Documente Academic
Documente Profesional
Documente Cultură
of Low Voltage
Networks with High Photovoltaic
Systems Penetration to
Transmission Network Faults
KonstantinosSkaloumpakas
Supervisors:
Prof.ir.MartvanderMeijden
Dr.ir.MadeleineGibescu
Dip.Ing.JensC.Boemer
M.Sc.ThesisinSustainableEnergyTechnology
FacultyofAppliedSciences
SpecializationinElectricalPowerEngineering
Delft,January2014
ii
PrefaceAcknowledgments
Nowadaystherapidgrowthoftechnologythatsurroundsrenewableformsofenergyis
one of the main discussion topics in the energy sector. For the purpose of securing our
future energy supplies, and given the realistic scenario of depletion of traditional energy
sources (oil, gas and coal), renewables seem to be one of the few viable alternatives. The
basis of research on renewables focuses on developing the technology in order to make
exploitation of solar, wind, biomass and the rest of clean energy sources economically
feasible.Inthisframeworkrenewablesneedtoachievehighefficiencyandaffordablecosts
forinstallationandmaintenance.However,despitethetechnologiesthemselves,thereisa
need to integrate the new forms of energy with their individual characteristics inside the
alreadyexistingpowersupplysystem.Asdiscussedin[1]theintegrationofPVinstallations
on all levels of the grid is going to introduce different characteristics in the energy
production.Researchisneededtospecifytheimpactonthesystemoperationalsecurity.
ThisthesishasbeenaveryinterestingjourneyinthefieldofPowersystems.Throughout
this journey I gained a lot of experience as a researcher by reading vast amounts of
publications,performingdetailedmodellingofnetworks,understandingthedetailsofpower
systemsintermsofpolicybutalsooperationandbehaviour.Iunderstoodtheimportanceof
planning and organization in order to maximize efficiency and output of work and most
importantlyIlearnedhowtomanagemyselfandmeetthedeadlinesevenintheexpenseof
sleep hours. This research of course would not be the same without the help of my daily
supervisor and mentor Jens C. Boemer that guided me throughout this project and freely
shared his knowledge and experience with me, spending long days and nights to help and
guideme.MoreoverIwanttothankAss.Prof.MadeleineGibescuformakingitpossiblefor
metobepartofthisprojectandforalwayskeepingacloseeyeonmemakingsureIamon
the right path. Furthermore Prof. Mart van der Meijden that despite his busy schedule
alwaysshowedinterestinmyworkandencouragedmyendeavours.FinallyIwanttothank
themembersoftheIEPGgroupoftheEWIfacultyatTUDelftfortheirsupportandalsomy
familyandfriends.
iii
iv
Abstract
The need for fault ridethrough (FRT) capability is not yet documented for the low
voltageconnectedDG.TherealizationthatanincreaseinthecapacityofinstalledDGpower
in the specific voltage level can trigger disconnection, in case of transmission faults, of
massiveamountsofpowerfromDGandjeopardizesystemstabilityisforming.
In order to investigate the response of low voltage distribution systems with high
penetrationofPVs,incaseofafaultinthetransmissionsystem,acomplexnetworkmodelis
createdconsistingofdifferentvoltagelevels,fromeHV(220kV)toLV(400V).Thismodelis
implemented as a benchmark system with DG capacities chosen to match the German
network. Also an elaborate PV model is used to simulate different behaviours during and
after fault and assist in examining the impact of different sensitivity factors on network
stability. The problem is analysed through root mean square (RMS) stability simulations
performed on the models of the network and the DG. The projected year that this study
takesplaceis2022andtheDGcapacityischosenforthatyear.Thecommerciallyavailable
DIgSilentPowerFactorysoftwareversion14.1isusedtocarryoutthesesimulations.
AsensitivityanalysisisperformedfordifferentcontrolmodesofthePVthatrangefrom
disconnection in the event of a fault to full dynamic network support during fault. Also
differentprefaultstatesofthenetworkareusedtoexaminetheeffectofthestateofthe
networkintheresults;thesestatesvaryfromnormaltopdownpowerflowtofullreverse
powerflowwhenPVsareatfullcapacity.Moreovertheeffectofdelayedpostfaultactive
powerrecoveryoftheLVconnectedDGisdiscussed.
The conclusions and analysis show the importanceof introducing FRT criteriaeven for
low voltage connected DG and give an insight on the behaviour of the system under the
circumstances described. The differences between control modes are affecting stability of
thesystemandthevariousphenomenathattakeplaceareanalysedandexplained.Finally
corrective steps that need to be taken in order to avoid future stability problems for the
network are outlined and justified and recommendations for future grid connection
requirementsforlowvoltageinstalledDGaregiven.
vi
Table of Contents
Nomenclature.......................................................................................................................................xi
Listofdefinitions................................................................................................................................xiii
Listofabbreviations............................................................................................................................xv
ListofFigures......................................................................................................................................xvii
ListofTables.........................................................................................................................................xxi
Listofequations.................................................................................................................................xxiii
1.1 Background..............................................................................................................................1
1.2 ProblemDefinition.................................................................................................................2
1.3 ObjectiveandResearchQuestions......................................................................................3
1.4 ResearchApproach................................................................................................................3
1.5 OutlineoftheThesis..............................................................................................................5
2.1 PowerSystemStability..........................................................................................................7
2.2 GridCodesandFaultRideThroughRequirementsStatusQuo....................................9
2.3 ImpactofFaultRideThroughandAdditionalReactiveCurrentInjectionDuringFault
onLineandGenerationProtection...................................................................................12
2.4 PhotovoltaicSystemsCurrentSettingsandOperationalPerformance....................16
2.4.1 EquationofaPVarray...........................................................................................16
2.4.2 ModellingofaPVsystem......................................................................................17
2.5 TypesofFaults......................................................................................................................19
2.6 Conclusions...........................................................................................................................19
3.1 Introduction..........................................................................................................................21
3.2 StructureoftheTestSystem..............................................................................................21
vii
3.3 Systemlevels.........................................................................................................................22
3.3.1 IEEE39Bus,10MachinesnewEnglandnetwork.............................................22
3.3.2 TypicalGermansubtransmission110kVringnetwork...................................23
3.3.3 CIGRmediumvoltagebenchmarknetwork....................................................24
3.3.4 CIGRlowvoltagebenchmarksystem................................................................26
3.3.4.1 Simplifications.........................................................................................26
3.3.4.2 Krontransformation(foroverheadlines,cables)...................................27
3.4 DistributedGenerationModelling.....................................................................................28
3.4.1 PhotovoltaicpowerParkmodule(PVPPM)........................................................28
3.4.1.1 Existingmodel(DCside,DC/ACinvertercontrol)...................................28
3.4.1.2 Networkfaultcontrolmodesandextensions........................................30
3.4.2 Windturbinegenerators.......................................................................................40
3.5 LoadModelling.....................................................................................................................41
3.6 Conclusions...........................................................................................................................42
4.1 Introduction..........................................................................................................................43
4.2 SystemConditions................................................................................................................43
4.2.1 Generaloverviewofthesystem..........................................................................43
4.2.2 ScenariosforinstalledPVcapacity......................................................................44
4.3 SensitivityFactors.................................................................................................................46
4.3.1 Load/Generationprofile.......................................................................................46
4.3.1.1 Basecaseunidirectionalpowerflow......................................................47
4.3.1.2 Partial(MVandHV)reversepowerflow................................................48
4.3.1.3 Complete(LVtoHV)reversepowerflow...............................................49
4.3.2 Networkfaultcontrolmodesettings..................................................................50
4.3.3 Overviewofsensitivitycases................................................................................50
4.4 PresentationandAnalysisofResults.................................................................................51
4.4.1 BaseCasewithnoLVRT........................................................................................51
4.4.1.1 Presentationofresults............................................................................51
4.4.1.2 Analysisofresults...................................................................................54
4.4.2 ComparisonbetweennoLVRT/ZPM/aRCI/aRACIforthebasecase..............57
4.4.2.1 Presentationofresults............................................................................57
4.4.2.2 Analysisofresults...................................................................................59
4.4.3 ComparisonbetweennoLVRT/ZPM/aRCI/aRACIinthecaseoffullreverse
powerflow..............................................................................................................62
4.4.3.1 Presentationofresults............................................................................62
4.4.3.2 Analysisofresults...................................................................................64
4.4.4 Effectofprefaultstartingpoint..........................................................................67
viii
4.4.4.1 Presentationofresults............................................................................67
4.4.4.2 Analysisofresults...................................................................................68
4.4.5 Effectofdelayedactivepowerrecovery(dAPR)...............................................69
4.4.5.1 Presentationofresults............................................................................69
4.4.5.2 Analysisofresults...................................................................................72
4.5 Conclusions...........................................................................................................................73
5.1 Introduction..........................................................................................................................75
5.2 Conclusions...........................................................................................................................75
5.2.1 VoltagedipdepthsensitivityinLVnetworkswithhighpenetrationofPV...75
5.2.2 Impactofreversepowerflow..............................................................................77
5.2.3 Recommendationsfornetworkfaultcontrolmodes.......................................77
5.2.4 Relevanceofdelayedactivepowerrecoveryafterfault..................................78
5.2.5 ConsiderationsontheapplicabilityoftheresultstotheGermannetwork..79
5.3 RecommendationsforFutureResearch...........................................................................81
References........................................................................................................................................83
ix
Nomenclature
Symbols
V voltage
I current
N numberofphotovoltaicpanels
R electricalresistance
DB deadband
K proportionalgain
T timeconstant
P,Q activeandreactivepower
phaseanglein[rad]
Scaler functionusedtoscalevariables
angleforcurrentinjection
X stateofasystemvariable
stateDerivative
functionofthestate
logical function that is equal to 1 when the system is in the postfault
state
a,b loadparameterusedinpolynomialandexponentialloadmodelling
e_aP,e_aQ loadparameters usedinexponentialloadmodelling
s Laplaceoperator
logicalfunctionthatisequalto1whenthesystemisinthefaultstate
Subscripts
pv photovoltaic
s connectedinseries
p connectedinparallel
d saturation
ph photogenerated
T junctionthermal
0 initialvalue
o outputsignals
g gridquantity
V voltage
d,q componentsofspacevectorinrotatingreferenceframe
a,b componentsofspacevectorinstationaryreferenceframe
ref referencevalue
max maximumvalueusedforprotectionrelaysorcontrollerlimitations
min minimumvalueusedforprotectionrelaysorcontrollerlimitations
xi
dc directcurrent
AC alternatingcurrent
aRCI caseofadditionalreactivecurrentinjection
prefault prefaultvalues
I current
I inputsignals
& caseofadditionalreactiveandactivecurrentinjection
functionofthetracker usedindelayedactivepowerrecovery
usedingainstodefineaslowtracker
ult usedingainstodefineafasttracker
Filter functionforfilter
polynomialstaticloadmodel
exponentialstaticloadmodel
P,Q activeandreactivepower
dyn dynamicload
r ratedpower
Superscripts
arbitrarycoordinatesystem
Ug Reference frame aligned to positive sequence grid voltage at
terminals
xii
List of definitions
Shortterminterruption
Short(forlessthan2seconds)disconnectionofawindturbinegeneratorfromthegrid
viaacircuitbreakerorpowerelectronicsofthemachinesstatoruntilthefaulthasbeen
clearedfollowedbyfastrestorationofprefaultpoweroutput;usedtoberelevantforWTG
designofspecificmanufacturers[2].
Shortcircuitratio
The ratio of the transmission systems threephase shortcircuit MVA, divided by the
ratedMWofthevariablegenerationplant[3].
Activesignconvention
Alsomentionedasgenerationconventionorproduceroriented,becausepositivepower
referstoproductionofpower.Positivecurrentexitsthehighervoltage(+)ofanelement.
Reversepowerflow
The phenomenon of producing power in a lower voltage level and transporting this
powertoahighervoltagelevelthroughthenetwork.
Voltagedipdepth
Thedifferenceofthevoltagevalueataspecificelement(busbar,connectionpoint,etc.)
immediatelyafterafaultoccurredandthelowerboundaryoftheadmissiblevoltagebandat
thatelement.
Maximumpowerpointtracking
Faultridethrough
Voltage control in the event of fault in the high and extrahigh voltage network that
results in voltage dip in the medium and low voltage network aiming at avoiding
unintentional disconnection of large power infeed. That means that generating facilities
mustbeableintechnicalterms:
nottodisconnectfromthenetworkintheeventofnetworkfaults[4],[5].
Lowvoltageridethrough
Theabilityofthepowergeneratingmoduletoridethrougheventsofundervoltagesat
PCCfordefinedtimeperiods.
xiii
Dynamicnetworksupport
Voltage control in the event of faults in the high and extrahigh voltage network that
result in voltage dips in the medium and low voltage network aiming at avoiding
unintentional disconnection of large power infeed. That means that generating facilities
mustbeableintechnicalterms:
nottodisconnectfromthenetworkintheeventofnetworkfaults,
tosupportthenetworkvoltageduringanetworkfaultbyfeedingareactivecurrent
intothenetwork,
not to extract from the medium voltage network after fault clearance more
inductivereactivepowerthanpriortotheoccurrenceofthefault[4],[5].
Deadband
Dead band is the range through which an input can be varied without initiating an
observableresponse.Deadbandisusuallyexpressedinpercentofspan.Thedeadbandis
usedinaRCIcontroltodeterminetheinsensitivityofthecontroltowardsvoltagechanges.
