Canada, June 15-19, 2015
the technicians and collaborators of Universit di Trento and [13] Bianchini, A, Balduzzi, F, Ferrara, G, and Ferrari, L,
Politecnico di Milano for their precious support before and Aerodynamics of Darrieus wind turbines airfoils: the impact
during the tests. of pitching moment, J. of Engineering for Gas Turbines and
Power, 2016, paper in press
REFERENCES [14] Menter, F, Two-equation Turbulence-models for
[1] Balduzzi, F, Bianchini, A, Carnevale, EA, Ferrari, L Engineering Applications, AIAA Journal, 32(8), 1994, pp.
and Magnani, S, Feasibility analysis of a Darrieus vertical- 15981605
axis wind turbine installation in the rooftop of a building, [15] Amet, E, Matre, T, Pellone, C, and Achard, JL, 2D
Applied Energy, 97, 2012, pp. 921929 numerical simulations of blade-vortex interaction in a
[2] Borg, M, Shires, A, and Collu, M, Offshore floating Darrieus turbine, J. of Fluids Engineering, 131(11), 2009,
vertical axis wind turbines, dynamics modelling state of the pp. 1-15
art. Part I: Aerodynamics, Renewable and Sustainable [16] Balduzzi, F, Bianchini, A, Ferrara, G, and Ferrari, L,
Energy Reviews, 39, 2014, pp. 1214-1225 Dimensionless numbers for the assessment of mesh and
[3] Paraschivoiu, I, Wind turbine design with emphasis on timestep requirements in CFD simulations of Darrieus wind
Darrieus concept, Polytechnic International Press, Montreal, turbines, Energy, 97(15 February 2016), 2016, pp. 246-261
Canada, 2002. [17] Bianchini, A, Ferrari, L and Magnani, S, Start-up
[4] Aslam Bhutta, MM, Hayat, N, Farooq, AU, Ali, Z, behavior of a three-bladed H-Darrieus VAWT: experimental
Jamil, ShR, and Hussain, Z, Vertical axis wind turbine A and numerical analysis, Proc. of the ASME Turbo Expo
review of various configurations and design techniques, 2011, Vancouver, Canada, June 6-10, 2011.
Renewable and Sustainable Energy Reviews, 16(4), [18] Bianchini, A, Balduzzi, F, Ferrara, G and Ferrari, L,
2012,1926-1939 Influence of the blade-spoke connection point on the
[5] Simao-Ferreira, C, van Zuijlen, A, Bijl, H, van Bussel, aerodynamic performance of Darrieus wind turbines, Proc.
G, and van Kuik, G, Simulating dynamic stall on a two- of the Turbo Expo 2016, Seoul, South Korea, June 13-17,
dimensional vertical-axis wind turbine: verication and 2016.