Beruflich Dokumente
Kultur Dokumente
1.P.SauerandM.Pai,PowerSystemSteadyStateStabilityandtheLoadFlowJacobian,IEEE
TransactionsonPowerSystems,Vol.5,No.4,Nov.1990
2.V.AjjarapuandC.Christy,TheContinuationPowerFlow:AToolforSteadyStateVoltage
StabilityAnalysis,IEEETransactionsonPowerSystems,Vol.7,No.1,Feb.,1992.
3.S.Greene,I.Dobson,andF.Alvarado,SensitivityoftheLoadingMargintoVoltageCollapse
withRespecttoArbitraryParameters,IEEETransactionsonPowerSystems,Vol.12,No.1,
Feb.1997,pp.232240.
4.S.Greene,I.Dobson,andF.Alvarado,ContingencyRankingforVoltageCollapsevia
SensitivitiesfromaSingleNoseCurve,IEEETransactionsonPowerSystems,Vol.14,No.1,
Feb.1999,pp.262272.
Voltage Security
Voltagesecurityistheabilityofthesystemtomaintain
adequateandcontrollablevoltagelevelsatallsystemloadbuses.
Themainconcernisthatvoltagelevelsoutsideofaspecified
range can affect the operation of the customerssloads.
rangecanaffecttheoperationofthecustomer
loads
Voltagesecuritymaybedividedintotwomainproblems:
1 L
1.Lowvoltage:voltagelevelisoutsideofpredefinedrange.
lt
lt
l li
t id f
d fi d
2.Voltageinstability:anuncontrolledvoltagedecline.
Youshouldknowthat
lowvoltagedoesnotnecessarilyimplyvoltageinstability
nolowvoltagedoesnotnecessarilyimplyvoltagestability
no low voltage does not necessarily imply voltage stability
voltageinstabilitydoesnecessarilyimplylowvoltage
Resources
Therehavebeenseveralindividualsthathavesignificantly
Th
h
b
l i di id l th t h
i ifi tl
progressedthefieldofvoltagesecurity.Theseinclude:
AjjarapufromISU
VanCutsem:SeethebookbyVanCutsemandVournas.
Van Cutsem: See the book by Van Cutsem and Vournas.
Alvarado,Dobson,Canizares,&Greene:
Thereareacoupleothertextsthatprovidegoodtreatmentsof
the subject:
thesubject:
CarsonTaylor:PowerSystemVoltageStability
PrabhaKundur:PowerSystemStability&Control
3
Ourtreatmentofvoltagesecuritywillproceedasfollows:
Voltageinstabilityinasimplesystem
Voltageinstabilityinalargesystem
Brieftreatmentofbifurcationanalysis
Continuationpowerflow(pathfollowing)methods
Sensitivity methods
Sensitivitymethods
Voltageinstabilityinasimplesystem
Considertheperphaseequivalentofaverysimplethree
phasepowersystemgivenbelow:
V1
V2
Node1
+
Z=R+jX
I
Node2
+
V2
V1
_
S12
SD=S12
Z = R + jjX Y = G jjB
NoteB>0
P12 =| V1 || V2 | B sin(1 2 )
Q12 =| V1 |2 B | V1 || V2 | B cos(1 2 )
NowwecangetSD=PD+jQD=(P21+jQ21)by
exchangingthe1and2subscriptsinthepreviousequations.
negating
g
g
PD = P21 = | V1 || V2 | B sin( 2 1 )
=| V1 || V2 | B sin(1 2 )
QD = Q21 = | V2 |2 B + | V1 || V2 | B cos( 2 1 )
= | V2 |2 B + | V1 || V2 | B cos(1 2 )
Define12 =1 2
PD =| V1 || V2 | B sin 12
QD = | V2 |2 B + | V1 || V2 | B cos 12
Define:
isthepowerfactorangleoftheload,i.e.,
p
g
= V2 I
ThenwecanalsoexpressS
jD as:
*
S D = V2 I =| V2 || I | e
=| V2 || I | (cos + j sin )
sin
=| V2 || I | cos (1 + j
)
cos
= PD (1 + j tan )
Define=tan.Then
S D = PD + jQD = PD (1 + j )
Notethatphi,and
thereforebeta,is
positiveforlagging,
negativeforleading.
ti f l di
8
Sowehavedevelopedthefollowingequations.
PD =| V1 || V2 | B sin 12
QD = | V2 |2 B + | V1 || V2 | B cos12
S D = PD + jQD = PD (1 + j )
EquatingtheexpressionsforPD andforQD,wehave:
QD = PD = | V2 |2 B + | V1 || V2 | B cos12
PD =| V1 || V2 | B sin 12
PD + | V2 |2 B =| V1 || V2 | B cos12
Squarebothequationsandaddthemtoget..
PD + ( PD + | V2 |2 B ) 2 =| V1 |2 | V2 |2 B 2 ((sin 2 12 + cos 2 12 )
2
PD + ( PD + | V2 |2 B ) 2 =| V1 |2 | V2 |2 B 2
2
Manipulationyields:
(| V | )
2 2
PD
2 PD
2
2
+
| V1 | | V2 | + 2 1 + 2 = 0
B
B
Notethatthisisaquadraticin|V2|2.Assuch,ithasthesolution:
| V1 |
PD | V1 | PD PD
2
| V2 | =
+ | V1 |
2
B
B B
4
2
1/ 2
10
Letsassumethatthesendingendvoltageis|V1|=1.0pu
andB=2pu.Thenourpreviousequationbecomes:
2 1 PD [1 PD ( PD + 2 ) ]
| V2 | =
2
1/ 2
Youcanmake
You
can make
thePVplotusing
thefollowing
matlabcode.
tl b d
%pf=0.97lagging
beta=0 25
beta=0.25
pdn=[00.10.20.30.40.50.60.70.78];
v2n=sqrt((1beta.*pdn sqrt(1pdn.*(pdn+2*beta)))/2);
pdp=[0.780.70.60.50.40.30.20.10];
v2p=sqrt((1beta.*pdp+sqrt(1pdp.*(pdp+2*beta)))/2);
pd1=[pdnpdp];
v21=[v2nv2p];
[
p]
%pf=1.0
beta=0
pdn=[00.10.20.30.40.50.60.70.80.90.99];
v2n=sqrt((1beta.*pdn sqrt(1pdn.*(pdn+2*beta)))/2);
pdp=[0.990.90.70.60.50.40.30.20.10];
v2p=sqrt((1beta.*pdp+sqrt(1pdp.*(pdp+2*beta)))/2);
pd2=[pdnpdp];
d2 [ d d ]
v22=[v2nv2p];
%pf=.97leading
beta=0.25
pdn=[00.10.20.30.40.50.60.70.80.91.01.11.21.3];
v2n=sqrt((1beta.*pdn sqrt(1pdn.*(pdn+2*beta)))/2);
pdp=[1.3
pdp
[1.31.21.11.00.90.70.60.50.40.30.20.10];
1.2 1.1 1.0 0.9 0.7 0.6 0.5 0.4 0.3 0.2 0.1 0];
v2p=sqrt((1beta.*pdp+sqrt(1pdp.*(pdp+2*beta)))/2);
pd3=[pdnpdp];
v23=[v2nv2p];
plot(pd1,v21,pd2,v22,pd3,v23)
11
Plotsofthepreviousequationfordifferentpowerfactors
|V2|
R l
Realpowerloading,P
l di PD
12
SomecommentsregardingthePVcurves:
1.Eachcurvehasamaximumload.Thisvalueistypicallycalledthe
maximumsystemloadorthesystemloadability.