[Owndefinition]
xiv
List of abbreviations
DG distributedgeneration
RPF reversepowerflow
FRT faultridethrough
OLTC Onloadtapchanger
rms rootmeansquare
HVdc highvoltagedirectcurrent
DSO distributionsystemoperator
LVRT lowvoltageridethrough
WPP windpowerpark
PEC powerelectronicconverter
ENTSOE Europeannetworkoftransmissionsystemoperatorsforelectricity
VDE|FNN Forumnetworktechnology/networkoperation
TSO transmissionsystemoperator
STI shortterminterruption
WTG windturbinegenerator
MPPT maximumpowerpointtracker
PLL phaselockedloop
PV photovoltaic
emt electromagnetictransient
CT clearingtime
PVPPM photovoltaicpowerparkmodule
eHV extrahighvoltage
HV highvoltage
MV mediumvoltage
LV lowvoltage
SC shortcircuit
VSC voltagesourceconverter
PCC pointofcommoncoupling
ZPM zeropowermode
aRCI additionalreactivecurrentinjection
aRACI additionalreactiveandactivecurrentinjection
nonLVRT not lowvoltageridethrough able
DFAG doublyfedasynchronous generator
FCG fullconvertergenerator
dAPR delayedactivepowerrecovery afterfault
xv
xvi
List of Figures
Figure1.1:Horizontallyoperatedpowersystem_________________________________2
Figure1.2:ReversepowerflowinsubstationinaGermany________________________2
Figure1.3:InstalledPVpowerinGermany(19982012)___________________________4
Figure1.4:ProjectedprogressininstalledPVpowerintheGermangrid(20132017) ___5
Figure2.1:Powersystemoperatingstates _____________________________________7
Figure2.2:Powersystemstabilityclassification _________________________________8
Figure2.3:FRTrequirementsfortypeIIgeneratorsintheGermanMVnetwork_______10
Figure2.4:FRTrequirementsfortypeIIgeneratorsintheGermanHVnetwork________11
Figure2.5:DGconnectedtothegridimportanceofimpedanceinaRCI______________13
Figure2.6:Reactivecurrentsupportduringfault_______________________________14
Figure2.7:Falsetripping __________________________________________________15
Figure2.8:Blindingofprotection____________________________________________15
Figure2.9:PVarraycurrentvoltagecurvewithpoweroutput_____________________17
Figure2.10:PVinvertersystem _____________________________________________18
Figure3.1:StructureoftheTestSystem_______________________________________21
Figure3.2:IEEE39bus,10machineNewEnglandtestsystem_____________________23
Figure3.3:110kVSubtransmissionring______________________________________24
Figure3.4:CigrMVNetwork ______________________________________________25
Figure3.5:CigrLVbenchmarknetwork______________________________________26
Figure3.6:PVPPMcomponents_____________________________________________28
Figure3.7:PVPPMsystemsetup____________________________________________29
xvii
Figure3.8:Powerfactorrelativetoactivepoweroutput _________________________30
Figure3.9:Typicalvoltagermsvalueplotdividedintotimeperiodsofthenetworkfault 30
Figure3.10:StateDiagramofPVPPMcontrol__________________________________31
Figure3.11:NoLVRTrequirementfortypeAgenerators,powergeneratingmoduleswitha
connectionpointbelow110kVandmaximumCapacityof0.8kWorless[13]by
Forumnetworktechnology/networkoperation(VDE)________________32
Figure3.12:DaxisandqaxiscurrentsofaPVPowerPlantsetforNoLVRTduringfault 32
Figure3.13:DaxisandqaxiscurrentsofaPVPowerPlantsetforZPMduringfault ___33
Figure3.14:DaxisandqaxiscurrentsofaPVPPMsetforaRCIduringfault__________35
Figure3.15:Inverterdaxisandqaxiscurrentdiagramsforpresentingscalingscheme_36
Figure3.16:DaxisandqaxiscurrentsofaPVPPMsetforaR&ACIduringfault_______37
Figure3.17:DaxisandqaxiscurrentinaRACIwithdelayedactivepowerrecovery____38
Figure3.18:Germangridstepresponse ______________________________________40
Figure4.1:InstalledloadandDGgenerationinthe2022ScenariofortheHV/MV/LV
networks _____________________________________________________46
Figure4.2:LoadandDGgenerationforHV/MV/LVnetworksforBasecasepowerflow_47
Figure4.3:UnidirectionalpowerflowfromeHVtoLVnetworkatBaseCase__________48
Figure4.4:LoadandDGgenerationforHV/MV/LVnetworksforHVandMVreversepower
flow__________________________________________________________48
Figure4.5:SplitreversepowerflowfromMVandHVtoeHVnetwork_______________49
Figure4.6:LoadandDGgenerationforHV/MV/LVnetworksforLVtoeHVreversepower
flow_______________________________________________________49
Figure4.7:FullreversepowerflowfromLVtoeHVnetwork_______________________49
Figure4.8:Voltageinp.u.atallbusbarsoftheeHVnetwork(differentcoloursrepresent
differentbusbars)_______________________________________________52
xviii
Figure4.9:Voltagesinp.u.atbusbarR1ofthelowvoltagenetworksconnectedtobusbar1
oftheMVnetworkswithR/Xratio4.4(differentcoloursindicatedifferentMV
networks)_____________________________________________________53
Figure4.10:Voltagesinp.u.atbusbarR15ofthelowvoltagenetworksconnectedtobusbar
1oftheMVnetworkswithR/Xratio0.43(differentcoloursindicatedifferentMV
networks)____________________________________________________54
Figure4.11:ActivePowerexchangeoveroneofthe38parallel500kVAMV/LVtransformers
connectedtobusbar1oftheMVnetworkinp.u.(differentcoloursindicate
differentMVnetworks,basevalue500kVA)_________________________55
Figure4.12:Activepoweroutputofall38parallelLVconnectedPVPPMsconnectedtobusbar
R1oftheLVnetworkconnectedtoMVbusbar1,inp.u.withanaggregated
basevalueof4560kVA_________________________________________55
Figure4.13:ReactivePowerexchangeoveroneofthe38parallel500kVAMV/LV
transformersconnectedinbusbar1oftheMVnetworkinp.u.(differentcolours
indicatedifferentMVnetworks,basevalue500kVA)__________________56
Figure4.14:Voltagesinp.u.atbusbarR1ofthelowvoltagenetworksconnectedatbusbar1
oftheMVnetworkswithR/Xratio4.4forfourdifferentcontrolmodes ___58
Figure4.15:Voltagesinp.u.atbusbarR15ofthelowvoltagenetworksconnectedatbusbar
1oftheMVnetworkwithR/Xratio0.43forfourdifferentcontrolmodes__59
Figure4.16:ActivePowerexchangeoveroneofthe38parallel500kVAMV/LVtransformers
connectedinbusbar1oftheMVnetworkforallMVnetworksinp.u.forall
controlcases(basevalue500kVA)________________________________60
Figure4.17:ReactivePowerexchangeoveroneofthe38parallel500KVAMV/LV
transformersconnectedinbusbar1oftheMVnetworkforallMVnetworksin
p.u.forallcontrolcases(basevalue500kVA) _______________________61
Figure4.18:Voltagesinp.u.atbusbarR1ofthelowvoltagenetworksconnectedatbusbar1
oftheMVnetworkswithR/Xratio4.4forfourdifferentcontrolmodesinfRPF
____________________________________________________________63
Figure4.19:Voltageinp.u.atbusbarR1ofthelowvoltagenetworkconnectedatbusbar10
oftheMVnetworkswithR/Xratio4.4forfourdifferentcontrolmodesinfRPF
____________________________________________________________63
Figure4.20:Voltagesinp.u.atbusbarR15ofthelowvoltagenetworksconnectedatbusbar
1oftheMVnetworkswithR/Xratio0.43forfourdifferentcontrolmodesin
fRPF ________________________________________________________64
xix
Figure4.21:ActivePowerexchangeoveroneofthe38parallel500kVAMV/LVtransformers
connectedinbusbar1oftheMVnetworkinp.u.forallcontrolcasesinfRPF
(basevalue500kVA)___________________________________________65
Figure4.22:ReactivePowerexchangeoveroneofthe38parallel500kVAMV/LV
transformersconnectedinbusbar1oftheMVnetworkinp.u.forallcontrol
casesinfRPF(basevalue500kVA) ________________________________66
Figure4.23:Voltagesinp.u.atbusbarR1ofthelowvoltagenetworksconnectedatbusbar1
oftheMVnetworkswithR/Xratio4.4duringZPMforallloadcases______67
Figure4.24:Voltagesinp.u.atnodeR15ofthelowvoltagenetworksconnectedatbusbar1
oftheMVnetworkswithR/Xratio0.43duringZPMforallloadcases_____68
Figure4.25:Voltageinp.u.atbusbarR1ofthelowvoltagenetworkconnectedtobusbar10
oftheMVnetworkforZPMwithandwithoutdelayedactivepowerrecovery
____________________________________________________________70
Figure4.26:Voltageinp.u.atbusbarR15ofthelowvoltagenetworkconnectedtobusbar10
oftheMVnetworkforZPMwithandwithoutdelayedactivepowerrecovery
___________________________________________________________70
Figure4.27:Voltageinp.u.atbusbarR1ofthelowvoltagenetworkconnectedtobusbar10
oftheMVnetworkforZPMwithandwithoutdelayedactivepowerrecoveryin
fRPF ________________________________________________________71
Figure4.28:Voltageinp.u.atbusbarR15ofthelowvoltagenetworkconnectedtobusbar10
oftheMVnetworkforZPMwithandwithoutdelayedactivepowerrecoveryin
fRPF ________________________________________________________72
xx
List of Tables
Table2.1:Timingfortrippingofgenerationprotection......................................................18
Table3.1:eHVimportantdataoverview............................................................................23
Table3.2:HVWindPowerParksRatings............................................................................23
Table3.3:MVWindPowerParksandPhotovoltaicPowerPlantsRatings.........................25
Table3.4:LVPhotovoltaicPowerPlantsRating..................................................................26
Table3.5:Scalingfactorsforlines,loads,transformersandPVPPMratingsperMVnetwork
nodeforalltypesofLVnetworks......................................................................27
Table3.6:ExponentialLoadParameters.............................................................................42
Table4.1:Detailsoffaultevent...........................................................................................43
Table4.2:Detailsoftransformerstapchangingstrategy...................................................44
Table4.3:MVconnectedDGinstalledpowersandconnectionpointsforthe2012scenario
.............................................................................................................................................45
Table4.4:InstalledpowerofLVconnectedPVPPMswiththeircorrespondingconnection
pointforthe2012scenario...............................................................................45
Table4.5:InstalledLoadandProductionintheeHVnetwork............................................46
Table4.6:MultipliersforinstalledloadandPVPPMspoweroutputforsettingupofload
cases..................................................................................................................47
Table4.7:ConnectionpointsofPVPPMsinthenetworkandangleofaRACIchosenforeach
point...................................................................................................................50
Table4.8:Thestudycasesformedbyvaryingthesensitivityfactorsofloadcasesand
PVPPM'scontrolmodes.....................................................................................50
TableA.1:LineparametersfortheeHVnetwork................................................................89
TableA.2:LoadparametersofeHVnetwork......................................................................90
xxi
TableA.3:ApparentpowerandpowerfactorforloadflowinitializationofeHVsynchronous
generatorsfornormalpowerflowscenario,alsodandqaxisreactanceinp.u.
...........................................................................................................................90
TableB.1:ParametersoflinetypeoftheHVnetwork........................................................91
TableB.2:LengthandtypeoftheHVnetworkslines..........................................................91
TableB.3:ParametersofeHVtoHVtransformers..............................................................91
TableC.1:LineparametersofMVnetwork.........................................................................92
TableC.2:HV/MVtransformerparameters........................................................................92
TableC.3:LoadparametersofMVnetwork........................................................................92
TableD.1:LineparametersofLVresidentialnetwork........................................................93
TableD.2:ConductorparametersforoverheadlinesoftheLVnetwork............................93
TableD.3:ConductorparametersforundergroundcablesoftheLVnetwork...................93
TableD.4:LineparametersofLVcommercialnetwork.......................................................94
TableD.5:LineparametersofLVindustrialnetwork..........................................................94
TableD.6:PhaseImpedancematrixafterKronreductionofallthreephaselinesintheLV
network..............................................................................................................95
TableD.7:MVtoLVtransformerparameters.....................................................................95
TableD.8:LoadparametersofallLVloads.........................................................................95
TableE.1:Tapchangerpositionsoftransformersperloadcase.........................................96
TableF.1:Schematicrepresentationofvoltageresults......................................................98
TableF.2:Schematicrepresentationoftransformerspowerexchangeresults..................98
TableF.3:SchematicrepresentationofDGandSGpoweroutputresults..........................99
TableG.1:Passivesignconvention....................................................................................145
TableG.2:Activesignconvention......................................................................................145
xxii
List of equations
1 __________________________________16
, 1 _________________________________________29
1 ________________________________________________29
| |
| | ,
____________34
,
,
__________________34
,
| |
&
| | _________________35
| |
&
| | _________________35
1, &
&
& , &
&
____________36
_ & _ &
0.2,
, 1
, __37
, 1
, 1
_ ________________________________________37
_ , 1
____________________________________________________40
, 1
_ _______________________________________________40
, 1
1 ____________________________41
1 ___________________________42
xxiii
_
_________________________________________________42
_
________________________________________________42
_ _
1 % _____________________________________42
_ _
1 % ____________________________________42
xxiv
Chapter 1 : Introduction
1.1 Background
Traditionallythepowersystemtransportspowerfromthepowergeneratingfacilitiesto
consumers,bothindustrialandresidential.Itiscommonpracticetoclassifythetransmission
network into the following subsystems: transmission system, subtransmission system,
distribution system [6]. The transmission system interconnects all major generation and
mainloadcentresinthesystem.Itformsthebackboneoftheintegratedpowersystemand
is operated at the highest voltage levels (typically in Europe, 150kV and above). Power is
transmitted to transmission substations where the voltages are stepped down to the sub
transmission level (typically in Europe, 110kV). The subtransmission system is the
intermediatesystemthattransferspowerfromthetransmissiontothedistributionsystem
and feeds large industrial customers. Finally the distribution system is the last stage of
energy transfer and delivers power to the individual customers. Voltages in primary
distribution feeders range from 4kV to 34.5 kV and deliver power to small industrial
customers whilst the secondary feeders supply residential and commercial customers at
240V[6].
Thistypeofhierarchicalpowersystemstructurecanbedefinedasaverticallyoperated
power systemas described in [7]. This working principle, however, has been changing and
willfurtherchangeinthefuture.Localdistributedgeneration(DG)sourcesareintroducedin
the network, mainly due to the integration of renewable sources. In general, DG can be
definedaselectricpowergenerationwithindistributionnetworksoronthecustomersideof
thenetwork[8].Thepowerratingofthesesourceshaslargevariationandmanyofthemare
connectedonthesubtransmissionanddistributionnetworks;evensomeatthesecondary
feedersofthedistributionnetwork,i.e.photovoltaic(PV)onroofs.
Figure1.1:Horizontallyoperatedpowersystem[ownrepresentationbasedon7]
Figure1.2:ReversepowerflowinsubstationinaGermany[9]
This transition from a vertically to a horizontally operated power system holds many
potentialchallengesforthesecurityandstabilityofenergytransfer.Animportantchallenge
isthebehaviouroftheprotectivesystemincaseofafault,andalsotheloadandgeneration
balanceincaseofmassivedisconnectionofDG.Ingeneralthebehaviourofsuchasystem
must be thoroughly investigated under all possible circumstances to ensure proper
operation.
An interesting case is a severe fault, i.e. system fault, loss of generation or circuit
contingency,inthetransmissionlevelataveryhighinfeedofDGpowerintothegrid.The
behaviour of the system for a serious disturbance during reverse power flow (RPF) is of
specialinterest.Thescenariomentionedabovewillresultinavoltagedipthatwillpropagate
tothedistributionsystemandmighttriggerthedisconnectionofconnectedDGindifferent
voltagelevelsdependingontheirfaultridethrough(FRT)capabilitiesandthereforeposea
threatto systemstability.Networkcodes (sets of rules that apply tooneormore partsof
2
the energy sector and provide for agreements by relevant network operators on various
technicalrequirements[12])forDGgeneratorsvariesfordifferentvoltagelevelsandtheFRT
requirementsarealreadyappliedforthehighandmediumvoltagelevelsbutnotforthelow
voltageconnectedDGs[1012].
The need for FRT capability is not yet documented for the low voltage distribution
system but a lot of discussions already refer to the possibility of including FRT capabilities
evenfortypeAgenerators,thatispowergeneratingmoduleswithaconnectionpointbelow
110kVandmaximumcapacityof0.8kWorless[13].Therealizationthatanincreaseinthe
capacityofinstalledDGpowerinthespecificvoltagelevelcantriggerdisconnection,incase
oftransmissionfaults,ofmassiveamountsofpowerfromDGandjeopardizesystemstability
isforming.
NumerousorganizationsarestartingtoconsiderFRTimportantforlowervoltagelevels
and smaller generators, amongst them the VDE Association for Electrical, Electronic and
InformationTechnologiese.V,theEuropeanNetworkofTransmissionSystemOperatorsfor
electricity (ENTSOe) and the European Committee for Electrotechnical Standardization
(CENELEC)[1418].
The objective of this thesis is to investigate the response of low voltage distribution
systemswithhighpenetrationofPVincaseofafaultinthetransmissionsystemanddefine
aFRTvoltagecurveappropriatetolowvoltageconnectedPV.
Theresearchquestionsare:
Ifcurrentrequirementsareappliedwillthesystemstaywithinreasonablestability
boundaries?Orwillitbeleadtoinstability?
Ifso,whatarethenecessarymeasuresthattheTSOsand/orDSOsneedtotakein
ordertoavoidthispossibility?IsFRTcapabilityimportantandrelevantforthelow
voltagedistributionsystem?
WhatarethecriticalpowerflowsituationsthatwouldleadLVconnectedDG
todisconnectincaseoftransmissionandsubtransmissionfaults?
What is the sensitivity of the voltage, during fault, in lowvoltage networks
with large amounts of generation from LV connected DG installations w.r.t.
networktopologyandbehaviourofthePVinstallation?
Whichamendmentstointerconnectionstandards(gridcodes)ofDGinlow
voltagenetworksarerecommended?
The problem will be analysed through root mean square (RMS) stability simulations
performedonappropriatemodelsofthenetworkandtheDG.Theprojectedyearthatthis
studytakesplacewillbe2022andtheDGcapacitywillbechosenforthatyear.Themodel
3
will be implemented as a benchmark system with DG capacities chosen to match the
Germannetwork.ThecommerciallyavailableDIgSilentPowerFactorysoftwareversion14.1
will be used to carry out these simulations. Furthermore a sensitivity analysis will be
performed w.r.t. load, generation and behaviour of PV to transmission system faults.
Selected simulation results will be presented in graphs and analysed. These results will
provide the basis for implications on grid connection requirements for low voltage
connected DGs and finally recommendations for future grid connection requirements for
lowvoltageinstalledDGwillbegiven.
Agoodcandidateforacasestudyisanetworkthathas,regionally,ahighpenetrationof
DG. In our case the German network, which has been pioneered in renewable integration
and has a major amount of photovoltaic installed power in the low voltage distribution
networkwillbeconsidered.Thefollowinggraphs(Figure1.3,Figure1.4)areindicativeofthe
development of installed PV power over the last decade and the foreseen development
based on a linear projection. Numerous reasons have led the increase of solar PV plants
capacityoverthelast15years.MostsignificantreasonsarecostreductionofPVmodules,
marketincentivesforelectricityproduction,socialawarenessaboutglobalwarmingetc.
Figure1.3:InstalledPVpowerinGermany(19982012)[19]
From these graphs it is clear that PV deployment has been massive in the German
network: The installed photovoltaic capacity in Germany on the 31st of January 2013 was
32,663Mwp according to [20]. Moreover by the end of 2010, approximately 80% of
cumulative installed photovoltaic (PV) capacity, roughly 14GW, was connected to the low
voltage distribution level, as reported by [21]. The high penetration of PV in the German
networkjustifiesthechoiceofthisnetworkforthisstudy.
Theconclusionsthatwillbedrawnfromthisresearchwillnotonlyberestrictedtothe
specific network but to all networks that share similar characteristics and have a high
amount of installed DG power. The choice of benchmark networks used ensures that all
othernetworkcharacteristicswillbesimilartomostoftheEuropeannetworks.
4
Figure1.4:ProjectedprogressininstalledPVpowerintheGermangrid(20132017)[22]
Chapter1givesanintroductiontothestudytopic.Inthischapterthebackgroundforthe
research topic is explained, and the objective and research questions of the thesis are
stated.
In Chapter 2 the literature survey is presented. The important, for this study, power
system stability issues are explained, and an analytical research on current grid codes is
presented.TheimpactofFRTcriteriaonlineandgenerationprotectionisdiscussedaswell
asthecurrentstateofphotovoltaicsysteminverters.
InChapter3thetestsystemusedforthisstudyispresented.Themodellingapproachof
different networks for different voltage levels is explained. The analytical model of the
PVPPM is presented along with all the additions forFRT capability. Finally theWPP model
andthemodellingoftheloadsaredescribed.
InChapter4thecasestudyandthesimulationresultsarepresented.Thescenariosfor
theinstalledDGcapacityofthesystemarepresented.Thechoiceofsensitivityfactorsthat
formulatethedifferentstudycasesisjustifiedandthesensitivityfactorsareexplained.The
presentationandanalysisofresultsfordifferentstudycasesfollows.
Finally, in Chapter 5 the conclusions drawn from the results are discussed. Also
recommendations for further research are proposed. The general findings of the research
are presented first and then the implications for German grid codes and the German
networkarepresented.Therecommendationsforfutureworkclosethischapter.
6
Chapter 2 : Literature Survey
Power systems stability is the ability of an electric power system, for a given initial
operating condition, to regain a state of operating equilibrium after being subjected to a
physical disturbance, with most system variables bounded so that practically the entire
systemremainsintact[23].
Inordertounderstandindepththetermpowersystemstability,itisimportanttohave
anoverviewoftheoperatingstatesofapowersystemasseeninFigure2.10.
Figure2.1:Powersystemoperatingstates[6]
The five operational states of the power system are: normal, alert, emergency, in
extremis, and restorative. In the normal state, all system variables are within the normal
rangeandthesystemisabletowithstandacontingencywithnooverloadingofequipment.
The alert state is the same as the normal state but a possible contingency can cause
problems in the network. If a sufficiently severe disturbance occurs when the system is in
thealertstatethentheemergencystateisentered,wherevoltagesatmanybusesarelow
and/orequipmentloadingsexceedshorttermemergencyratings.Thenemergencycontrol
actions take place: fault clearing, excitation control, generation tripping, generation run
back, HVDC modulation, and load curtailment. If the above measures are ineffective, the
system is in extremis; theresult is cascading outages and possiblya shutdown of amajor
portionofthesystem.
Control actions, such as load shedding and controlled system separation, are aimed at
savingasmuchofthesystemaspossiblefromawidespreadblackout.Finallytherestorative
stateisreachedwhenactionsaimatconnectingallthesystemandrestoringtheproblems
[6].ThestateweareinterestedinistheemergencystateandhowtodefinecorrectlyFRT
criteriainordertoavoidmovingintotheinextremisstate.
Thereforeforourmodelling,faultclearingtimeandfastcontrolactionsarerelevantto
describe the situation during and immediately after a fault. We propose classifying the
stability studied in this thesis as large disturbance voltage stability for the transient fault
7
responseofthesystem.Weareinterestedincheckingwhetherbusvoltagesarekeptwithin
limitsinordertoavoidmajorlossofDGgenerationduetoundervoltageprotectionduringa
fault. The study period is short term and can reach up to tens of second if the postfault
systemisincluded.ThisclassificationisshowninFigure2.2,wheretherelationbetweena
shorttermstudyperiodandlargedisturbancevoltagestabilityisshown.
Figure2.2:Powersystemstabilityclassification[23]
Most definitions for voltage stability refer to longer time frames and incorporate the
responseofOLTCs(OnLoadTapChangers)andgeneratorfieldcurrentlimiters[6].However
transient stability refers to the short transient period but deals mainly with rotor angle
stability and not voltage profiles. Solidstate inverters do not behave like rotating
generators. Solidstate inverters have no inertia in their output and can respond
immediatelytochangesintheacpowersystem.[24].Basedonthisclaimwecansaythat
thelackofinertiaofaninvertermaybethereasonwhywecannoteasilyclassifyourstability
issue. However the classification of stability that is based on [6, 23] in our study is large
disturbancevoltagestability.
On the other hand DGs ability to solve network problems has been debated for
distributionnetworks.Unfortunately,thedistributionsystemoperator(DSO)hasnocontrol
or influence on DG location and size below a certain limit. Nevertheless, DG placement
impacts critically the operation of the distribution network. Inappropriate DG placement
may increase system losses and network capital and operating costs. On the contrary,
optimalDGplacementcanimprovenetworkperformanceintermsofvoltageprofile,reduce
flowsandsystemlosses,andimprovepowerqualityandreliabilityofsupply[25].
TherearemorebenefitsthatDGscanofferintermsofpowersystemstability.Ifpower
electronic converter (PEC) coupled DGs are disconnected from the network during fault,
then maximum rotor acceleration is restricted and thus transient stability is benefited. On
the other hand large penetration levels of DG result in imbalance of load and production
duringfaultifthesamepracticeisfollowed,resultinginpoorfrequencyandvoltagestability.
However,keepingtheDGconnectedcanhelpdampoutoscillationsfasterpostfaultdueto
theimmediatereactionofthePECthatisnotboundbyitsinertia,whichcansupplyreactive
8
powerduringfaultandcontributetothefaultcurrent.Incontrasttoelectricalmachinesthe
short circuit current of a PEC is limited by its protection and exceeds the nominal current
onlyslightly[26].Alltheseresultsarepresentedin[27].
TomitigatetheeffectofDGonpowersystemstabilitygridoperatorshavedefinedfault
ridethroughcriteriaforDG(seesection2.2).Thesecriteriaarealreadyineffectforpower
plants connected to higher voltage levels but now also become operational for DG
connectedtodistributiongrids.Faultridethroughcriteriadefineforwhatvoltagedipswith
specific duration the DG have to stay connected to the grid. Depending on the size of the
DGunit the fault ridethrough criteria can also require additional grid support, such as
voltageandfrequencysupport,duringandafteradisturbance.
An important feature of grid connected power generating facilities is the Fault Ride
Through(FRT)orLowVoltageRideThrough(LVRT)capability,definedin[1]specificallyfor
PVsasthe:PVinverters'capabilityofremainingconnectedtothegridintheeventofgrid
failures,ofnotsupplyinganyactivepowerduringagridfault,andofdeliveringactivepower
directlyafterclearingthefault,thusstabilizingthegrid.Secondlythereisthecapabilityof
injectinganadditionalreactivecurrentintothegridincaseofgridfault,withthepurposeto
providevoltagesupportduringfaultconditions.
SomeGermangridcodes(GC)havealreadyspecificrequestsforinvertercoupledDGs.
TheserequirementshavearisenfromthelargepenetrationofPVtechnologyintheGerman
grid.ThewayDGwasregardedbyTSOSinthepastwasasapassivecomponent;thusnot
allowed to take any type of action during fault. However more contemporary grid codes
changethisviewandintroduceFRTcriteriaforPVplantsconnectedtotheMV(since2009)
[5] subtransmission network and transmission network (since 2007) [28]. Within these
codessolarplantsneedtobeabletosustaincertainamountofvoltagedipsforspecifictime
frameswithoutdisconnectingfromthegrid.Thebasictypeofrequirementisprovidedasa
voltageversustimegraphat the grid connection point for a generating facilitydepictingthe
behaviourofasystemwithincertainareasofthe limiting voltage curves, an example can
be seeninFigure2.3[28].Analternativeapproachispresentedat[3]whereadifferenttype
of curve for voltage profiles that defines FRT requirements is presented. The cumulative
severitydurationtypeofspecificationismorecloselyassociatedwiththeactualstresseson
equipmentandtheactualbehaviourofmanyprotectiverelays[3].
Besides meeting the specifications for FRT curves during the fault period there is a
requirementforsolarplantstoprovideadditionalreactivecurrentinjectioninordertoaidin
maintaining the voltage profile of the network. Maintaining the voltage profile of the
network by limiting the dipdepth is desirable during a fault because it prevents
disconnectionofoldgenerationplants(e.g.windfarms)thatarenotabletoridethrougha
faultandingenerallimitstheseverityofthefaultasseenbyallpowergeneratingfacilities.
IntheGermanGCformediumvoltagenetworks,enteringintoforceonApril1st,2011,
PV plants have to be capable to participate in complete dynamic network support during
9
grid faults [29]. Complete dynamic network supports is defined as voltage control in the
eventoffaultsinthehighandextrahighvoltagenetworkthatresultinvoltagedipsinthe
medium and lowvoltagenetworks aiming at avoiding unintentional disconnectionof large
feedinpower,andthusnetworkcollapse.Thatmeansthatgeneratingfacilitiesmustbeable
intechnicalterms:
Nottodisconnectfromthenetworkintheeventofnetworkfaults
Tosupportthenetworkvoltageduringanetworkfaultbyfeedingareactivecurrent
intothenetwork,
Not to extract from the medium voltage network after fault clearance more
inductivereactivepowerthanpriortotheoccurrenceofthefault[10],[5].
The voltage profiles in the following figures refer to a type II generating unit [28],
definedasallgeneratorunitsthatarenotsynchronousgeneratorsdirectlyconnectedtothe
network.
InFigure2.3theMVvoltageFRTrequirementsareshowninnovelgraphsbasedonthe
technicalguidelineforgeneratingplantsconnectedtothemediumvoltagenetworkbythe
German Association of Energy and Water Industries (BDEW) [5]. The requirements specify
that voltage dips with values above the red dashed line must not lead to instability or to
disconnection of the generator from the network. In the area between the two red lines
generating units should pass through the fault without disconnecting from the network.
Furthermore, in consultation with the network operator, a shortterm interruption (STI)
fromthenetworkisallowedifthegeneratingplantcanberesynchronizedwithin2seconds,
at the latest, after disconnecting. The STI term was originally used for a wind turbine
generator(WTG)ofspecificmanufacturerswherethegeneratorconceptusedacrowbarat
thestator[2].Afterresynchronization,theactivepowermustbeincreasedwithagradient
of at least 10% of the nominal capacity per second. Below borderline 2, a shortterm
disconnectionofthegeneratormaybecarriedoutwithlongerreconnectiontimesandlower
gradientsoftheactivepowerincomparisonwiththepreviouscase.Belowthebluelineno
requirementsforFRTapply.
Figure2.3:FRTrequirementsfortypeIIgeneratorsintheGermanMVnetwork
10
InFigure2.4therespectiveHVFRTrequirementsareshownbasedon[28].Thegreen,
yellowandgreyareascorrespondtothesamerequirementspresentedfortheMVnetwork.
The area between 1.5 and 2.4 seconds corresponds to stepwise tripping of the faulted
generatorswithintheplant.Themaindifference,however,isthelackofthebluelineand,
therefore,theHVFRTrequirementsareslightlystricterthantheMVFRTrequirements.
Figure2.4:FRTrequirementsfortypeIIgeneratorsintheGermanHVnetwork
ThereforeintheMVandHVnetworksFRTcriteriaarealreadyineffectforDG.However
that does not hold true for the LV distribution network. In the LV network DG units are
requiredtointentionallydisconnectfromthegridduringadip.Thetwomostrelevantand
important standards are the IEEEStd. 1547 that was created in 2003, and the VDEARN
4105Std. from 2011. Both of these standards require a mandatory disconnection of LV
connected DG during fault. The disconnection is based on the OR logic, by monitoring all
phases separately, if one phase is below the voltage threshold set from the protection
devicethenthemoduledisconnects[10,11].
Thefundamentaldriversforthesestandardsweremany.Animportantconcernwasthe
possibleinterferenceoftheDGunitswiththeprotectionschemesofthenetwork.Another
reason is the sensitivity to overcurrents and overvoltages by the power electronic
converters(PECs)couplingtheDGunitstothenetwork[26].Furthermorethesestandards
were created at a time when high penetration of DG was not envisioned. Therefore as
observedin[30]themandatorydisconnectionrequirementnegativelyimpactsbulksystem
reliabilityinahighDGpenetrationscenariowhereexcessivelossofgenerationmightresult
insysteminstability.MoreovercertaincategoriesofDG,e.g.PV,canusethereactivepower
capabilitythatthepowerelectronicsconverter(PEC)providestoensureminimumdeviation
intheirvoltageprofileatthepointofcommoncoupling(PCC),pointinthepublicnetwork
closesttothecustomersystemtowhichfurthercustomersystemsareconnectedorcanbe
connected[10],howeverbasedonIEEE1547standardsDGmaynotchangetheirreactive
poweroutputindirectresponsetothemagnitudeofthevoltage[31].
Atthispointitisgoodtoobservethatthesestandardsaremerelyguidelinesthatneed
toberevisedandreconsideredcontinuouslytobekeptuptodate.Thehypothesisthatin
networks with a large DG penetration level the disconnection of all DG units can cause
11
serious problems as it may result in a large power generation deficit and in stability
problems is worthy of investigation. In addition a draft of the ENTSOE Network Code
RequirementsforAllGeneratorsincludedarequirementevenfortypeAunits,whichare
generatingunitswith0.8KWratingormoreconnectedin110KVorlowervoltagelevel,not
to trip in vast numbers simultaneously in response to a single fault at the highest voltage
level[13].EventhoughthisopinionoftheENTSOEdidnotbecomepartofthefinalNetwork
Code,itrevealstheneedforadeeperinsight.Furthermorein[15]ENTSOEwelcomesthe
intentionofCENELECtointroduceaFRTrequirementforgeneratorsconnectedatmedium
and low voltages (>16A/phase), which would also extend to include type A generators.