2.Iftheloadisincreasedbeyondtheloadability,thevoltageswill
declineuncontrollably.
3.Foravalueofloadbelowtheloadability,therearetwo
voltagesolutions.Theupperonecorrespondstoonethatcanbe
reachedinpractice.Theloweroneiscorrectmathematically,butI
p
y,
donotknowofawaytoreachthesepointsinpractice.
4.Inthelaggingorunitypowerfactorcondition,itisclearthatthe
voltagedecreasesastheloadpowerincreasesuntiltheloadability.
g
p
y
Inthiscase,thevoltageinstabilityphenomenaisdetectable,i.e.,
operatorwillbeawarethatvoltagesaredecliningbeforethe
loadability is exceeded.
loadabilityisexceeded.
5.Intheleadingcase,oneobservesthatthevoltageisflat,orperhaps
evenincreasingalittle,untiljustbeforetheloadability.Thus,in
the leading condition voltage instability is not very detectable
theleadingcondition,voltageinstabilityisnotverydetectable.
Theleadingconditionoccursduringhightransferconditionswhenthe
loadislightorwhentheloadishighlycompensated.
13
QVCurves
Weconsideroursimple(lossless)systemagain,withtheequations
PD =| V1 || V2 | B sin 12
QD = | V2 |2 B + | V1 || V2 | B cos12
Now,againassumethatV1=1.0,andforagivenvalueofPD
andV2,compute12 fromthefirstequation,andthenQfromthe
secondequation.RepeatforvariousvaluesofV2 toobtainaQV
curveforthespecifiedrealloadPD.
YoucanmakethePVplotusingthefollowingmatlabcode.
v1=1.0;
b=1.0;
pd1=0.1
v2=[1.1,1.05,1.0,.95,.90,.85,.80,.75,.70,.65,.60,.55,.50,.45,.40,.35,.30,.25,.20,.15];
sintheta=pd1./(b*v1.*v2);
theta=asin(sintheta);
qd1=v2.^2*b+v1*b*v2.*cos(theta);
plot(qd1,v2);
Thecurveonthenextpageillustrates.
14
QVCurve
|V2|
QD
15
Homework
1.DrawthePVcurveforthefollowingcases,andforeach,determinetheloadability.
a.B=2,|V1|=1.0,pf=0.97lagging
b B=2 |V1|=1 0 pf=0 95 lagging
b.B=2,|V1|=1.0,pf=0.95lagging
c.B=2,|V1|=1.06,pf=0.97lagging
d.B=10,|V1|=1.0,pf=0.97lagging
Identifytheeffectonloadabilityofpowerfactor,sendingendvoltage,andlinereactance.
2.DrawtheQVcurvesforthefollowingcases,andforeach,determinethemaximumQD.
a.B=1,|V1|=1.0,PD=0.1
b.B=1,|V1|=1.0,PD=0.2
c. B=1,
c.B
1,|V1|
|V1|=1.06,
1.06,P
PD=0.1
0.1
d.B=2,|V1|=1.0,PD=0.1
IdentifytheeffectonmaximumQD ofrealpowerdemand,sendingendvoltage,andline
reactance.
t
16
SomecommentsregardingtheQVCurves
Inpractice,thesecurvesmaybedrawnwithapowerflowprogram
by
1.modelingatthetargetbusasynchronouscondenser(a
g
g
y
(
generatorwithP=0)havingverywidereactivelimits
2.Setting|V|toadesiredvalue
3 Solving the power flow
3.Solvingthepowerflow.
4.ReadingtheQofthegenerator.
5.Repeat24forarangeofvoltages.
QVcurveshaveoneadvantageoverPVcurves:
Theyareeasiertoobtainifyouonlyhaveapowerflow(standard
powerflowswillnotsolvenearorbelowthenoseofPVcurves
p
buttheywillsolvecompletelyaroundthenoseofQVcurves.)
17
Voltageinstabilityinalargesystem:
Influentialfactors:
Loadmodeling
Reactivepowerlimitsongenerators
Lossofacircuit
Loss of a circuit
Availabilityofswitchableshuntdevices
Twoimportantideasonwhichunderstandingoftheabove
i fl
influencesrest:
1. Voltageinstabilityoccurswhenthereactivepowersupply
cannotmeetthereactivepowerdemandofthenetwork.
Transmissionlineloadingistoohigh
Reactivesources(generators)aretoofarfromloadcenters
Generatorterminalvoltagesaretoolow.
I ffi i
Insufficientloadreactivecompensation
l d
i
i
2.Reactivepowercannotbemovedveryfarinanetwork
(varsdonottravel),sinceI2Xislarge.
Implication:TheSYSTEMcanhaveavarsurplusbutexperience
voltageinstabilityifalocalareahasavardeficiency.
18
Loadmodeling
Inanalyzingvoltageinstability,itisnecessarytoconsiderthenetwork
undervariousvoltageprofiles.
Voltagestabilitydependsonthelevelofcurrentdrawnbytheloads.
Thelevelofcurrentdrawnbytheloadscandependonthevoltageseen
bytheloads.
Therefore,voltageinstabilityanalysisrequiresamodelofhowthe
loadrespondstoloadvariations.
Thus,loadmodelingisveryinfluentialinvoltageinstabilityanalysis.
19
Exponentialloadmodel
Atypicalloadmodelforaloadatabusistheexponentialmodel:
V
P = P0
V0
V
Q = Q0
V0
wherethesubscript0indicatestheinitialoperatingconditions.
Theexponents and arespecifictothetypeofload,e.g.,
Incandescentlamps
1.54
Roomairconditioner
0.50
F
Furnacefan
f
0 08
0.08
Batterycharger
2.59
Electroniccompactflorescent1.0
Conventionalflorescent
2.07
2.5
16
1.6
4.06
0.40
3.21
20
Polynomialloadmodel
TheZIPorpolynomialmodelisaspecialcaseofthemoregeneral
exponentialmodel,givenbyasumof3exponentialmodelswith
i l
d l i
b
f3
i l
d l ih
specifiedsubscripts:
V 2
V
P = P0 p1 + p2 + p3
V0
V0
p1 + p2 + p3 = 1.0
V 2
V
Q = Q0 q1 + q2 + q3
V0
V0
q1 + q2 + q3 = 1.0
whereagainthesubscript0indicatestheinitialoperatingconditions.