Furthermore CENELEC states that the industry is ready for this requirement as many
technologies have this capability already. Finally new requirements of FRT for PV systems
connected to a single phase on the 200V low voltage distribution feeder were recently
investigated and proposed to cope with expected future large penetration. This analysis
occurred as part of Japanese research governmental program contracted with the NEDO,
JapanslargestpublicR&Dmanagementorganization[32].
Whendefiningnewrequirementsfornetworkconnecteddevicesitisveryimportantto
take into account the impact these measures will have on the device itself and the whole
network. This section focuses on the latter. Important parts of the network are the
protection devices and schemes used to ensure proper function of the network during
faults.Inordertounderstandtheimpactofthesecriteriaontheprotectionsystemofthe
networkabetterunderstandingofhowprotectiondevicesworkisneeded.
Onthedistributionlevelsimpleprotectiveschemesareusedandtheiroperationfocuses
onthedetectionofafaultcurrentsignificantlylargerthanthenominalcurrentofthegrid
components [7]. The protection relays used for sucha task are the overcurrent relays. To
operatethisrelayonlythecurrenthastobemeasured.Whenthecurrentexceedstheset
limit,ittriggersatripsignalthatissenttothecircuitbreakerdisconnectingthefaultedline.
However, different relay characteristics based on the principle of operation can be
developed by varying pickup current and time response of the device. Two important
featuresoftheprotectivesystemareselectivityandsensitivity.Meaningthattheprotective
systemmustbesensitiveenoughtofaultsbutalsodisconnecttheleastpartofthenetwork
needed to clear a fault without interfering with the healthy part of the network. For that
reason, most protective systems overcurrent relays have to be coordinated with other
protectiondevices[7].
It is not in the focus of this Thesis to go into lengths analysing the protection
mechanisms but rather to provide a fundamental insight of the principle of operation in
order to assist in the understanding of the interference of the FRT and reactive current
injection of the DG units on protection. However the most important protection schemes
areexplained:
12
Reclosurefusecoordination
Basedonovercurrentrelaysthatarecommonlyusedonoverheadlineswherebrief
interruption and quick reconnection is desirable due to nonpermanent faults, e.g.
lightning.
Differentialprotection
Distanceprotection
AprotectiondeviceisusedwhichmakesuseofthelinearrelationbetweenUandI.
Voltageandcurrentaremeasuredinordertocalculateimpedance.Achangeintheratio
indicatesaphysicalalterationintheexistingcircuitandthusitisdeenergized[7].
Forreadersthatwishtolearnmoreonprotection,chapter3.4from[7]hasadescription
ofaprotectivesystemforsimpleradialandmeshedgridsindicativeofEuropeangrids.
TheaddedrequirementthatisalreadyincorporatedintotheMVandHVgridcodesof
Germany is the additional reactive current injection (aRCI) from DG generators during a
fault.ThefundamentalconceptbehindthisrequirementisthatduetothehighX/Rratioof
the transmission and distribution networks, reactivecurrent raises thevoltagelevel of the
network. As seen in [33] the higher the X/R ratio the reactivecurrentcontributesmoreto
raisingthenetworkvoltage.Theconceptcanbeeasilydemonstratedthroughasimplyradial
networkrepresentationasshowninFigure2.5.
Figure2.5:DGconnectedtothegridimportanceofimpedanceinaRCI
Assume that during the prefault period active power is pushed into the transmission
systembytheaggregatedDGandinordertolimitthevoltageriseDGisconsumingreactive
power.Duringafaultinsidethetransmissionnetworkthevoltageofthebusbarislowered
significantly by the propagation of the voltage dip to the distribution system. In order to
13
raisethevoltageareactivecurrentcomponentadditionaltotheprefaultvalueisprovided
bytheDGand,therefore,thetotalreactivecurrentconsumedatthebusbarisdecreased
and can even become negative, i.e. reactive power is fed into the network. Therefore the
reactivecurrentthatispassingthroughthereactanceofthelinesisraisingthevoltagelevels
atbusbarsbetweenthefaultandthePCC.TherequirementsforaRCIareshowninFigure
2.6.
According to this requirement DG generators have to supply at least 1.0 p.u. reactive
current when the voltage falls below 50%. A dead band of 10% is introduced to avoid
undesirable control actions [34] and to prevent conflicts with power factor control during
steadystate.
Figure2.6:Reactivecurrentsupportduringfault[34]
AfterexplainingtheprinciplesofprotectionandintroducingaRCIwehaveallthetoolsto
understand the reason FRT requirements can disrupt the current protection mechanisms.
IntegrationofDGnotonlyaltersthepowerflowbutwillalsochangethefaultcurrentsinthe
distribution grid. These changing fault currents can affect proper operation of the current
protective system. The three main issues we can identify are false tripping, blinding of
protectionandreversefaultcurrents.
Forthesereasonsthestandardpractice,uptoafewyearsback,takenbygridoperators
wastodemandimmediatedisconnectionoftheDGunitsincaseofafaultorshortcircuit.
Immediate disconnection of DGunits restore the distribution grid to a grid with only one
source of supply and the protective system can function as it was implemented to do so
duringthedesignstageofthedistributiongrid[7].TheundervoltageprotectionoftheDG
woulddisconnectthemfromthenetwork,howeverduetothelargepenetrationlevelofDG
this practice is not applicable any more for the HV and MV networks. False tripping can
occur in case of a faulted feeder when DG connected to this or adjacent feeders can
contributetothefaultcurrentandasaresult,theprotectivesystemoftherespectivefeeder
wouldunnecessarydisconnecthealthyfeeders.TheschematicrepresentationofFigure2.7
canassistinunderstandingthisphenomenon.TheDGlocatedontheadjacentfeederofthe
14
faultedonefeedscurrenttothefaultlocation.Thiscurrentcanexceedthepickupcurrentof
theDGfeederprotectionanddisconnectthehealthy(DG)feeder.
Figure2.7:Falsetripping[7]
Figure2.8:Blindingofprotection[7]
Thepreviousissueswhilevalidareoutofthescopeofthisthesisastheyarereferringto
faults on the distribution level, while we are considering faults on the transmission level.
However there are a lot of proposed solutions to overcome these issues and interested
readerscanfindthesesolutionsin[7,24].Morren[33]evendisagreeswiththestatement
that DG can prevent the proper operation of the feeder protection and proves that
converterbasedDGcanlimittheircurrentduringafaultand,hence,minimizetheinfluence
onthenetworkprotection.TheissuethatisrelevantforthisThesisisthereversepowerflow
duringfaultthatcandisrupttheprotectivesystem.
Inradialdistributiongridsnormallythereisonlyonesourceofsupplyandapplicationof
overcurrent relays as protection devices are prevalent. However, selective protection with
overcurrentrelaysofdistributiongridswithameshedstructure,orwhenbidirectionalfault
currentsoccurcancauseundesireddisconnectionofcircuits.Toprovideproperprotection
ofthesegridstructuresandcopewithbidirectionalfaultcurrentsadirectionalelementhas
15
tobeaddedtotheovercurrentrelay.Thisdirectionalelementdeterminesthedirectionof
thepowerflowintheassociateddistributionfeeder.Thepowerflowdirectionisdetermined
withareferencesignalwhichisoftenavoltage.
Theserelayswillissueatripsignaltothecircuitbreakerwhenboththedirectionsetting
andthecurrentpickupvaluearemet.Sotheserelaysaresetinthesamemannerassimple
overcurrentrelaysbutalsoaspecificdirectionischosen.Byusingtworelaysetups,eachfor
different direction, for each line the protective system can be adapted to handle bi
directionalcurrentswhileensuringsensitivityandselectivity.Thisadaptationisadequateto
ensurethattransmissionfaultswithFRTcriteriaforDGontheLVdistributionsystemwillnot
disrupttheprotectivemechanismsofthenetwork[7].
TheenergycapturedfromaPVarrayisproportionaltotheirradianceanddependsupon
the location of the operating point of the solar panels. The formula that describes the
relationshipbetweencurrentandvoltageinaPVarrayisgiven(Equation21).
exp 1 Equation21[35]
With V :PVarrayoutputvoltage
I :PVarrayoutputcurrent
n :Numberofpanelsinseries
n :Numberofpanelsinparallel
I :Darksaturationcurrent
R :Cellseriesresistance
I :Photogeneratedcurrent
V :Junctionthermalvoltage
In Figure 2.9 a graphic representation of this formula for a given temperature and
irradiance is depicted. The operating point in this curve can be shifted and the power
productionischangedaccordingly.
16
Figure2.9:PVarraycurrentvoltagecurvewithpoweroutput[35]
Formaximizationofproducedpowertheusualpracticeistouseanalgorithmthatshifts
theoperatingpointofthecurveofFigure2.9.Thealgorithmiscalledmaximumpowerpoint
tracking(MPPT)anditisappliedontheDC/DCconverterthatconnectstheoutputofthePV
panels to the DC/AC inverter that is used for grid connection. This setup is called a two
stagePVsystems,whichisaPVsystemwithaDC/DCconverterandaDC/ACinverter.Inone
stage PV systems (only inverter) this function is performed by the DC/AC inverter. This
algorithmismodulatingthefrequencyoftheswitchingelementsinsidetheDC/DCconverter
toregulatetheDClinkvoltage.
Thephaselockedloop(PLL)algorithmthatisimplementedontheDC/ACinverterisof
great importance to the PV grid connection. This is because it must guarantee the
synchronization of the inverter voltage vector with the grid voltage vector, through the
detectionofitsphaseangle.Thegridvoltagesaretheinputintothealgorithmandtheyare
transformed into synchronous reference frame by means of the abc / dq transformation
module.Thephaselockingisrealizedbycontrollingthedvoltagecomponenttobezero.Ud
iscomparedwithitsreferencevalueandwithaPIregulatorthevalueofgridfrequencyis
calculated. After the integration of the grid frequency, the voltage angle is obtained.
Nowadays, the PLL technique is the stateoftheart method to extract the phase angle of
thegridvoltages[1].
ArepresentationofthestateoftheartPVinvertersystemcanbeseeninFigure2.10.It
usesbothPLLandMPPTalgorithms.ConcerningLVRTcapability,wemustconsiderthatthe
main reasons for inverter disconnection during voltage dips are excessive DC voltage,
overcurrent due to low voltage (in constant P operation), and loss of synchronism [1]. In
particular, overcurrent occurs because the DC link voltage control s generally designed to
keep the DC voltage equal tothe reference (MPP). In this way, the active power is always
constant resulting in distorted current when the voltage is unbalanced. However modern
invertersarecapableofenforcingFRTwithoutencounteringproblems.
17
Figure2.10:PVinvertersystem[32]
InourmodellingofthePVinverterwearegoingtouseaDigSILENTPowerFactorymodel
developedbySamadi[36].InthepaperbySamadithemodelisvalidatedincomparisonwith
a PSCAD model to ensure that the results from an average modelling program such as
PowerFactorydonothaveanysignificantdifferencewithanEMTmodel.ThePLLalgorithm
and the MPPT are included, as well as active and reactive power control.More details for
themodelwillbediscussedlateron.
IngeneraltheprotectionmechanismsthatdisconnecttheDGinlowvoltagedistribution
networksareunderandovervoltage(andalsofrequency)protection.Itiswithinthescope
ofthemodellingtorepresenttheprotectionmechanismsthatdisconnectthePVplantfrom
thenetwork,alsoknownasgenerationprotection.ThetimingsaccordingtotheVDEARN
4105standardforlowvoltageconnectedgeneratorsarepresentedinTable2.1.
When we apply the standards used nowadays, the protection is set to disconnect DG
fromthenetworkintheeventofafaultbasedonTable2.1values.ButwhenFRTcriteriaare
developed these values have to be adapted to ensure that the protection devices do not
disconnecttheDGfromthenetworkbeforeasettime.
FurthermoretheexistenceofaDC/DCconverterwillnotbemodelledbecause,although
itispartoftheDGinterfacewiththenetwork,itsdynamicresponseissofastthatitcanbe
ignoredwithoutaproblem.Typicalresponsetimesareintheorderof5 asknownfrom
discussionswithSMAandTUBraunschweig[37].
Finallyitisworthnoticingthatin[32],resultshaveshownthatpostfaulttherearesome
inverters that disconnect due to under voltage of the DC link. As explained in [32], this
phenomenonoccursatvoltagerecoveryduetorapidacvoltageriseatPCCresultinginaDC
undervoltage. However the discussions with SMA and TU Braunschweig revealed that for
theEuropeanPVmarketportfoliosuchbehaviourisnotrelevant.
18
2.5 Types of Faults
Thefaultsweareconsideringinthisthesisaretransmissionsystemfaults.Thesefaults
occur at the HV network and the related voltage dip may propagate throughout the sub
transmissionanddistributionnetwork.Avoltagedippropagationresultsinanundervoltage
in the bus bars of the low voltage connectedDG, whose behaviour we wish to investigate
under fault. The clearance time (CT) of the fault in the HV network is very small. Values
between 90 and 120 are common [7], however to ensure a proper safety margin and
basedontheexistingFRTcurvesavalueof150 forCTischoseninthesimulations.
Thetypesoffaultsthatwewillconsiderareonlysymmetricthreephasefaultswhichare
themostsevereonesfromthenetworkperspective.Thesetypesoffaultsarealsoeasierto
study given that we have a three phase PV system model. Also the propagation of the
voltagethroughdifferenttransformersdoesnotaffectthevoltageprofileinthesetypesof
faults.
2.6 Conclusions
In this chapter the literature survey was presented. The nature of the investigation in
termsofnetworkstabilitywasexplained,alsothecurrentgridcodestandardswereanalysed
aswellasthecurrentstateofPVinverterstechnology.Theimportanceofprotectionscheme
choicewasanalysedandthepossibilityofinterferencewithDGbehaviourincaseofafault.
Finallythereasoningforthechoiceoftypeoffaultstudyispresented.
19
20
Chapter 3 : Test System and Distributed Generation Modelling
3.1 Introduction
Inthischapterthemodellingapproachofthestudyisexplainedandadetailedreviewof
all components of the test system is included. Initially, the structure of the test system
comprisingofdifferentvoltagelevelsispresented.Thissetupallowsforasystemicapproach
and analysis of photovoltaic power park modules (PVPPM) contribution to system
performanceandstability.ThesecondessentialcomponentisarealisticmodelofaPVPPM
with adequate control functionalities able to perform different levels of dynamic voltage
support. This model will aid in understanding and evaluating the impact of low voltage
connected,invertercoupled,generatorscontributiontosystemstabilityduringtransmission
faults.Thewindpowerplant(WPP)usedisalsodescribedaswellasloadandlinemodelling.
ThestructureofthetestsystemispresentedinFigure3.1.Thedifferentvoltagelevels
aremarkedfromtheextrahighvoltage(eHV)network(blue)tothelowvoltage(LV)
network(green).
Figure3.1:StructureoftheTestSystem
The eHV network (220kV) has one connection point (two transformers) to a HV
transmission ring (110kV). This transmission ring connects to six identical MV networks.
21
Each one of them connects to two LV networks. This structure represents a complex
multilevelsystemapproachtoresembleanactualpowersystem.
The aim is to investigate how the low voltage connected PVPPM controls will affect
system stability following a network fault in the transmission system. The added value of
usingsuchacomplicatednetworkistheexpectationtobeabletorecognisemorerealistic
interactions that are present in an actual network rather than focus on the fundamental
relationshipsbetweendifferentcomponents.
TheeHVnetworkchosenforthisstudyisaDIgSILENTPowerFactorymodelofthe39bus
10machines New England network validated by IEEE, for stability controlled performance
[38].ThesinglelinediagramofthenetworkisshowninFigure3.2.
The network has 10 generators connected in different busbars (BB) via step up
transformers.Thenominalfrequencywas60Hz.Voltagelevelofthereceivednetworkwas
originally345kVbutwaschangedinto220kVand50HzinordertoresembletheGerman
transmission system. The voltage at the generation buses is 22kV. The connection point
withtheunderlayingintegratedHV,MV,LVnetworkisplacedatBB8(bluearrows)viatwo
150MVAtransformers.
The connection point for the HV network was chosen after short circuit power (SC)
calculations. The SC capacity of a busbar is the product of the magnitude of the prefault
voltage and fault current.It is ameasureofthe strength of the system atthat connection
point and, therefore, of the ability to retain its voltage under system faults; on the other
hand, the busbar with the lowest SC capacity will be the one affected the most by the
responseofthedistributionsystem.Inordertoconsidertheworstcase,theweakestBBis
chosen. The most important parameters of the network are presented in Table 3.1, more
detailscanbefoundinAppendixA.
22
~
G
Area 1 ~
W ind Lignite Area 3 ~
G G
Coal
41_WPP 37
~
44 46
G 26 29 38
Gas 41
43 45
47 ~
30 28 24 G
42 CCG T 38
40 29
25 27
16 35
5 4 28
31 37
26 25
21 22
02 17
35
15
1 3 48
18
7 30
24 36
01 03
27
40_WPP
6 14
2
23
9 Wind
G
~
39 04
12 19
8 33
23
G 05
~ 32
33
C oal 10 22 21 39
06 20
13 G as
13 G
11 ~
17 12
11
14 34
to HV 07
19
36
18
34
15 CC GT
31 10 G
~
08
CC GT
G
16 Wn
i d 20 ~
G
~
32
09 to HV Gas
G
Area 2 ~
Figure3.2:IEEE39bus,10machineNewEnglandtestsystem[39]
TheHVnetwork,Figure3.3,isa110kVsubtransmissionringimplementedaccordingto
[40].ItrepresentsatypicalsubtransmissionleveloftheGermannetworkconsistingofsix
buses. Every bus connects to a different MV network through two transformers with
25MVA of capacity each. Buses three and five are connected to the eHV network via the
threewinding transformers. Buses two and six havea WPP connected each, the ratings of
whicharegiveninTable3.2.TheanalyticaldataforthisnetworkcanbeseeninAppendixB.
Bus WPPRating
3 50 MW
5 50 MW
Table3.2:HVWindPowerParksRatings
23
to MV 110kV / 50MW to eHV to MV to MV
WPP (new)
No01
02 03 04
01 05
to eHV
to MV
06
to MV
110kV / 50MW
WPP (new)
to MV No02
Figure3.3:110kVSubtransmissionring[40]
The MV network used in this study was chosen after an extensive comparison of
different systems presented in literature [4147]. The Cigr medium voltage (MV)
distribution benchmark network is derived from a physical MV network in southern
Germany,whichsuppliesasmalltownandthesurroundingruralarea.Duetothefocusof
this thesis on the German situation and the previous use of this system in DG integration
studiestheCigrMVbenchmarkwaschosen,Figure3.4,aspresentedin[47].
Thenetworkconsistsof14nodeswitharatedvoltagelevelof20kV.Itisconnectedto
theHVnetworkthroughtwo110/20kVtransformersof25MVAcapacity.EachofthesixMV
network clones used in this study connect to a different busbar of the HV network. The
line modelling along with the load modelling and values of installed capacity were
implementedbasedonthe[47].ThedetailscanbefoundinAppendixC.
In each of the MV networks one wind power plant is connected at busbars 8 and 9
respectivelyandtwophotovoltaicpowerplantsareconnectedoneachofbusbars6and12.
TheratingsandconnectionpointsareshowninTable3.3.Thechoicefortheinstalledpower
oftheDGfollowedastatisticalanalysisofdatafromtheGermansituation[48].
24
Figure3.4:CigrMVNetwork[47]
SomeoftheMVloadsneedtobereplacedbyLVnetworksinordertobeabletoinclude
lowvoltageconnectedPVPPMandloads.Thechoiceofloadstobereplacedneedstoensure
validityfortheresultswithoutovercomplicatingthestructureofthesystem.BBs1and10
of the MV networks are connected to a LV network each, creating one connection point
close to the transformer and one deep in the MV voltage network. This strategy allows
mitigationoftheriskofbiasintheresults,whileensuringareasonablelevelofcomplexity
for the network topology.The connectionwas doneby replacing the loadson the specific
BBswiththeLVsystems.Scalingtookplace,asitwillbeexplainedinsection3.3.4,inorder
to maintain the power profile of the MV network as described in [47]. Details for the
modellingofthenetworkcanalsobefoundintheAppendixC.
25
3.3.4 CIGR low voltage benchmark system
3.3.4.1 Simplifications
The Cigr low voltage (LV) benchmark network is used in this study, Figure 3.5. The
networkconsistsofthreeradialfeedersrepresentingtheresidentialloads,theindustrialand
thecommercialones.Ontheresidentialfeeder,atjunctionnodesR1andR15,onePVPPMis
included.ThesizeofthePVPPMswasbasedonstatisticaldataoftheGermannetworkand
thevaluesarepresentedinTable3.4.Thenetworkismodelledbasedonthe[47].
Figure3.5:CigrLVbenchmarknetwork[47]
TheLVnetworkisclonedtwelvetimesanditisconnectedtobusbars1and10ofevery
oneofthesixMVnetworkasmentionedinthepreviousparagraph.InordertokeeptheMV
loadconstanttheresidentialpartofthelowvoltagenetworkneedstomatchtheresidential
load on the MV busbar and the same holds for the industrial/commercial part of the
network.
This procedure leads to the following results: clones of the low voltage network
connectedtobusbar1ofeveryMVnetworkincludeascalingoftheresidentialfeederby38
timesandofthecommercial/industrialfeederby15times.Thescalingprocedureincludes
26
the rating of the transformer, the installed load and the line current capacity. Also in the
residentialpartofthenetworktheoutputpowerofthePVPPMsisscaledaccordingly.For
the clones of the LV network connected to busbar 10 no scaling is necessary but the
commercialfeederisremovedinordertomatchtheindustrialMVload.Thedetailsofthe
network modelling can be found in the Appendix D along with the scaling manual for the
PVPPMmodel[49],AppendixG.
*XmarksthetypeofLVnetworkthatincludesPVPPM,inthiscaseonlyresidential.
The line modelling for the low voltage system was done in a detailed fashion. Two
differenttypesoflinesareusedbasedonthetypeofLVnetwork.Thecommercialnetwork
consistsofoverheadlinesandtheresidentialandindustrialnetworkofundergroundcables.
Themodellingforeachtypeispresentedbelow,theparametersandvaluescanbefoundin
AppendixD.
OverheadLines
For modelling power lines a number of modelling approaches are available in the
commercialmodellingsoftwareused(DigSILENTPowerFactoryv14.1).Themostsuitablefor
overhead lines is the socalled Typtow and Typcon elements [50]. The Typtow element
allowstheusertodefinethezerosequenceandpositivesequencereactanceandresistance
of the line as well as the number of line circuits that are included in the overhead
transmissionsystem.ItdoesnotgivetheoptionofincludinganeutralwirewhilsttheCigr
LV benchmark network data is given with a neutral wire configuration and is a 4wire
network.TosolvethisissuetheKronreducedimpedancematriceswereused,thattakeinto
account the neutral contribution on the phase impedance. Inside the Typtow element a
conductortypemustbedefined,theTypconelement.Theelectricalparametersneededfor
definingthiselementwereavailablefromtheCigrreport,allexceptthedcresistanceofthe
wireat20oCelsius.ThesevalueswereacquiredfromtheIEC60287[51].Anoverviewofthe
electricalparametersusedcanbefoundinAppendixD.
Undergroundcables
For the modelling of underground cables another element type was used. The TypLne
element is used which allows the modelling of the neutral wire and therefore the
parametersgivenintheprimitiveadmittancematrixwereused[47].Thecapacitanceofthe
cables was not considered due to the small lengths of the cables, as instructed in the
benchmarkreport[47].
27
3.4 Distributed Generation Modelling
As stated in section 2.4, the model used to represent the PVPPM originates from a
publicationofSamadi[36].ThisPVPPM,aspresentedinthesinglelinediagramofFigure3.6,
consists of the dc current source representing the PV array, the dclink capacitor and the
voltagessourceconverter(VSC).ItisathreephasesinglestagePVPPMconnectedthrougha
transformertothedistributiongrid.
Figure3.6:PVPPMcomponents[36]
The setup of the model is presented in Figure 3.7. The PV array seen consists of solar
cells which are, combined in series and in parallel to form PV modules, which in turn are
interfacedtoformsolarpanelsthatarealsoconnectedtoformthePVarray.Finallytheend
product,thePVarray,hasadequatevoltageandpowerforgridconnection.
Thedescriptionoftheoperationofthemodelfollows:ThedcoutputpowerofthePV
array is fed to the dc link, where a parallel connected capacitor is used to stabilize the
voltage. The voltage source converter (VSC) inverts the current and voltage waveforms to
connecttotheacside.TheVSCterminalsareconnectedtothepointofcommoncoupling
(PCC) via the interface reactor, and a stepup transformer. The transformer serves two
purposes: it provides an isolated ground for the PVPPM and it raises the voltage level to
matchthevoltageoftheacnetwork.Theshuntcapacitorontheacsideincombinationwith
the inductance of the transformer and interface reactor forms a filter that absorbs any
undesired current harmonics. The PLL algorithm is used to synchronize the control system
withthegridfrequencybyatransformationfromthea/b/creferenceframetoad/qframe
reference.
28
Figure3.7:PVPPMsystemsetup[36]
Themodelisequippedwithavarietyofcontrolfunctionalities.Theonesrelevantforthis
studywillbeexplainedinthischapter.ThedaxisandqaxisaredecoupledthroughthePLL
andthereforeactive(d)andreactivepower(q)controlareaswell.
The active power control strategy is done through controlling the dc link voltage. The
reference setpoint is calculated through the PV equation (Equation 21) and then is
perturbedbythemaximumpowerpointtracker(MPPT)tofindthePVoperationpointwith
maximumefficiency,thekneeoftheVIcurveofthePV.Theerrorsignalbetweentheset
point and the actual value is fed to a PI controller that calculates the daxis current. The
equationcanbeseenbelow.
1
I , K 1 U U Equation31[36]
Ts
Inthisstudythedynamicpowerfactorschemeisusedforsteadystatereactivepower
control,asproposedbyGermangridconnectionrequirements[10].Thecalculatedreactive
power reference setpoint is compared with the reactive power output and the deviation
signal is fed to a PI controller that calculates the qaxis current of the VSC. The reactive
power setpoint calculation is done based on the power factor. By measuring the active
poweroutputandapplyingthecharacteristicpresentedinFigure3.8thetargetpowerfactor
iscalculated.ByapplyingEquation32thereactivepowersetpointiscalculated.Therefore
when the PVPPM is exporting maximum active power then it is also consuming reactive
powerinordertolowerthevoltageatthePCC.
1 Equation32
Q P 1
cos
29
Figure3.8:Powerfactorrelativetoactivepoweroutput[36]
This section describes extensions of the PVPPM model necessary for a realistic
representationofvoltageridethrough(LVRT)behavior.Thestabilitysimulationsperformed
in the present study are based on rms (average value), positivesequence models because
the focus lies on the power system behaviour rather than specific issues of PV systems or
converterdesigns,thereforesomesimplificationswillalsobeincluded.
Figure3.9:Typicalvoltagermsvalueplotdividedintotimeperiodsofthenetworkfault
ThenetworkfaultresponseofPVPPMscanbedifferentiatedintothreetimeperiodsas
defined in Figure 3.9. Theprefault period spans from the start of the simulation until the
occurrenceofthefault.Inthistimeperiodthepreexistingprefaultcontrollerisusedand
allsystemvariablesareinsteadystate.Thefaultperiodstartswiththefaultoccurrenceand
endswiththefaultclearance;itischaracterizedbyheavytransients.Inthistimeframefour
differentcontrolmodesareaddedtothePVPPMasexplainedbelow.Thepostfaultperiod
followsthefaultperiodanddependingonthechoiceofcontrolmodecaneitherusethepre
30
fault controller or disconnect the PVPPM from the network . In addition during that last
periodanewcontrolstrategyisintroducedthatwillbeexplainedbelow.
InordertoshowtheoverviewofthecontrolfunctionalitiesavailableinthePVPPM,the
statediagramofFigure3.10ispresented.
Figure3.10:StateDiagramofPVPPMcontrol
Asdepictedinthediagram,theprefaultstateisthestartingpointofthesimulation.The
prefault controller function is described in 3.4.1.1. The second state of the system is the
fault state in which four different controlmodes are possible.If the controlmode is set to
NoLVRTthen avoltage drops below0.8p.u. willmove the system to the NoLVRT. For all
othercontrolsettingsthevoltageboundaryof0.9p.u.isusedforstatetransition.Whena
statetransitionisenergizedtheprefaultcontrollerisdecoupledfromthesystem.Thepre
faultcontrollermayagainbecoupledwiththesystempostfaultdependingonthechoiceof
control mode.The four control modes during the fault state are described in the following
paragraphs.
As described in section 2.2 the PVPPM installations in the low voltage network are
currentlynotabletoridethroughlowvoltages(nonLVRT).SoasseeninFigure3.11whena
faultinthenetworkresultsinthevoltageatthePCCofaPVinstallationtodropbelow0.8
p.u. then the under voltage protection is triggered and with a maximum timedelay of0.2
seconds the plant is disconnected from the network. Furthermore after voltage recovery,
within60secondstheplantisreconnectedtothenetwork[10].Thiscontrolfunctionalityis
available in some of the PVPPM used for this study and represents the behaviour of PV
installationscommissioneduntil2012.
31
Figure3.11:NoLVRTrequirementfortypeAgenerators,powergeneratingmoduleswithaconnection
pointbelow110kVandmaximumCapacityof0.8kWorless[13]byForumnetworktechnology/network
operation(VDE)[10]
The control setting chosen for modelling the undervoltage protection sets the voltage
boundary at the PCC at 0.8p.u. resulting in termination of active and reactive power
exchange with the network for 60 seconds, should the voltage drop below this boundary.
Thetimedelayfortheprotectionrelayresponseis0.1s.Thisvalueischosenastofullfillthe
requirement of disconnection within 0.2s without being borderline compliant. Figure 3.12
demonstrates this operation for type A generators, which are power generating modules
withaconnectionpointbelow110kVandmaximumCapacityof0.8kWorless[13].
1,1
0,6
0,0
-0,6
-1,1
-0,1 0,0 0,1 0,2 0,3 [s] 0,4
Controller: Id_ref
Controller: Iq_ref
Figure3.12:DaxisandqaxiscurrentsofaPVPowerPlantsetforNoLVRTduringfault
Source:ownsimulation
32
ThedaxisandqaxiscurrentsofaPVPPMaredepictedduringafault.Thefaultoccursat
t 0 and within 0.1 s the PVPPM ceases power exchange with the network. The rise in
active current during the first 100ms of the fault is due to the response of the prefault
controller trying to balance the ac/dc power imbalance at the terminals of the inverter
causedbythevoltagedropattheacside.Thecorrespondingdropinreactivecurrentisalso
theresponseoftheprefaultcontrollerapplyingthedynamicpowerfactorcontrolstrategy.
Thezeropowermode(ZPM)isafaultridethrough(FRT)modeduringwhichthepower
generating module does not infeed any power (active or reactive) into the network. This
control mode is desired by many distribution system operators (DSO) to avoid conflicts of
DGwithnetworkprotection[5].Thereforewhenthevoltagedropsbelow0.9p.u.atthePCC
then the VSC of the PVPPM stops any power exchange with the network while staying
connected.Thedifferencewiththestatusquoisthatbystayingconnectedwiththenetwork
the PLL synchronization algorithm is applied continuously and as soon as the voltage is
restored, the active power of the PVPPM recovers instantaneously and therefore is not
disturbingtheactivepowerbalancepostfaultbystayingdisconnected.