Usually,valuesp2 andq2 arethelargest.
Sothismodeliscomposedofthreecomponents:
constantimpedancecomponent(p
constant impedance component (p1,q
q1)
) lighting
constantcurrentcomponent(p2,q2) motor/lighting
constantpowercomponent(p3,,q3) loadsservedbyLTCs
21
EffectofLoadmodeling
Understandingtheeffectofeachcomponentonvoltageinstability
Understanding
the effect of each component on voltage instability
dependsonunderstandingtwoideas:
1.Voltageinstabilityisalleviatedwhenthedemandreduces.This
l
i
bili i ll i d h
h d
d d
hi
isbecauseIreducesandI2Xreactivelossesinthecircuitsreduce.
2.Sincevoltageinstabilitycausesvoltagedecline,alleviationof
voltageinstabilityresultsifdemandreduceswithvoltagedecline.
This gives the key to understanding the effect of load modeling
Thisgivesthekeytounderstandingtheeffectofloadmodeling.
constantimpedanceload(p1)isGOODsincedemand
reduces with square of voltage
reduceswithsquareofvoltage.
constantcurrentload(p2)isOKsincedemandreduces
withvoltage.
Constantpowerload(p
C
l d ( 3)isBADsincedemanddoes
) i BAD i
d
dd
notchangeasvoltagedeclines.
22
Someconsiderationsinloadmodeling
Theeffectsofvoltagevariationonloads,andthusofloadson
The
effects of voltage variation on loads and thus of loads on
voltageinstability,cannotbefullycapturedusingexponentialor
polynomialloadmodelsbecauseofthefollowingthreeaspects.
Thermostaticloadrecovery
g pp g
Inductionmotorstalling/tripping
Loadtapchangers
23
Thermostaticloadrecovery
Heatingloadisthemostcommontypeofthermostaticload,andit
g
yp
,
isoneforwhichweareallquitefamiliar.Althoughmuchheatingis
donewithnaturalgasastheprimaryfuel,someheatingisdone
electrically and even gas heating systems always contain some
electrically,andevengasheatingsystemsalwayscontainsome
electriccomponentsaswell,e.g.,thefans.
Otherthermostaticloadsincludespaceheaters/coolers,water
h h
i l d i l d
h
/ l
heaters,andrefrigerators.
Whenvoltagedrops,thermostaticloadsinitiallydecreaseinpower
consumption.Butaftervoltagesremainlowforafewminutes,the
load regulation devices (thermostats) will start the loads or will
loadregulationdevices(thermostats)willstarttheloadsorwill
maintainthemforlongerperiodssothatmoreofthemareonat
thesametime.Thisisreferredtoasthermostaticloadrecovery,
andittendstoexacerbatevoltageproblemsatthehighvoltage
di
d
b
l
bl
h hi h l
24
level.
Inductionmotorstalling/tripping
Threephaseinductionmotorscompriseasignificantportionof
Three
phase induction motors comprise a significant portion of
thetotalloadandsoitsresponsetovoltagevariationisimportant,
especiallysinceithasaratheruniqueresponse.
Considerthesteadystateinductionmotorperphaseequivalent
model.
Za=R1+jX1
V1
Zb=
Rc//j
//jXm
X2
I2
R2+R2(1-s)/s
=R2 / s
25
Inductionmotorstalling/tripping
The(referredtostator)
I '2
rotor current is given by:
rotorcurrentisgivenby:
Zb
Vth = V1
where
Z a + Zb
Vth
=
Z th + ( R'2 / s ) + jX '2
and
Z a Zb
Z th = Z a // Z b =
Z a + Zb
Undernormalconditions,theslipsistypicallyverysmall,lessthan0.05
Under
normal conditions the slip s is typically very small less than 0 05
(5%).Inthiscase,R2/s>>R2,andI2 issmall.
ButasvoltageV
B
t
lt
V1 decreases,theelectromagnetictorquedeveloped
d
th l t
ti t
d l
d
decreasesaswell,themotorslowsdown.Ultimately,themotormaystall.
Inthiscase,s=1,causingR2/s=R2.Thus,oneseesthatthecurrentI2 is
muchlargerforstalledconditionsthanfornormalconditions.BecauseofX
hl
f
ll d
di i
h f
l
di i
B
f X1
andX2 oftheinductionmotor,thelargestallcurrentrepresentsalarge
reactiveload.
Largemotorshaveundervoltagetrippingtoguardagainstthis,butsmaller26
motors(refrigerators/airconditioners)maynot.
Tapchangers:
LLoadtapchangers(LTC,OLTC,ULTC,TCUL)aretransformersthat
d
h
(LTC OLTC ULTC TCUL)
f
h
connectthetransmissionorsubtransmissionsystemstothedistribution
systems.Theyaretypicallyequippedwithregulationcapabilitythat
allowthemtocontrolthevoltageonthelowsidesothatvoltage
deviationonthehighsideisnotseenonthelowside.
t:1
V1 andtare
giveninpu.
HVside
V1
V1/t
LVside
Inperunit,wesaythatthetapist:1,where
tmayrangefrom0.851.15pu
asinglestepmaybeabout0.005pu(5/8%=0.00625isverycommon)
a single step may be about 0 005 pu (5/8%=0 00625 is very common)
achangeofonesteptypicallyrequiresabout5seconds.
thereisadeadbandof23timesthetapsteptopreventexcessivetapchange
Underlowvoltageconditionsatthehighside,theLTCwilldecreaset
U
d l
l
di i
h hi h id h LTC ill d
inordertotryandincreaseV1/t.
27
Tapchangers:
Thus,aslongastheLTCisregulating(notatalimit),avoltage
declineonthehighsidedoesnotresultinvoltagedeclineatthe
load in the steady state so that even if the load is constant Z
load,inthesteadystate,sothateveniftheloadisconstantZ,
itappearstothehighsideasifitisconstantpower.Soasimple
loadmodelforvoltageinstabilityanalysis,forsystemsusingLTC,
isconstantpower!
Thereare2qualificationstousingsuchasimplemodel(constantpower):
There
are 2 qualifications to using such a simple model (constant power):
1.Fastvoltagedipsareseenatthelowside(sinceLTC
actiontypicallyrequiresminutes),andifthedipislowenough,
induction motors may trip resulting in an immediate decrease in
inductionmotorsmaytrip,resultinginanimmediatedecreasein
loadpower.
2.OncetheLTChitsitslimit(minimumt),thenthelowside
voltagebeginstodecline,anditbecomesnecessarytomodel
28
theloadvoltagesensitivity.
Generatorcapabilitycurve:
Fieldcurrentlimitduetofieldheating,
enforced by overexcitation limiter on If.
enforcedbyoverexcitationlimiteronI
Q
Qmax
Typical
approximation
usedinpower
di
flowprograms.
Armaturecurrentlimitdueto
Armature
current limit due to
armatureheating,enforcedby
operatorcontrolofPandIf.
P
Qmin
Limitduetosteadystateinstability(small
internalvoltageEgivessmall|E||V|Bsin),
andduetostatorendregionheatingfrom
inducededdycurrents,enforcedby
underexcitationlimiter(UEL).
29
Effectofgeneratorreactivepowerlimits:
1.Voltageinstabilityistypicallyprecededbygeneratorshittingtheirupperreactive
limit,somodelingQmax isveryimportanttoanalysisofvoltageinstability.
2.MostpowerflowprogramsrepresentgeneratorQmax asfixed.However,this
i
isanapproximation,andonethatshouldberecognized.Inreality,Q
i
i
d
h h ld b
i d I
li Qmax isnotfixed.
i
fi d
ThereactivecapabilitydiagramshowsquiteclearlythatQmax isafunctionofPand
becomesmorerestrictiveasPincreases.AfirstorderimprovementtofixedQmax
is to model Qmax asafunctionofP.
istomodelQ
as a function of P
3.Qmax issetaccordingtotheOvereXcitationLimiter(OXL).Thefieldcircuithasa
ratedsteadystatefieldcurrentIfmax,setbyfieldcircuitheatinglimitations.Since
2
heating is proportional to I f dt, we see that smaller overloads can be tolerated
heatingisproportionalto,weseethatsmalleroverloadscanbetolerated
overload
forlongertimes.
time
Therefore,mostmodernOXLsaresetwithatimeinversecharacteristic:
4.AssoonastheOXLactstolimitIf,thenno
furtherincreaseinreactivepowerispossible.
2.0
OXLcharacteristic
WhendrawingPVorQVcurves,the
If
actionofageneratorhittingQmax,will
Irated
manifestitselfasasharpdiscontinuity
120
1.0
inthecurve.
10
Overloadtime(sec) 30
EffectofOXLactiononPVcurve:
Onegeneratorhitsreactivelimit
|V|
o
Noreactive
limitsmodeled
P
(demand)
Note:GeorgiaPowerCo.modelsitsloadability
limitatpointx,notpointo.
31
L
Lossofacircuit
f i it
Comparereactivelosseswithandwithoutsecondcircuit
AssumebothcircuitshavereactanceofX.
I/2
/
I
X
I/2
Qloss=(I/2)
(I/2)2X
X+(I/2)
(I/2)2X
X=II2X/2
P
Qloss=II2X
Implication:Lossofacircuitwillalwaysincreasereactivelosses
in the network. This effect is compounded by the fact
inthenetwork.Thiseffectiscompoundedbythefact
thatlosingacircuitalsomeanslosingitslinecharging
32
capacitance.
Kundur,onpp.979990,hasanexcellentexamplewhichillustrates
manyoftheaforementionedeffects.Theillustrationwasdoneusing
alongtermtimedomainsimulationprogram(Eurostag).
33
I
P
|V|
Without
capacitor
it
Withcapacitor
P
(demand)
34
But,shuntcompensationhassomedrawbacks:
Itproducesreactivepowerinproportiontothesquareofthe
the voltage therefore when voltages drop so does the reactive
thevoltage,thereforewhenvoltagesdrop,sodoesthereactive
powersuppliedbythecapacitor.
Ithasamaximumcompensationlevelbeyondwhichstable
It h
i
ti l l b
d hi h t bl
operationisnotpossible(Seepg.972ofKundur,andnextslide).
(A
(AsynchronouscondenserandanSVCdonothavethese2drawbacks)
h
d
d SVC d
th
th
2d
b k)
ItresultsinaflatterPVcurveandthereforemakesvoltage
instabilit less detectable Therefore as the load gro s in areas
instabilitylessdetectable.Therefore,astheloadgrowsinareas
lackinggeneration,moreandmoreshuntcompensationisusedto
keepvoltagesinnormaloperatingranges.Bysodoing,normal
operatingpointsprogressivelyapproachloadability.
35
V1=1.0
EachQVcurve/Capacitorcharacteristic
intersectionshowstheoperatingpoint.Note
thatforthefirstthreeoperatingpoints,a
small increase in Q comp (indicated by
smallincreaseinQcomp(indicatedby
arrows)resultsinvoltageincrease,butfor
thelastoperatingpoint(950),moreQcomp
(say960)resultsinavoltagedecrease.
V2
PL
QL=0
S=|V2|2B*Sbase
with|V2|=1.0
675Mvar 450Mvar
|V2|
300Mvar
1.2
950Mvar
1.0
QV-curves drawn
using synchronous
condensor approach.
0.8
0.6
1600
1400
1200
1000
800
600
CapacitiveMvars
400
200
36
Bifurcationanalysis(ref:A.GaponovGrekhov,Nonlinearitiesinactionand
alsoVanCutsem&Vournas,Voltagestabilityofelectricpowersystems.)
Abifurcation,foradynamicsystem,isanacquisitionofanew
qualitybythemotionthedynamicsystem,causedbysmallchanges
in its parameters. A power system that has experienced a bifurcation
initsparameters.Apowersystemthathasexperiencedabifurcation
willgenerallyhavecorrespondingmotionthatisundesirable.
Considerrepresentingthedynamicsofthepowersystemas:
x& = F ( x, y, p )
0 = G ( x, y , p )
Eqts.1
Adifferentialalgebraicsystem(DAS):
H
Herex
representsstatevariablesofthesystem(e.g.,rotorangles,rotor
i bl
f h
(
l
speed,etc),y representsthealgebraicvariables(busvoltagemagnitudes
&voltageangles),andp representstherealandreactivepowerinjections
ateachbus.ThefunctionF representsthedifferentialequationsforthe
generators,andthefunctionG representsthepowerflowequations. 37
Typesofbifurcations
Thereareatleasttwotypesofbifurcation:
Hopf:twoeigenvaluesbecomepurelyimaginary:
abirthofoscillatoryorperiodicmotion.
Saddlenode:adisappearanceofanequilibriumstate.
The stable operating equilibrium coalesces with an unstable
Thestableoperatingequilibriumcoalesceswithanunstable
equilibriumanddisappears.Thedynamicconsequenceofa
genericsaddlenodebifurcationis:
a monotonic decline in system variables
amonotonicdeclineinsystemvariables.
Sowethinkitisthesaddlenodebifurcationthatcauses
voltage instability
voltageinstability.