In that way the ZPM is providing service to the network postfault by reconnecting
faster and taking part in the restoration of the power balance. This control strategy is
implemented during the fault by blocking the inverter switches and directing the power
productionfromthePVPPMtothedcchopper.Bydoingso,thepowerisdissipatedoverthe
series dump resistance of the chopper. It can, therefore, be concluded that this control
modeisimplementedmainlyduringfaultbuttheservicetothegridismanifestedduringthe
postfaultbehaviourofthesystem.
1,1
0,6
0,0
-0,6
-1,1
-0,1 0,0 0,1 0,2 0,3 [s] 0,4
Controller: Id_ref
Controller: Iq_ref
Figure3.13:DaxisandqaxiscurrentsofaPVPowerPlantsetforZPMduringfault
33
In Figure 3.13 the daxis and qaxis currents of the PVPPMs inverter are shown. The
fault occurs at t 0 and the controller is driving the power exchange/currents to zero.
Afterfaultclearance,150mslater,theprefaultcontrolleriscoupledwiththenetworkand
thecurrentsreturntotheprefaultsetpoints.Nodelayisimplementedintherecoveryof
activepowerinordernottointeractwiththeprefaultcontroller.Thevoltage,notshown,is
recovering immediately and enables the fast response of the controller. The difference
between the NoLVRT and ZPM is evident by comparing Figure 3.12 and Figure 3.13. The
zeropowerexchangeisachievedinlesstimeasthecontrollerisfasterthantheprotection
relay.Alsotheripplesseenatfaultrestorationareduetothevoltageoscillationpostfault
thattriggerstheactionofthesteadystatecontroller.
Anotherrequirementforapplyingdynamicnetworksupportduringfault,asdefinedby
the German grid connection requirements [5], is the additional injection of current during
fault to assist in raising the voltage at thePCC. Usually this additional current depends on
thetypeofnetworktheDGisconnectedto.Thereforewhenmediumvoltageorhighvoltage
networksareconsideredthecurrentispurelyreactiveasthenetworksX/Rratioishighand
the strongest impact in raising the voltage can be achieved by injecting an additional
reactivecurrentcomponent.
PriortothefaulttheinverterofthePVPPMisinjectingactiveandreactivepowerinto
the grid which corresponds to the active and reactive current. The vector sum of those
currentsformsthetotalcurrentinjectedintothegrid.Duringfaultitisdesirabletomaintain
theprefaultoperatingpointandinjectanadditionalreactivecurrentcomponent.
Assoonasthefaultisdetected,theinputstotheprefaultcontrollerarefrozentothe
prefaultvalues.ThenanadditionalproportionalcontrollerisaddedtotheoutputofthePI
qaxis current controller. The additional controller uses the deviation of the ac voltage to
raise the qaxis current. The equations for the reactive and active current are shown in
Equation33andEquation34,respectively.ThegainparameterK isadjustable.Thetotal
currentisthenpassedthroughalimitertoensurethatthecurrentcapacityoftheinverteris
notexceeded.Forthelimiteravalueof1.1p.u.isusedforthesumofthecurrents.Incase
thesetpointsareexceedingthecapacitythenthedaxiscurrentisscaleddowninorderto
remain within the inverter capacity. The scaling function of the limiter can be seen in
Equation34.dV isthedeviationofthevoltagefromthenominalvalueinp.u.wheredbis
thedeadbandaroundthevoltagedeviation.
| |
| | , Equation33
,
,
Equation34
,
34
Afterfaultclearancethestateofthecontrollerisunchangeduntilthevoltagerecovers
to0.9p.u.atthePCC.Afterthatpointtheadditionalcontrollerisdisabledandtheprefault
controllertakesover.Asmalladjustmentfollows,beforetheprefaultcontrollerreturnsto
thepointofoperationpriortothefault.Figure3.14depictstheprocessdescribed.
1,1
0,6
0,0
-0,6
-1,1
-0,1 0,0 0,1 0,2 0,3 [s] 0,4
Controller: Id_ref
Controller: Iq_ref
Figure3.14:DaxisandqaxiscurrentsofaPVPPMsetforaRCIduringfault
The final control functionality implemented for the fault period is the mixed aR&ACI.
ThiscontroloptionissimilartotheaRCI,whiletheadditionalcomponentofthecurrentalso
hasanactivepart.ThereasonforthisadditionstemsfromtheX/Rratioofthelowvoltage
networkthatcanbelowerthanoneunliketheMVorHVguide.Thistranslatesintoactive
currenthavingadominantimpactonthevoltagesupport.Theadditionalcurrentisinjected
atanadjustableanglebetweenactiveandreactivecurrent.
Theprefaultcontrollervaluesofthedaxisandqaxiscurrentsarefrozenasdescribed
in 3.4.1.2. During fault a proportional controller is added for both active and reactive
current. The reactive and active current equations are Equation 35 and Equation 36,
respectively.Theadditioninbothequationsisthecosineandsinusterms,whichtakecareof
theangleofinjectionifthesameKfactorisused(K K .
| |
&
| | sin Equation35
| |
&
| | cos Equation36
35
1, & &
& , &
&
Equation37
_ & _ &
TheequationforthescalingcanbeseeninEquation37.Thegraphicalexplanationfor
thescalingprocedureispresentedinFigure3.15.
Figure3.15:Inverterdaxisandqaxiscurrentdiagramsforpresentingscalingscheme
Theinitialsetpointforbothactiveandreactivecurrentismarkedinthegraphwiththe
blue arrow, from that starting point there is an additional directional current component
addedtothetotalcurrentthatstemsfromthefaultproportionalcontrollers(orangearrow).
Ifthetotalcurrentoftheinverterexceedstheinverterscapacity,scalingistakingplace.The
magnitudeofthetotalcurrentisinterceptedwiththecorrespondingstraightlinefromthe
origin of the two axes. Moving on that line the setpoint is shifted until it is inside the
capability limit of the inverter. With this scaling technique the current setpoint is shifted
during the fault period until the angle between they axis and the straight line connecting
the origin and the current setpoint can reach the maximum of the additional current
injectingangle,whilststayingwithinthecapabilitylimitsoftheinverter.
ThesameprocedurecanbeseeninFigure3.16forthesameangle(45)aspresentedin
Figure3.15.Thefaultoccursatt 0 andiscleared150mslater.Thevoltagedropatthe
PCC determines if the maximum angle will be reached. In this case the current is not
36
exceeding the limit of the inverter (1.1p.u.) and therefore the final angle differs from the
injectionangle.
1,1
0,6
0,0
-0,6
-1,1
-0,1 0,0 0,1 0,2 0,3 [s] 0,4
Controller: Id_ref
Controller: Iq_ref
Figure3.16:DaxisandqaxiscurrentsofaPVPPMsetforaR&ACIduringfault
The final control strategy added to the model is the postfault active power recovery.
The restoration of the active power postfault value at the end of the faultperiod to the
prefaultactivepowersetpointisachievedthrougharampwithaspecificinclination.The
Germangridconnectionrequirementsaskforarestorationofactivepowertotheprefault
statewitha20%increaseofactivepowereveryseconduntiltheprefaultactivepoweris
reached[28].Thiscontrolstrategyisimplementedwithafilterfunction.Theresponseofthe
controllerisalmostimmediateinprovidingtheprefaultsetpoint.Thisresponseisfiltered
through the tracker by a first order filter function with the required time delay to achieve
therampfeaturesrequired.Therampisactivatedpostfaultanddeactivatedwhenthepre
fault active power is reached. The tracker equations can be seen below (Equation 38,
Equation39).
0.2, _
, 1
_ , _
Equation38
_ , 1
, 1
_ Equation39
_ , 1
The format that the equations are presented is the state space, and the final output
comes from Equation 39. The condition is a boolean term which, if true,
indicatesthatthesystemiswithinthepostfaultperiod.Thisconditionistruefor5seconds
37
after the fault in order to ensure that the daxis current will be restored to the prefault
value even if we are in the worst case (5 seconds needed for 01 pu, with 20% raise per
second).Thevalue0.2ischosenasthespecifiedrateofincrease.TheKvaluesforthefast
and slow trackers are 104 and 103 respectively. The results are shown in Figure 3.17 as
implemented by the PVPPM modelled. The fault is cleared at 2.15s and then the delay in
activepowerrestorationaffectsthereactivepoweraswellasitisafunctionofactivepower
(Equation32).
Figure3.17:DaxisandqaxiscurrentinaRACIwithdelayedactivepowerrecovery
DCChopper
[32]reportsundesireddisconnectionofPVPPMsduringorafternetworkfaultstriggered
bytheinternaldccircuitprotection,evenifthePVPPMsweresupposedtoperformLVRT,as
follows.ThedclinkvoltageinitiallyrisesduringafaultontheacsideofthePVPPMdueto
theactivepowerimbalancebetweentheacanddcsideofthevoltagesourceconverter.The
controller responds by increasing the ac current and thereby tries to restore the active
powerbalance.However,thecurrentlimitsoftheinvertermaybereachedbeforeanactive
power balance can be restored and that ultimately leads to a further increase of the dc
voltage. This behaviour is reversed in the postfault period. The ac voltage is recovered
almostinstantlyatfaultclearancewithoutgivingthecontrollersufficienttimetoreducethe
acsidecurrenttoitsprefaultlevel.Theresultisanactivepowerimbalancebetweentheac
and dc side of the inverter. Since the active power flow through the grid sideconverter is
larger than the one from the dc source, the dc capacitor is discharged, i.e. the dc voltage
decreases.Incertaincasesthedcvoltagemaydropbelowtheminimumallowedvalueand
that would trigger the internal dc circuit protection to disconnect the PVPPM from the
network.
38
Discussions between TU Delft with Laudahn (TU Braunschweig) indicated that such
issuesareunlikelytooccurforPVPPMinverterssoldontheGermanmarket[37].Proprietary
control strategies of the grid side converter and/or dcdc converter would take care of
levelling the dc voltage within the allowable voltage band and prevent disconnection of
PVPPMsduringorafternetworkfaultsduetotheinternaldccircuitprotection.
Theactivepoweriscontrolledbyadjustingthedclinkvoltageandthereactivepoweris
calculatedviathedynamicpowerfactorapproach,asdescribedinsection3.4.1.1.Thusitis
obviousthatthedclinkvoltageisanimportantparameterforcontrollingthePVbehaviour.
The importance of the dc voltage is further stressed by the constraints imposed by the
protection of the PVPPM. If the dc voltage exceeds certain limits the protection relay will
disconnect the PVPPM from the grid. Therefore in order to be able to safely apply any
controlschemeduringfaultwemustensurethatthedclinkvoltagewillstaywithincertain
limits.
Giventhepowersystemfocusofthepresentstudyandthelackofinformationregarding
different proprietary control strategies for PVPPMs, simplifcations in the modelling of the
PVPPMsarerequiredandjustified.Originatingfromthedetailedrepresentationofboththe
dcandacsidesofaPVPPM[36]andaimingatastableoperationduringandafternetwork
faultswhiledisregardinganyspecificandpotentiallyproprietarycontrolstrategies,anactive
dcchopperinthedccircuitisassumedduringLVRToperationandforacertaintimeafter
voltagerecovery.Thechoppercomprisesoftwoantiparalleldiodesconnectedtothedcbus
and in series to a voltage source behind an impedance (see Figure 3.6). For steadystate
operationnochopperisassumed.Thechopperkeepsthedcvoltageattheprefaultvalue
during the fault. For an adjustable time after fault clearance it is assumed to continue
operationwhichfurtherpreventsundesireddcvoltagefluctuationspostfault.
AnotheradditiontothePVPPMmodelisthecompliancewiththereactivecurrenttime
performance as required by German grid codes and presented in Figure 3.18. The step
responsedefinedbyGermangridcodesisthemarginallyacceptedbehaviourofagenerator
performingacontrolinducedchangeinitscurrent/powersetpoint.ThereforeeveryPVPPM
thathastoconnecttotheGermangridmustprovethatthestepresponsetoachangeofthe
current setpoint can be performed within the margins specified in Figure 3.18. In detail,
90% of the final setpoint value must be achieved within 30ms (rise time) and the final
signalmustbesettledbetween90%and120%ofthefinalvaluewithin60ms(settlingtime)
[52].
39
Figure3.18:Germangridstepresponse[52]
The PVPPM model was already complying with these requirements. However it is
unrealistictoconsiderthatalltheinstalledPVPPMshaveaveryfastresponseastheywere
installed in the past and therefore it could be that they do not comply to these
requirements.Inordertoensurethatthemodelisrealisticanditconsiderstheworstcase
ratherthanthebestafilterwasimplementedsothatthestepresponserequirementismet
marginally.Fortheimplementationofthefilterafirstorderfunctionwasusedasitisshown
in Equation 310 and Equation 311. The parameter tuning was done with a heuristic
approachinordertoachievethedelaysof30msand60msforeachsetpointpercentage.
The variable is a boolean which is true only during the fault period. The diagrams
presentedforallthecontrolmodesincludethisfilter.
Equation310
, 1 Equation
_
, 1 311
The WPPs models used in this study originate from [40]. The models for WPPs used
consistof:
Doublyfedasynchronousgenerators(DFAG),and
Fullconvertergenerators(FCG).
TheWPPconnectedineachofthesixMVnetworksconsistofbothtypeofgenerators.
The 10MW WPP consists of a 6MW aggregated FCG generator and a 4MW aggregated
DFAGgenerator.Thesmaller6MWWPPconsistsofa4MWaggregatedFCGgeneratorand
a2MWaggregatedDFAGgenerator.
40
AllWPPsareabletoridethroughfaultsandinjectreactivecurrenttothenetworkinthe
eventof a fault. The type of fault control applied to these WPPs is the additional reactive
current injection (aRCI) control mode similar to thePVPPMs. The gain is set at 2 and this
behaviour is in compliance with current grid codes for MV and HV connected DG [5],[28].
ThemodellingdetailsoftheWPPswillnotbediscussedastheyarenotinthefocusofthis
thesis but rather support the realistic representation of the German network. However
detailsinthemodeofoperationofthesemodelscanbefoundin[53]fortheDFAGmodel
and in [54] forthe FCG model as they are both based onthe corresponding PowerFactory
templates.
Through an extensive literature survey on load modelling, it became clear that the
model structure of the load and the identification of the load parameters is an important
factor in stability studies [55], [5662].Different load modelstructures maymake stability
studyresultstotallydifferent[55].
A load can be modelled as either static, dynamic or aggregated. A static load model
describestherelationoftheactiveandreactivepoweratanytimewiththevoltageand/or
frequencyatthesameinstantoftime[56].Astaticloadcanhaveeitherapolynomial(ZIP)or
exponential form. On the other hand, a dynamic load model expresses this relation as a
function of the voltage and/or frequency time history, including normally the present
moment[56].Dynamicloadmodelscanberepresentedbyinductionmotormodelsorinput
outputmodelsthatarebasedondifferentialequations[55].Theaggregatedloadcombines
bothmodellingtypesbyapercentagesplitbetweendynamicandstaticload.
PowerFactory provides a variety of options to model loads. For the static load model
both polynomial and exponential load modelling is possible. Also the model can be part
dynamicandpartstaticwherethedynamicpartcanbebasedonlinearornonlinearsecond
orderdifferentialequations.Finallythereisalwaystheoptiontouseaninductionmotorin
order to represent dynamic load [62]. However the parameter identification for load
modellingisadifficultandtimeconsumingprocedurethatstemsfromstatisticalanalysisof
historical data combined with optimization algorithms for curve fitting. The choice of
parameters for this study is done based on the work of [55, 63] and [64]. The frequency
dependentpartoftheloadisignored.
Forthestaticpartoftheloadthepolynomial(ZIP)approachispresentedfirst.Inthis
approachtheloadcanbeclassifiedintoconstantpower,constantcurrentandconstant
impedanceload,dependingonthepowerrelationtothevoltage[56].Asitcanbeseenin
Equation312andEquation313forconstantimpedanceload,thepowerdependenceon
voltageisquadratic,foraconstantcurrentitislinear,andforaconstantpower,thepoweris
independentofchangesinvoltage.