38
TheunreducedJacobian:
The Jacobian matrix of eqts 1 is
TheJacobianmatrixofeqts.1is
F X
J =
G X
FY
GY
anditisreferredtoastheunreducedJacobianoftheDAS,where
x
x&
0 = J y
Eqt 2
Eqt.2
39
ThereducedJacobian:
Wemayreduceeq.2byeliminatingthevariable y
x& F X
0 = G
X
F Y x
y
GY
Thismeansweneedtoforcethetoprighthandsubmatrixto0,whichwecan
dobymultiplyingthebottomrowbyFYGY1 andthenaddingtothetoprow.
x& F X F Y G Y G X 0 x
0=
G Y y
G X
Thisresultsin:
1
x& = F X F Y G Y G X x
1
SothatthereducedJacobianmatrixisaSchurscomplement:
1
A = F X F Y GY G X
40
Stability:
Fact1:Theconditionsforasaddlenodebifurcationare
x& = F ( x, y, p)
1. Equilibrium:
F X
J =
0 = G ( x, y , p )
GX
2. SingularityoftheunreducedJacobian
Singularity of the unreduced Jacobian
det(J)=0(a0eigenvalue,J noninvertible).
FY
G Y
Implication1:ThestabilityofanequilibriumpointoftheDASdependson
theeigenvaluesoftheunreducedJacobianJ.Thesystemwillexperiencea
SNBasparameterp increaseswhenJ hasazeroeigenvalue.
Fact2:ThedeterminantofaSchurscomplementtimesthedeterminantof
p
GY givesthedeterminantoftheoriginalmatrix:det(J)=det(A)*det(GY)
ifGY isnonsingular.
Implications2:
Implications
2:
1. IfGY isnonsingular,thensingularityofA impliessingularityofJ so
thatwemayanalyzeeigenvaluesofA toascertainstability.
2. ThefactthatG
h f
h
b
l
singular,meansthat
l
h
Y maybenonsingular,yetA
loadflowconvergenceisnotasufficientconditionforvoltage 41
stability.
SingularityofloadflowJacobian:
IImplications2:
li i
2
1. IfGY isnonsingular,thensingularityofA impliessingularityofJ so
thatwemayanalyzeeigenvaluesofA toascertainstability.
2. ThefactthatGY maybenonsingular,yetA singular,meansthat
loadflowconvergenceisnotasufficientconditionforvoltage
stability
stability.
GY
Singular
Singular
Nonsingular
Nonsingular
A
Singular
(unstable)
Nonsingular
(stable)
42
SingularityofloadflowJacobian:
SSovoltageinstabilityanalysisusingonlyaloadflowJacobianmayyield
l
i
bili
l i
i
l l d fl J bi
i ld
optimistic resultswhencomparedtoresultsfromanalysisofA,
thatis,stablepoints(basedonGy)maynotbereallystable.
=>However,Ibelieveitistruethatpointsidentifiedasunstableusing
theloadflowJacobianwillbereallyunstable(Schurscomplement
does not support that singularity of GY impliessingularityofJ,
doesnotsupportthatsingularityofG
implies singularity of J
however,becauseitisonlyvalidifGY isnonsingular).
GY
Singular
(unstable)
Nonsingular
Nonsingular
(stable)
Singular
(unstable)
Nonsingular
g
(stable)
Singular
Singular
(unstable)
Nonsingular
(stable)
Note:SauerandPai,1990,provideanin
Note:
Sauer and Pai 1990 provide an indepth
depthanalysisoftherelation
analysis of the relation
betweensingularityofGY andsingularityofJ,andshowsomespecial43
casesforwhichsingularityofGY impliessingularityofJ.
SingularityofloadflowJacobian:
SSo,weassume thatloadflowJacobiananalysisprovidesanupper
h l d fl J bi
l i
id
boundonstability.
Fact:Thebifurcation(zeroeigenvalueofGY)oftheloadflow
Jacobian corresponds to the turnaround
Jacobiancorrespondstothe
turn aroundpoint
point(i.e.,the
(i e the nose
nose
point)ofaPVorQVcurvedrawnusingapowerflowprogram.
Thiscanbeprovenusinganoptimizationapproach.
Seepp.218220ofthetextbyVanCutsemandVournas.
WehavepreviouslydenotedthepowerflowequationsasG(x,y,p)=0,
butnowwedenotethemasG(y,p)=0,withoutthedependenceonthe
statevariablesx (whichrelatetothemachinemodelingandinclude,
minimally, and ofeachmachine).
44
Soweturnourefforttoidentifyingthesaddlenodebifurcation
(SNB)forthepowerflowJacobianmatrix.
TheJacobiancanreachaSNBinmanyways.Forexample,
increasetheimpedanceinakeytieline
p
y
increasethegenerationlevelatageneratorwithweaktransmission,while
decreasinggenerationatallothergenerators.
||V||
g
increasetheloadatasinglebus
increasetheloadatallbuses.
Inallcases,wearelookingforthenosepointofthe
V curve,where istheparameterthatisbeingincreased.)
Mostapplicationsfocusonthelastmethod(increaseloadatallbuses).
Key questions here are:
Keyquestionshereare:
directionofincrease:arebusloadsincreasedproportionally,orinsomeotherway?
dispatchpolicy:howdothegeneratorspickuptheloadincrease?
Wewillassumeproportionalloadincreasewithgovernorloadflow
(generatorspickupinproportiontotheirrating)
45
Define:criticalpoint
D
fi
iti l i t theoperatingconditions,characterized
th
ti
diti
h
t i d
byacertainvalueof,beyondwhichoperationisnot
acceptable.
Question1:
Whatcancausethecriticalpointtodiffer
p
fromtheSNBpoint?
|V|
Question2:
Question
2:
Howcanknowledgeofthecriticalpointprovideasecurity
measure?
Question3:
DoesthePVcurveprovideaforecast ofthesystemtrajectory?
46
i=i+1
g
goto2
elseifnotsolved,
*=(i+1)
endif
4.End
Approach2:Usethecontinuationpowerflow(CPF).
Predictor step
Predictorstep
Correctorstep
p
No
No.
Pass*?
SSelect
l
continuation
parameter
Yes.
Stop
48
Thepredictorstep:
Thepowerflowequationsarefunctionsofthebusvoltagesand
busanglesandthebusinjections:
0 = G ( y, p)
Augmentthepowerflowequationssothattheyarefunctionsof
(dependenceonp iscarriedthroughthedependenceon).
pp0
0 = G ( y, )
N
Nowrecognizethat
i th t y =
sothat
0 = G ( , V , )
IfwewanttocomputethechangeinthepowerflowequationsdG
If
we want to compute the change in the power flow equations dG
duetosmallchangesinthevariables,V,and,
thatmoveusclosertotheloadabilitypoint
aswemovefromonesolutionitoanotherclosesolutioni+1,then
as we move from one solution i to another close solution i+1 then
dG=G((i),V(i),(i)) G((i+1),V(i+1),(i+1))=0 0=0
49
dG
dG
dG
dG =
d +
dV +
d
d
dV
d
Here,eachsetofpartialderivativesareevaluatedattheoperatingconditions
correspondingtotheoldsolution.Ifthepowerflowequationsarelinearwiththe3
sets of variables in the region between the old solution and the (close) new one the
setsofvariablesintheregionbetweentheoldsolutionandthe(close)newone,the
followingissatisfied:
dG
dG
dG
dG =
d +
dV +
d = 0
d
dV
d
Eq.3
q
[G
GV
G ]d V = 0
d
BUT,wehaveaddedoneunknown,,
BUT
h
dd d
k
tothepowerflowproblemwithoutaddinga
t th
fl
bl
ith t ddi
correspondingequation,i.e.,inG(,V,)=0,thereareareNequationsbutN+1
variables,sothatineq.3,thematrix[G GV,G],hasNrows(thenumberofeqts
beingdifferentiated)andN+1columns(thenumberofvariablesforwhicheacheqt
isdifferentiated). Soweneedanotherequationinordertosolvethis.Whattodo?
50
Theanswertothiscanbefoundbyidentifyinghowwewillbeusingusingthe
solution to eqt. 3. Note the solution corresponding to the new
solutiontoeqt.3.Notethesolutioncorrespondingtothe
new pointis:
point is:
(i +1, p ) (i ) d '
(i +1, p ) (i )
= V + d V '
V
(i +1, p ) (i ) d '
Herethe
Here
the p
p indicates
indicates
thatthisisthe
predictedpoint.
Ifwedefine tobethestepsize,thenwecanrewritethisas
(i +1, p ) (i )
d
(i +1, p ) (i )
where d '
d
=
V
V
+
d
V
d V ' = d V
(
i
+
1
p
)
(
i
)
d '
d
51
Wecalltheupdatevector(withthedifferentials)the
tangentvector,denotedbyt.
d
t = dV
d
Thisvectorprovidesthedirection tomoveinorder
tofindanewsolution(i+1,p)fromtheoldone(i).
We can think of this in terms of the following picture
Wecanthinkofthisintermsofthefollowingpicture..
52
Tangentvector
|V|
53
Note:Inspecifyingadirectionusinganndimensionalvector,onlyn1ofthe
elementsareconstrained oneelementcanbechosentobeanyvaluewe
y
like.
Forexample,considera2dimensionalvector.
x2=x1tan(30)so:
x2
Direction
=30o
x1
thedirectionisspecifiedby
selectingx1=1,x2=0.5774,
thedirectionisspecifiedby
selectingx
g 1=0.5,x
, 2=0.2246.
Sowecansetoneofthetangentvectorelementsto
anyvaluewelike,thencomputetheotherelements.
l
lik h
h
h
l
Thisprovidesuswithourotherequation.
54
Supposethatwesetthekthparameterinthetangentvectortobe
1.0.Thenourequationgivenaseq.3canbeaugmentedtobecome:
GV
ek
d
G 0
dV = 1
d
where
ek = [ 0
...
...
0]
k
Toselect,wewouldhave:
ek = [ 0
...
...
1]
Whichwould
forced=1.
55
Theparameterforwhichweselectkiscalledthecontinuation
parameter,anditcanbeanyloadlevel(orgroupofloadlevels),
oritcanbeavoltagemagnitude.Initially,whenthesolutionis
farfromthenose,thecontinuationparameteristypically.
(i +1, p ) (i )
d
(i +1, p ) (i )
dV
V
=
V
+
(i +1, p ) (i )
d
( i +1, p )
=y
(i )
+ t
Theparameter iscalledthestepsize,anditcanbeselected
usingvarioustechniques.Thesimplestoftheseistojust
setittoaconstant.Letstrythisonoursimpleproblem
formulatedatthebeginningoftheseslides.
56
HOMEWORK#2,DueMonday,Jan26.
1.Usingtheequationsatthebottomofslide7,withthelefthandside(PD
andQD)andalsoV1givenbytheproblemstatement,weknoweverything
except V2 and theta.
exceptV2andtheta.
2.Now,justbringtherighthandsideofthese2equationsovertothe
lefthandside,andyouhavethe2equationsthatcorrespondtoG(y,p)=0.
3.Solvetheseequationstogetthecorrespondingpowerflowsolution(but
youdonotneedNewtonRaphson todothis youcanjustusetheequation
atthebottomofslide10).
t th b tt
f lid 10)
4.NowyouneedtoreplacethevaluespecifiedintheequationsforPD
(whichis0.4accordingtotheproblemstatement)with0.4*lambda.This
givesyoutheequationsintheformofslide49:0=G(theta,V,lambda).
Note,however,thatGisreallytwoequations:G1andG2.
y
q
5.Nowyouneedtoformulatetheequationsontheslide55.Thisisamatter
oftakingderivativesandthenevaluatingthosederivativesatthe
solutionthatyouobtainedabove.Note,however,theeachelementinthe
matrixofslide55actuallyrepresents2elements.Thatis:
PD =| V1 || V2 | B sin 12
QD = | V2 |2 B + | V1 || V2 | B cos12
|dG1/dtheta
dG1/dV
dG1/dlambda|
|dG2/dtheta
dG2/dV
dG2/dlambda|
#9and#10willbe
|0
0
1
|
6.Evaluateeachoftheabovematrixelementsatthesolutionobtainedin
explainedinnext
p
step3.
fewslides.
7.Thensolvetheseequationsforthetangentvector.Youcandothisby
invertingtheabovematrix(usematlab oracalculatortodothis)and
thenmultiplytherighthandsidebythisinvertedmatrix.
8 Then take a step
8.Thentakea
step usinganappropriatelychosenstepsizeperthe
using an appropriately chosen step size per the
equationonslide56.
9.Beginningfromyourpredictedpointthatyouidentifiedinstep8of#2a,develop
equationsforapproacha,solvethem,andidentifytheresultingcorrectedpointintermsofvoltageandpower.
10.Repeat#9exceptimplementapproachb.
57
Correctorstep
Note,however,thatthepredictedpointwillsatisfythe
powerflowequationsonlyifthepowerflowequationsare
linear, which they are not.
linear,whichtheyarenot.
Soourpointneedscorrection.Thisleadstothecorrectorstep.
Therearetwodifferentapproachesforperformingthe
correctorstep.
Approacha: Perpendicularintersectionmethod.
Approachb: Parameterizationmethod
58
Approacha:perpendicularintersection
Here,wefindtheintersectionbetweenthepowerflow
equations(thePVcurve)andaplanethatisperpendicularto
the tangent vector
thetangentvector.
|V|
Solvesimultaneously,
fory(i+1)
0 = G( y
{y
( i +1)
( i +1)
( i +1, p )
, )
y(i)
t
y(i+1,p)
y(i+1)
} t = 0
UseNewtonRaphsontosolvetheabove(requiresonly13iterationssincewehave
goodstartingpoint).Ifnoconvergence,cutstepsize()byhalfandrepeat.