1 Equation312
41
1 Equation313
For the exponential approach Equation 314 and Equation 315 give the active and
reactive power dependency of the load on voltage as found in [65]. For this thesis the
exponential approach was chosen as parameters for exponentially modelled loads for
differentvoltagelevelsandtypeofnetworkswereeasiertobefound.
_
Equation314
_
Equation315
The values for load parameter chosen for each type of network are presented in
Table3.6.TheeHVstaticloadparameterspresentedaretakenfrom[64]andfortherestof
theloadsfrom[63].
ForthedynamicpartoftheloadthenonlineardifferentialequationmodelofDIgSILENT
PowerFactoryisavailable(Equation316,Equation317)[62].Asitcanbeseenthevoltage
deviationfrom1p.u.playsaroleintheinstalledpowerwhichistimedependent.However
inthisstudynoconsiderationofdynamicloadwilltakeplaceasitwouldfurthercomplicate
thestudyandparametersfortimeconstantsaredifficulttofind.
_ _
1 % Equation316
1
1 % Equation317
1
3.6 Conclusions
Concludingthemodellingchapter,anoverviewofthestudynetworkstructurewasgiven
andeveryvoltagelevelwasexplainedinmoredetail.Thecomponentsincludedinthestudy
networkwereidentifiedandthemodellingprocedurewasdescribed.ThePVPPMmodelwas
presented in detail along with all the control functionalities that were included for the
purpose of the simulations to be performed in Chapter 4. The WPP models were briefly
mentionedaswellasloadmodellingandchoicesmade.
42
Chapter 4 : Case study and Sensitivity Analysis
4.1 Introduction
Inthischapterthestructureofthecasestudiesisexplainedindetail.Theprinciplesof
operation of the study system are defined and a general outlook of the study is given.
Moreoverthechoiceofsensitivityfactorsthatformulatethecasestudiesisexplainedand
justifiedandfinallyselectedresultsarepresentedandanalysed.Theresultspresentedare
themostimportantasaselectionhadtobemadeduetothevolumeofcasesanalysed.
The test system described in Chapter 3 is used to simulate transmission system faults
andinvestigatethesystemperformanceunderdifferentconditions.Theparametersthatare
notvariedunderdifferentcasestudiesarepresentedinthissection.
Thelocationofthefaultisplacedatthemiddleofline9oftheeHVnetwork(seeFigure
3.2).Thechoiceofthefaultlocationismadeinordertoensurethatthefaultisclosetothe
busbar that connects the eHV network with the subtransmission and distribution system
alsoduetothelowshortcircuitratioofthatlocationtoensureaseverevoltagedrop.The
faultdurationischosentobe150msinaccordancewiththefindingsoftheliteraturesurvey
presentedinsection2.5.Thefaultoccursat0sandiscleared150mslater.Thetypeoffault
that will be analysed is a threephase fault with no fault impedance. This type of fault is
chosen inorderto investigate the most severe case that would bring thesystemcloser to
instability.ThedetailsofthefaultlocationanddurationarepresentedinTable4.1.
FaultLocation Middleofline9ofeHVnetwork
Faultduration 150ms(0150 ms)
FaultImpedance 0 Ohm
Distancefrombusbar8 2.5km
Table4.1:Detailsoffaultevent
Furthermore due to the use of different load and generation profiles for the starting
pointofthesimulation,thepositionsofthetransformertapchangersneedtobedefined.In
order to ensure a proper initial voltage for all busbars of the network in all load and
generationcasesthefollowingstrategywasdevised.
The eHV/HV transformers are equipped with automatic tapchanging and regulate the
secondarysidebusbar,inthiscasethebusbaroftheHVnetworktheyareconnectedto,ata
targetvoltageof1.0p.u.TheHV/MVtransformersuseremotevoltagecontrol,andregulate
busbarnumber10oftheMVnetwork(seeFigure3.4),atatargetvoltageof1.0p.u.Finally
theMV/LVtransformersarenotequippedwithautomatictapchangersbuthavemanualtap
changers. Therefore for the starting point of the simulation the residential LV network
43
transformersfixtheirtapchangersinordertoensurearemotetargetvoltageof1.0p.u.in
terminalR15oftheLVnetwork(seeFigure3.5).ThetransformersoftheLVindustrialand
commercialnetworksregulatetheLVconnectedbusbarwithatargetvoltageat1.0p.u.All
transformers equipped with automatic tap changers have a time constant of 20s for
changing their taps as a response to a change of voltage at their respective control node.
TheschemefortransformertapchangerscanbeseeninTable4.2.
With regard to the behaviour of DG installations and in order to respect the already
defined German grid codes, the MV connected PVPPMs and the respective HV and MV
connectedWPPsareequippedwithFRTcapabilityandadditionalreactivecurrentinjection
during fault. In this way a more realistic approach equivalent to todays behaviour of the
Germansystemismodelled.AlsointheLVnetworkpart,theoldPVPPMsinstallationsare
notabletoperformFRTanddisconnectintheeventofafaulttomodelthealreadyexisting
PVPPMsthatareconnectedupto2012andarenotabletoridethroughafault.
PVPPMsconnectedtotheLVnetworkafter2012andupto2022aregiventheoptionof
FRT and current injection during a fault. Those are the PVPPMs that will change their
respectivecontrolinthecontrolmodevariationsthatwillbeexplainedinsection4.3.2.
Finally in the event of a transmission system fault the shortage of active and reactive
powerofthenetworkiscoveredbygenerator2oftheeHVnetworkwhichismodelledasa
slackgenerator(seeFigure3.2).
Thisstudyisperformedasapossiblescenariofortheyear2022.Thereforethesystemis
modelledfortheyear2022but2012DGinstallationsarealsoconsidered.Intheyear2012
the system already includes the DG connected today that are compliant with current grid
codes.Intheyear2022anadditionalprojectedinstalledcapacityofPVPPMsisaddedinthe
LVandMVnetworks.Thedetailsforthenetworkateachyearareexplainedbelow.
Inthe2012GermanscenariothetotalinstalledcapacityofPVPPMsperMVnetwork
was7640kW.AsitcanbeseeninTable3.3thesizesofPVinstallationsfortheruraland
suburban part of the MV network are 120kW and 3700kW respectively. So in order to
matchthecapacitieswiththetotalsizeoftheinstalledPVPPMs,twoofeachsizeofPVPPMs
where connected in their respective network. For simplification of the study an aggregate
model of 240kWwas created forthe suburban partof theMV network and an aggregate
modelof7400kWfortheruralpartoftheMVnetwork.Theconnectionpointsofthetwo
44
PVPPMsarebusbar6forthesuburbanpartoftheMVnetworkandbusbar12fortherural
part.ThesePVPPMsareabletoridethroughafaultandprovidereactivecurrentwithagain
factorof2asdemandedbyGermangridcodes[5].FortheMVandHVconnectedWPPsthe
ratingsandconnectionpointsarekeptthesameasdescribedinTable3.2andTable3.3.The
sameFRTandreactivecurrentinjectioncontrolsareincludedfortheWPPsaswell.
FortheLVconnectedPVPPMsthetotalinstalledcapacityperLVnetworkwas180kW.In
2012thesizesoftheLVconnectedPVPPMscanbeseeninTable3.4.Inordertomatchthe
capacitieswiththesizeofinstalledPVPPMs,12installationsof10kWand12installationsof
5kWarechosenperLVnetwork.TheaggregatingprocedureoutcomeisaPVPPMof120kW
of installed power connected in busbar R1 of the LV network and a PVPPM of 60kW
installedpowerconnectedinnodeR15oftheLVnetwork(seeFigure3.5).
FurthermoreLVconnectedPVPPMsforthe2012scenarioabidesbycurrentGermangrid
codes[10]thataskformandatorydisconnectionintheeventofafault.Thedisconnection
timeischosentobe100msaftertheoccurrenceofthefaultinordertorepresenttheworst
case.
Inthe2022scenariotheinstalledsynchronousgeneration,WPPsandloadarekeptthe
sameasin2012.TheinstalledcapacityofPVPPMsintheMVandLVnetworksisdoubledin
comparison to the 2012 scenario presented in Table 4.4. So all busbars and nodes of the
45
networkthatwereconnectedtoaPVPPMofacertainratinginthe2012scenario,areinthe
2022scenarioconnectedtotwoPVPPMsofthesamerating.ThenewlyinstalledPVPPMsare
abletoperformdifferentcontrolstrategieswhilsttheonesconnectedin2012cannotride
through fault. Doubling the installed PV capacity in both the MV and LV network over 10
years is a realistic assumption as it can be seen in the literature survey chapter for the
projectedgrowthofPVPPMsinstalledcapacityinGermany.
In Figure 4.1 the installed load and DG production can be seen for all the MV and LV
networks in the 2022 scenario. Due to having an aggregated and a single LV network
connectedtodifferentbusbarsoftheMVnetworktheirinstalledloadsandproductionsare
presented separately. The load and production of the eHV and HV networks are shown in
Table4.5.TheHVnetworkhasnoload.
Figure4.1:InstalledloadandDGgenerationinthe2022ScenariofortheHV/MV/LVnetworks
The first sensitivity factor presented is the load and generation profile. Namely as
explained before the system has a fixed installed generation and load power. However in
differentperiodsoftimeonlyapartoftheloadisactuallyconnectedtothenetworkdueto
thetimevaryingneedsofconsumers.AlsorenewablebasedDGpoweroutputisdependent
on solar irradiation for PVPPMs and wind speed for WPPs. Therefore only a part of the
installed power of DG is actually fed to the network at different times. Synchronous
generationalsotakesintoaccounttheneedsofthenetworkintermsofactiveandreactive
powerandproducestherestofthepowerneeded.
46
The aim is to define three prefault operating points for the network, based on load
consumptionandPVPPMsproduction.Theloadcanbeeitherat1p.u.wheninthehighload
caseorat0.5p.u.atthelowloadcaseasseeninTable4.6.Similarlytheirradiationisatthe
corresponding value to produce 1p.u. active power output from the PVPPMs in the high
generation case and is halved so that PVPPMs produce 0.5p.u. active power in the low
generationcase(Table4.6).WPPproductionisconsideredthesameforallcasesandisfixed
at 1p.u. for all WPPs. The three prefault operating points of the network are created by
varyingtheloadandgenerationcases.Eachofthemcreatesadifferentpowerflowasitwill
beexplainedinthenextthreesections.
Byusingafactorof1forinstalledload(highload)and0.5forinstalledPVPPMspower
output (low generation) the load and distributed generation profile seen in Figure 4.2 is
realised.AsitcanbeseentheDGintheMVandLVnetworksarenotabletocovertheload
demand within their respective voltage levels and the 100MW of power that the HV
networkcanprovidealsodoesnotcoverthetotalloaddemand.Hencethepowerremaining
will be provided by the eHV network. This load/generation profile will therefore create a
verticalpowerflowinsidethestudynetworkstemmingfromtheeHVnetworkdowntoallLV
networks.Sointhiscasethesystemsprefaultconditionwillbeaunidirectionalpowerflow
asseeninFigure4.3.
Figure4.2:LoadandDGgenerationforHV/MV/LVnetworksforBasecasepowerflow
47
Figure4.3:UnidirectionalpowerflowfromeHVtoLVnetworkatBaseCase
Byusingafactorof0.5p.u.forpeakload(lowload)and0.5forinstalledPVPPMspower
output (low generation) the load and distributed generation profile seen in Figure 4.4 is
realised.
Figure4.4:LoadandDGgenerationforHV/MV/LVnetworksforHVandMVreversepowerflow
As it can be seen the DG in the LV networks are not able to cover the load demand
withintheirrespectivevoltagelevels,contrarytotheMVnetworksthatevenhaveasurplus
ofpower.The49MWofexcesspowerintheMVnetworksareenoughtocoverthealmost
18MW of shortage in the LV networks and export power towards the HV level. The HV
network having no installed load will export active power towards the eHV network. This
load/generation profile will therefore create a split reverse power flow inside the study
network, resulting in reverse power flow from the MV and HV voltage levels as seen in
Figure4.5.
48
Figure4.5:SplitreversepowerflowfromMVandHVtoeHVnetwork
By using a factor of 0.5p.u. for peak load (low load) and 1p.u. for installed PVPPMs
power output (high generation) the load and distributed generation profile seen in Figure
4.6isrealised.AsitcanbeseentheDGintheLVandMVnetworksareabletocoverthe
loaddemandwithintheirrespectivevoltagelevelsandexporttheexcessofactivepowerto
theHVnetwork.TheHVnetworkhavingnoinstalledloadwillexportactivepowertowards
theeHVnetwork.Thisload/generationprofilewillthereforecreateafullreversepowerflow
inside the study network, resulting in a reverse power flow from the LV, MV and HV
networkstowardstheeHVnetwork,asseeninFigure4.7.
Figure4.6:LoadandDGgenerationforHV/MV/LVnetworksforLVtoeHVreversepowerflow
Figure4.7:FullreversepowerflowfromLVtoeHVnetwork
49
4.3.2 Network fault control mode settings
The PVPPMs network fault control settings that are explained in section 3.4.1.2 are
applicable only to the low voltage connected PVPPMs installed after 2012. The sensitivity
towards the behaviour of PVPPMs during and after the fault is applied by changing the
controlstrategyfollowedbytheaforementionedPVPPMsbetweennoLVRT(noLowVoltage
Ride Through), ZPM (Zero Power Mode), aRCI (additional Reactive Current Injection) and
aRACI(additionalReactiveandActiveCurrentInjection).
Alsotheoptionofdelayedpostfaultactivepowerrecoverycanbeimplementedinany
of the three control modes except for the noLVRT. In total this allows for seven different
controlmodesthatwillaidinunderstandingthesensitivityofthevoltageintheLVnetwork
based on the behaviour of the PVPPMs during and shortly after fault and the benefit of
applyingdynamicvoltagesupportattheLVlevelthroughtheLVconnectedDG.
ForthecaseofaRACIthechoiceofangleforinjectingthemixtureofactiveandreactive
currentduringthefaultisbasedonthenetworkimpedanceangleseenfromthePCC(point
ofcommoncoupling)ofeachLVconnectedPVPPM.Thepossiblefourdifferentlocationsof
the connection point of PVPPMs in the low voltage network are shown in Table 4.7 with
theirrespectiveangleforperformingaRACI.
MVbusbar/LVbusbarofconnectedPVPPM aRACIangle
01/R1 77
01/R15 26
10/R1 73
10/R15 15
Table4.7:ConnectionpointsofPVPPMsinthenetworkandangleofaRACIchosenforeachpoint
Table4.8:ThestudycasesformedbyvaryingthesensitivityfactorsofloadcasesandPVPPM'scontrolmodes
50
Thefinaloverviewofsensitivitiesthatstemsfromthevariationoftheprefaultnetwork
setpoint (load case) and the variation of the control strategy followed by PVPPMs
connectedintheLVnetworkafter2012formsthefinal21studycases(Table4.8).
Due to the high number of study cases only one load case will be presented fully and
then comparative graphs will be shown between different control modes and power flow
scenarios.TheresultsforthestudycasesnotpresentedhereareincludedintheAppendixF.
Also for the cases with delayed postfault active power recovery only an overview of the
effectofthiscontrolstrategywillbepresentedthroughselectedgraphs.
Four cases will be presented in this section. The structure of presenting and analysing
theresultsiskeptthesamethroughoutthecases.Thevoltagesatdifferentbusbarsofthe
lowvoltagenetworkarepresentedandthenwiththeaidofsupportinggraphstheanalysis
of the voltage pattern takes place. Firstly the noLVRT control scenario for the base case,
normalloadflow,willbepresentedasitshowstheprojectedstateofthenetworkifGerman
grid codes stay unchanged until 2022. Following this case the comparative results for all
control scenarios, besides the ones with dAPR, in the base case will be presented to
demonstrate theeffectofdifferent controlmodes on thevoltage profile. Thethird case is
goingtobeacomparativecasebetweenallcontrolmodesfortheLVtoeHVreversepower
flowscenario.Finallytheeffectofdelayedactivepowerrecoverywillbepresented.