59
Approachb:Parameterization
Thecorrectorstepisperformedby
identifyingacontinuationparameter (seeslide62) canbe
fixingitatthevaluefoundinthepredictorstep;
g
p
q
thensolvingthepowerflowequations.
|V|
Solvesimultaneously,
Solve
simultaneously
fory(i+1)
G ( y , )
=0
( i +1)
yk
( i +1)
y(i)
Verticalcorrections
correspondtoafixed
loadcontinuation
parameter,horizontal
correctionstoafixed
voltagecontinuation
parameter.
t
(i 1 )
y(i+1,p)
y(i+1)
Detectionofcriticalpoint:
Wewillknowthatwehavesurpassedthecriticalpoint
whenthesignofd inthetangentvectorbecomes
negative,becauseitisatthispointwheretheloading
reachesamaximumpointandbeginstodecrease.
|V|
increasing
decreasing
61
Selectionofcontinuationparameter:
The continuation parameter is selected from among
Thecontinuationparameterisselectedfromamong
andthestatevariablesiny accordingtotheonethatis
changingthemostwith.Thiswillbetheparameterthat
Theonechanging
themostwith is
most sensitive and
mostsensitiveand
representsa
variablethatwe
wanttobecareful
withaswelookfor
anothersolution,
soitmakessenseto
keepitconstant.
hasthelargestelementinthetangentvector.
relativelyunstressedconditions(farfromnose):generally
relativelystressedconditions(closetonose):generallythe
relatively stressed conditions (close to nose): generally the
voltagemagnitudeoftheweakestbus,asitchangesagreat
dealas ischanged,whenwearecloseto*.
Typically,yk
isgoingtobe
oneofthese.
(i +1, p ) (i )
d
(i +1, p ) (i )
V
= V + dV
(i +1, p ) (i )
d
62
Selectionofcontinuationparameter(unstressedcondition):
The continuation parameter is selected from among
Thecontinuationparameterisselectedfromamong
andthestatevariablesiny accordingtotheonethatis
changingthemostwith.Thiswillbetheparameterthat
hasthelargestelementinthetangentvector.
relativelyunstressedconditions(farfromnose):generally.
=>Thislookslikebelow.
> This looks like below
|V|
y(i)
y(i+1,p)
y(i+1)
Here, isfixed.
63
Selectionofcontinuationparameter(stressedcondition):
relativelystressedconditions(closetonose):generallythe
y
(
) g
y
voltagemagnitudeoftheweakestbus.Here,thevoltagebeing
plottedischosenasthecontinuationparameter.
|V|
y(i)
(i+1 p)
y(i+1,p)
y(i+1)
Here,|V|isfixed.
Essentially,avariableisfixedasaparameter(thevoltage),and
p
( )
p
g
theparameter()istreatedasavariable.Thisprocessofselecting
avariabletofixissometimescalledtheparameterizationstep.
ScottGreene,Ph.D.dissertation,1998.
64
Acentralquestion:
Howdoesthecontinuationtechniquealleviatetheill
conditioning problem experienced by a regular power flow ?
conditioningproblemexperiencedbyaregularpowerflow?
Refer to the solutions procedures for the two corrector approaches.
Refertothesolutionsproceduresforthetwocorrectorapproaches.
Perpendicularinteresection
Solvesimultaneously,
for y(i+1)
fory
0 = G( y
{y
( i +1)
( i +1)
, )
y (i +1, p ) t = 0
Parameterization
Solvesimultaneously,
for y(i+1)
fory
G ( y (i +1) )
=0
yk (i +1)
IInbothcases,weuseNewtonRaphsontosolve,soweneedtoobtainthe
b th
N t R h
t
l
d t bt i th
Jacobian.ButtheJacobianisslightlydifferentthaninnormalpowerflow.
65
TheJacobianofthepowerflowequationsisjustGy,butthe
Jacobian of the equations in the two corrector approaches
Jacobianoftheequationsinthetwocorrectorapproaches
willhaveanextrarowandcolumn.
G y
C y
G xk
C xk
Here,Cistheadditionalequation,andxk istheselected
continuationparameter.
ThisadditionofarowandcolumntotheJacobianhasthe
effectofimprovingtheconditioningsothatthepreviously
ff
fi
i
h
di i i
h h
i l
singularpointscaninfactbeobtained.Inotherwords,the
additionalrowandcolumnprovidesthatthisJacobianis
nonsingularat
i
l
* wherethestandardJacobianissingular.
h
h
d d J bi i i
l
66
Knowncodesforcontinuationmethods:
1. ClaudioCanizarresatUniversityofWaterloo:Ccode
Seehttp://www.power.uwaterloo.ca/~claudio/claudio.html
UWPFLOWisaresearchtoolthathasbeendesignedtocalculatelocalbifurcationsrelatedtosystem
limitsorsingularitiesinthesystemJacobian.Theprogramalsogeneratesaseriesofoutputfilesthat
allowfurtheranalyses,suchastangentvectors,leftandrighteigenvectorsatasingularbifurcation
point,Jacobians,powerflowsolutionsatdifferentloadinglevels,voltagestabilityindices,etc
2. IhaveMatlabcodethatdoesit fromScottGreene.
3. VenkataramanaAjjarapu(ISU):Fortrancode
4. Powertechhasaprogram
67
Calculationofsensitivitiesforvoltageinstabilityanalysis
Whatisasensitivity?
Itisthederivativeofanequationwithrespecttoavariable.
Itshowshowparameter1changeswithparameter2.
Itis:exactwhenparameter2dependslinearlyonparameter1.
Itisapproximatewhenparameter2dependsnonlinearlyonparameter1,
butitisquiteaccurateifitisonly
usedclosetowhereitiscalculated.
68
G
y
ConsiderthesystemcharacterizedbyG(y).Then
is the sensitivity of the equation G with respect to y
isthesensitivityoftheequationGwithrespecttoy,
evaluatedaty*.
G(y)
y*
SlopeisG/y
evaluatedaty*.
y y*
y
Itsusefulnessisthatonceitiscalculated,itcanbeusedto
QUICKLYevaluatef(y)fromG(y)G(y*)+(G/y|y*)y,
BUTONLYASLONGASyISCLOSETOy*.
69
Considerparameterp:wedesiretoobtainthesensitivityof
Consider
parameter p: we desire to obtain the sensitivity of
G(y,p)top.Typicalparameterspwouldbeabusload,abus
powerfactor,oragenerationlevel.
Veryimportanttodistinguishbetween
voltagesensitivities
voltageinstabilitysensitivities
Whatisthedifferencebetweenthemintermsof
whattheymean?
howtocomputethem?
how to compute them ?
70
|V|
Foragivencontinuationparameter,theycanbeobtained
from the first predictorstepinthecontinuationpowerflow.
fromthefirst
predictor step in the continuation power flow
Recallthatthisprovidesuswith d
dV
t
=
thetangentvector,givenby:
d
Current
operatingpoint
Thetangentvectoristhevectorof
Th
t
t
t i th
t
f
sensitivitieswithrespecttoasmall
changein,sotheportionofthevector
designatedasdV isexactlythevoltage
sensitivities.
71
Sensitivitiesforvoltageinstability
Here,itisimportanttorealizethatthemeasureofvoltageinstability,
Here
it is important to realize that the measure of voltage instability
theloadingmargin,dependsonanoperatingcondition
differentfromthepresentoperatingcondition.
Theimplicationisthatwemustlookatsensitivities
of the loading margin not of the voltage
oftheloadingmargin,notofthevoltage.
|V|
Loadingmargin
Current
operatingpoint
Sowewantthesensitivities
evaluated at this point, i.e.,
evaluatedatthispoint,i.e.,
theSNBpoint.
72
DerivationofloadingmarginsensitivitiesatSNBpoint.
LetS bethevectorofrealandreactiveloadpowers,
andk
d k bethedirectionofloadincrease.
b h di
i
fl di
S = S0 + k
Also,defineLastheloadingmargin(ascalar),sothat
theloadpowersresultingintheSNBpointaregivenby:
S = S 0 + Lk
WedesiretofindthesensitivityoftheloadingmarginLtoa
changeintheparameterp.WedenotethissensitivitybyL
h
i th
t
W d
t thi
iti it b Lp.
73
Considerthesystemcharacterizedby
G(y,S,p)=0
Wewantthesensitivityof
the loading margin to p.
theloadingmargintop.
Assumption:thesystemhasaSNBat(y
p
y
(y*,,S*,,p
p*),
),i.e.,:
,
1.G(y*,S*,p*)=0 (anequilibriumpoint)
2.Gy(y*,S*,p*)issingular(zeroeigenvalue),and
w isalefteigenvectorofGy(y*,S*,p*),corresponding
tothezeroeigenvaluesothat(bydefinitionoftheleft
eigenvector)
wT Gy(y*,,S*,,p
p*))=0wT=0
NotethatGy(y*,S*,p*),beingsingular,cannotbeinverted,but
wecancomputeit(thatis,Gy(y*,S*,p*)),anditseigenvectors.
3.wT GS(y*,S*,p*) 0
74
Thepoints(y,S,p)satisfyingnumbers1and2correspondtoSNB
points,
andwecanobtainacurveofsuchpointsby
varyingpaboutitsnominalvaluep*.
LinearizationofthiscurveabouttheSNBpointresultsin
G y y + G S * S + G p p = 0
*
wherethenotation|* indicatesthederivativesareevaluatedattheSNBpoint.
P
Premultiplicationbythelefteigenvectorw
lti li ti b th l ft i
t
resultsin:
lt i
w G y y + w G S * S + w G p p = 0
T
By#2onthepreviousslide,thefirsttermintheaboveiszero.So...75
w G S * S + w G p p = 0
T
Eqt.*
Nowrecalltherelationoftheloadpowerstotheloadingmargin.
S = S 0 + Lk
S = L k
Substitutingthisexpressionfortheloadpowersintoeqt.*,
w G * L k + w G p p = 0 L w G * k = w G p p
T
L w G p *
=
= T
p w G S k
T
Lp
Sopmaybe,forexample,
realpowerloadatabus(to
detect the most effective load
detectthemosteffectiveload
shedding)orreactivepower
atabus(todeterminewhere
76
tositeashuntcap).
w Gp
T
SomecommentsaboutcomputingLp
Lp =
*
w GS k
*
Thelefteigenvectorw mustbecomputedforthe
Jacobian GyevaluatedattheSNBpoint.
JacobianG
evaluated at the SNB point
Youonlyneedtocomputew andGS once,independentof
how many sensitivities you need. Methods to compute the left eigenvector
howmanysensitivitiesyouneed.Methodstocomputethelefteigenvector
w includeQRorinverseiteration.
Thevectorofderivativeswithrespecttotheparameterp,whichisGp, is
typicallysparse.Forexample,ifyouwanttocomputethesensitivitytoa
buspower,thentherewouldbeonly1nonzeroentryinGp.
Thematrixofderivativeswithrespecttotheloadpowers,GS,usingconstant
powerloadmodels,isadiagonalmatrixwithonesintherowscorresponding
toloadbuses.ThisisbecauseaparticularloadvariablewouldONLYoccur
i th
intheequationcorrespondingtothebuswhereitislocated,andforthese
ti
di t th b
h
it i l t d d f th
equations,thesevariablesappearlinearlywith1ascoefficient.
77
Lp =
*
w Gp
w GS k
*
Gettingmultiplesensitivitiescanbeespeciallyattractivewhenwewanttofind
thesensitivitytoseveralsimultaneouschanges.Onegoodexampleistofindthe
sensitivitytochangesinmultipleloads.
AspecialcaseofthisistofindthesensitivitytochangesatALLloads,whichis
verytypical,givenaparticularloadingdirectionk.Then
l
l l
k h
Lall loads * = ki L pi
i
Asensitivitytoalineoutagemaybeobtainedbylettingp
A
iti it t li
t
b bt i d b l tti
containelements
t i l
t
correspondingtotheoutagedlineparameters.
78
Somecommentsaboutextensions
Asensitivitytoalineoutagemaybeobtainedbylettingp
A
iti it t li
t
b bt i d b l tti
containelements
t i l
t
correspondingtotheoutagedlineparameters:R(seriesconductance),X(series
reactance),andB(linecharging).Thenusethemultipleparameterapproach.
w Gp
Zpq=R+jX
Lp =
*
w GS k
*
L = L p * p
q
jB
jB
Here,p =[RXB]T.
Notethatp isNOTSMALL!ThereforeLmayhaveconsiderableerror.
Forthatreason,thisoneneedstobecarefulaboutusingthisapproachto
computetheactualloadingmarginsfollowingcontingencies.
However,itcertainlycanbeusedforRANKINGcontingencies.Onemight
considerhavingaquickapproximationandalongexactriskcalculation. 79
1
L = L p p + L pp (p ) 2
*
2
*
Itrequiressignificantlymorecomputationbutcanprovidegreater
accuracy over a larger range of p
accuracyoveralargerrangeofp.
InvariantSubspaceParameticSensitivity(ISPS)byAjjarapu.
Advantages:
basedondifferentialalgebraicmodel
g
providessensitivitiesatANYpointonthePVcurve
80