In the base case the simulation starts from a prefault state of topdown power flow
from the eHV to the LV network. This case represents the most commonly experienced
power flow state of the network nowadays. The parameters that usually ensure a vertical
powerflowareahighloaddemandandalowactivepowerinfeedfromDGconnectedto
thedistributionnetwork.
Inthisscenarioathreephasefaultistriggeredat0sinline9oftheeHVnetworkand
150mslaterthefaultiscleared.HVandMVconnectedDGabidewiththecurrentgridcode
requirements (GCR) and inject reactive current during the fault to support the voltage. LV
connected PVPPMs abide with the LV GCR and disconnect from the network 100ms after
thefaultisdetectedviathevoltagedropatthePCC.AlsointhiscaseMVandLVconnected
PVPPMsareproducingathalfofratedactivepoweroutputduetotheassumedirradiation
atthetimeofthefault.Duringthisperiodof100mstheprefaultcontrollerofthePVPPMs
isoperational.
InFigure4.8thevoltageprofilesofallbusbarsoftheeHVnetworkareseen.Thefault
causes different voltage dips at different locations in the eHV network depending on the
distance from the fault and the shortcircuit ratio of each location. As it can be seen the
rangeofvoltagedropvariesfrom0.1p.u.toalmost0.8p.u.
51
Figure4.8:Voltageinp.u.atallbusbarsoftheeHVnetwork(differentcoloursrepresentdifferentbusbars)
Thefaultatbusbar8oftheeHVnetwork,wherethedistributionsystemisconnected,is
theredlineofthegraphandthevoltagedipatthatpointreachesdownto0.4p.u.Dueto
thereactivecurrentinjectionduringfaultthereisariseinthevoltageinthatperiodbutitis
not significant. Then the disconnection of the PVPPMs at 100ms causes a slight voltage
decrease, and at 150ms the fault is cleared and the voltage is restored at a slightly lower
steadystatevaluethantheprefault.
The oscillation of the voltage is caused by the oscillation in the infeed of active and
reactivepowerbythesynchronousgeneratorsoftheeHVnetwork.Thisbehaviouristypical
ofSGafterfaults.
For the LV connected PVPPMs there are four different positions in which they can
connect.EveryoneofthesixMVnetworksofthetestsystemisconnectedto2LVnetworks;
oneatbusbar1andtheotheratbusbar10oftheMVnetwork(seeFigure3.4).Eachofthe
twoLVnetworkshastwoconnectionpointsforPVPPMs,oneatbusbarR1andoneatnode
R15 of the LV network. Due to the similarity in the behaviour of the two different LV
networks only the voltage profile of the LV networks connected to busbar 1 of the MV
networkswillbepresented.TherestcanbefoundinAppendixF.
Firstly,inFigure4.9,thevoltageprofileofbusbarR1oftheLVnetworkconnectedtoMV
busbar1isdepictedforallMVnetworksindifferentcolours.
52
Figure4.9:Voltagesinp.u.atbusbarR1ofthelowvoltagenetworksconnectedtobusbar1oftheMV
networkswithR/Xratio4.4(differentcoloursindicatedifferentMVnetworks)
Asseenthestartingpointofthevoltageisat1p.uinallnetworks.Attheoccurrenceof
thefaultthevoltagedropsto0.46p.u.andthenforthefirst100msofthefaultperioditis
beingraisedto0.54p.u.At100msasecondvoltageplateauisdepictedloweringthevoltage
marginallyabove0.5p.u.Itshouldbenotedthatthevoltagetracesareverysimilaramong
the6MVnetworks.
Finally at 150msthe fault is cleared and thevoltage rises close to the prefault value.
However, even though the swing of the voltage that is seen postfault complicates the
graph, the steady state voltage postfault will be slightly lower than the prefault starting
point. The total lossof active power per aggregateLV network is in theorderof 7MW or
approximately35%oftheirpeakload(seeTable4.4).
InFigure4.10thevoltageprofileofnodeR15ofthesameMVconnectedLVnetworkas
thepreviousgraphisshown.ThesamepatternofbehaviourasinFigure4.9isdepicted.The
prefault voltage is almost at 1 p.u. and at fault occurrence the voltage drops to 0.48p.u.
The voltage is then raised during the first 100ms of the fault to the value of 0.56p.u. In
addition,at100msavoltageplateauat0.48p.u.isnoticed.At150msthefaultiscleared
and the voltage is raised slightly above 0.9p.u. Then the voltage oscillates due to the
oscillationofthesynchronousgeneratorsoftheeHVnetwork.
53
Figure4.10:Voltagesinp.u.atbusbarR15ofthelowvoltagenetworksconnectedtobusbar1oftheMV
networkswithR/Xratio0.43(differentcoloursindicatedifferentMVnetworks)
Firstly in Figure 4.9, the starting point of the voltage is at 1 p.u. due to the remote
controloptionoftheMV/HVtransformers.AsexplainedbeforetheMV/HVtransformersof
theMVsuburbannetworktargetbusbar10oftheMVnetwork.Sincethehighestloadofthe
MV suburban network is located at busbar 1 the power flow inside the MV network is
directedtobusbar1andthecorrespondingvoltagedroplowersthevoltageatthatbusbar
andconsequentlyattheLVnetworkthatisconnectedtothatspecificbusbar.Inthesame
figurethevoltageriseseenduringthefaultisanoutcomeofthereactivecurrentinjection
performed by the MV and HV connected DGs. Even though there is no reactive current
injectioninsidetheLVnetworkthevoltageisraisedasthevoltageofthecorrespondingMV
busbarisraised.
At100msthevoltagedipseenisanoutcomeofthedisconnectionofPVPPMsintheLV
network.TheactivepowerthatthePVPPMswereinjectingintothenetworkduringthefirst
100msofthefaultisnowlost.ThispowerisprovidedfromtheMVnetworkandcreatesa
larger voltage drop over the impedance of the MV lines and the MV to LV transformer in
ordertosupplytheload.
This phenomenon is also noticeable in Figure 4.11 where the active power exchange
overtheMV/LVtransformercanbeseenforthisLVnetwork.At100msaraiseintheactive
powerinfeedfromtheMVtoLVnetworkisseenasaresponsetothelossofthePVPPMs
activepowerproduction.Thebluelineindicatestheareasofnormalpowerflowandreverse
powerflow.
54
NPF
RPF
Figure4.11:ActivePowerexchangeoveroneofthe38parallel500kVAMV/LVtransformersconnectedto
busbar1oftheMVnetworkinp.u.(differentcoloursindicatedifferentMVnetworks,basevalue500kVA)
Figure4.12:Activepoweroutputofall38parallelLVconnectedPVPPMsconnectedtobusbarR1oftheLV
networkconnectedtoMVbusbar1,inp.u.withanaggregatedbasevalueof4560kVA
Finallythedecreaseinthepostfaultvoltageascomparedwiththeprefaultvalueisalso
explained by Figure 4.11. In this figure the postfault active power received from the MV
networkisraised,from0.4p.u.toabout0.8p.u.,tocoverthedeficitofactivepowerinside
theLVnetworkcausedbythePVPPMsdisconnection.Thisriseinactivepowerreceivedfrom
theMVnetworkresultsinaslightlyhighervoltagedropoverthetransformer,loweringthe
voltageatbusbarR1andinasignificantlyhighervoltagedropovertheLVcablesresultingin
alowervoltageatnodeR15.
55
In Figure 4.10 the voltage profile of node R15 of the examined LV network shows
differencesascomparedtoFigure4.9ofbusbarR1.Themaindifferenceisthatthevoltage
drop at theoccurrenceofthe fault is slightlysmaller, due to the longer distance from the
faultlocationasnodeR15issituateddeeperintheLVnetwork.AlsothevoltageatnodeR15
is raised higher as compared to busbar R1 during the first 100ms of the fault due to the
activecurrentinjectionperformedbythePVPPMoftheLVnetwork,whichkeepsacertain
amount of active power infeed during the first 100ms of the fault (Figure 4.12) prior to
disconnecting.
FurthermorethedisconnectionofthePVPPMscausesalargervoltagedipandalower
postfault voltage as compared to Figure 4.9. The reason for both these behaviours is the
factthatthePVPPMisconnecteddeeperintotheLVnetworkandexperiencesthecombined
lossofactivepowerinfeedfromthePVPPMconnectedhigherintheLVnetworkatR1and
its own loss. Also this power needs to be provided by the MV network causing an even
highervoltagedropthanbeforeduetothelowX/RratioofthelinesinsidetheLVnetwork.
This state is prolonged postfault as PVPPMs stay disconnected and therefore the voltage
remainsatalowervalueascomparedtotheprefaultone.
Figure4.13showsthereactivepowerexchangeovertheMV/LVtransformerconnected
at busbar 1 of the MV network in p.u. The rating of the transformer is 500kVA and the
positivesignisattributedtoapowerflowfromtheMVtotheLVlevel.Asseentheprefault
andpostfaultstatesarethesameasnoPVPPMwasconsumingorproducinganyreactive
power prefault. During fault, the voltage drop lowers the reactive power consumption of
the voltagedependent loads and the reactive power losses over the lines. In general the
reactivepowerexchangedidnothaveanyimportantinfluenceinthisloadcase.
Figure4.13:ReactivePowerexchangeoveroneofthe38parallel500kVAMV/LVtransformersconnected
inbusbar1oftheMVnetworkinp.u.(differentcoloursindicatedifferentMVnetworks,basevalue500kVA)
56
Finallyitisworthnoticingthatthefaultcreatedavoltagedropdownto0.4p.u.inthe
eHVbusbar8,howevertheLVnetworkinsidethedistributionsystemconnectedtobusbar8
experienced a voltage drop to 0.5p.u. and in the case of the LV networks connected to
busbar 10 of the MV network to 0.6p.u. (as this is located even further from the fault
location). This shows the high retained voltage in the LV networks during a transmission
systemfault.
Also this fault created a loss in photovoltaic power infeed of 42MW (see Figure 4.2)
whichcorrespondsto35%ofthepeakloadofallLVnetworks.IfeverybusbaroftheeHV
networkhadanidenticaldistributionsystemattachedtoitthesamefaultwouldpotentially
causealossof1600MWofphotovoltaicactivepower!
After presenting the noLVRT base case and performing an analysis of the results, a
comparison between the four basic control modes for this load case is presented. In this
comparison only the voltage profiles of the LV network connected to busbar 1 of the MV
networkarepresentedastheyaresimilartotheotherLVnetwork.
Thestartingvoltageforallcontrolcasesis0.97p.u.,at0sthefaultoccursanddrivesthe
voltagedowntoapproximately0.46p.u.forallcases.HoweveritisnotablethatintheZPM
(greenline)casethevoltagedropsevenlower(0.44p.u.).Duringthefirst100msofthefault
thevoltagerisesinallcasesbutwithadifferentslope.TheriseisgreaterinthecaseofaRCI
(blue line) followed closely by aRACI (black line), both of these cases achieve a voltage of
almost0.58p.u.TheZPMexhibitsthelowestriseinvoltageaswellasthenoLVRT(redline)
whichmanagetoraisethevoltagetoonly0.54p.u.whereastheothercontrolmodesriseto
0.58p.u.At100msasecondvoltageplateauisseenforallcases.Postfaultthevoltagestays
atlowerlevelwithrespecttotheprefaultvalueinallcasesandevenlowerinthenoLVRT
case.
57
Figure4.14:Voltagesinp.u.atbusbarR1ofthelowvoltagenetworksconnectedatbusbar1oftheMV
networkswithR/Xratio4.4forfourdifferentcontrolmodes
InFigure4.15thevoltageatnodeR15ofthelowvoltagenetworkconnectedtobusbar1
oftheMVnetworkispresented.Againthesamepatternisseen,withthevoltagestarting
point slightly lower than in Figure 4.14. The voltage drop at fault occurrence reaches 0.48
p.u.,andinthecaseofZPM(greenline),0.46p.u.Inthefirst100msofthefaulttherisein
voltage is steeper for the aRACI reaching almost 0.6p.u. while, aRCI (blue line) also has a
steep slope, followed by noLVRT (red line) at 0.56p.u. and the least rise in voltage is
exhibitedbyZPMwherethevoltageis0.53p.u.At100msthevoltageexhibitsaplateauin
all cases but as seen in the graph the ZPM and noLVRT drop below0.5p.u. whereas aRCI
andaRACIstayhigher.
58
Figure4.15:Voltagesinp.u.atbusbarR15ofthelowvoltagenetworksconnectedatbusbar1oftheMV
networkwithR/Xratio0.43forfourdifferentcontrolmodes
As seen in Figure 4.14 for the ZPM (green line) the voltage drop is lower at the
occurrence ofthe fault (0.44p.u.) with regard tothe other control modes(0.46p.u.). Also
thevoltageduringthefaultismarginallylowerfortheZPMcase(0.54p.u.)relativetothe
noLVRT(redline)andsignificantlylowerifcomparedwiththeaRCI(blackline)andaRACI
(blue line), 0.58p.u. This behaviour of the ZPM control case can be explained from Figure
4.16andFigure4.17.
InFigure4.16theactivepowerexchangeovertheMV/LV500kVAtransformercanbe
seenforallcontrolcases.AsseeninthisgraphattheoccurrenceofthefaulttheZPMcurve
(greenline)stayshigherthantherestexceptfortheaRCIcurve(blueline).Thismeansthat
forthefirst100msthePVPPMsoftheLVnetworkstopanyactivepowerexchangewiththe
network and therefore this power is now provided by the MV network creating a larger
voltagedropoverthetransformersandlines,thusloweringthevoltageevenfurther.Inthe
caseofnoLVRT(redline)thePVPPMsproducesomeactivepowerforthefirstperiodofthe
fault and therefore less power is required from theMV network,making thevoltage drop
smaller.InthisfiguretheaRACIcase(blackline)isinjectingactivecurrentatthesamelevel
asthenoLVRTcasewhilstaRCI(greenline)farlessasthelimitersareenabledafter50ms
duetoreactivecurrentinjectionpriorityduringfault.
59
0,8
0,6
0,4
NoLVRT
0,2 ZPM
a RCI
a RACI
0,0
0,0 0,3 0,6 0,9 1,2 [s] 1,5
Figure4.16:ActivePowerexchangeoveroneofthe38parallel500kVAMV/LVtransformersconnectedin
busbar1oftheMVnetworkforallMVnetworksinp.u.forallcontrolcases(basevalue500kVA)
AtthedisconnectionofthePVPPMsat100msZPMandnoLVRTareattheexactsame
level as in both case at that time no PVPPMs are producing any active power in the LV
network.FortheaRCIalsoactivepowerproductionislowbutthereissome,ashalfofthe
PVPPMsarestillconnected.FinallytheaRACIproducesthemostactivepowerofallcontrol
modesduringfault.
ThepatternseeninFigure4.16wouldleadtoahighervoltageduringfaultfortheno
LVRTandaRACI,however,thevoltageprofileofFigure4.14ismeasuredclosetotheMV/LV
distributiontransformerandtheX/Rratioofthenetworkatthatpointishigherthan1due
tothereactivenatureofthelinesoftheMVnetwork.Thereforereactivepowerexchange
overthetransformerwillinfluencethevoltageprofilestrongerthanactivepower.
In Figure 4.17 the reactive power exchange over the MV/LV 500kVA transformer is
depictedforallcontrolmodes.PrefaulttheLVnetworkisconsumingreactivepower,which
duetothevoltagedropattheoccurrenceofthefaultisloweredsignificantly.
60
0,3
0,2
0,1
NoLVRT
0,0 ZPM
aRCI
aRACI
-0,1
0,0 0,3 0,6 0,9 1,2 [s] 1,5
Figure4.17:ReactivePowerexchangeoveroneofthe38parallel500KVAMV/LVtransformersconnected
inbusbar1oftheMVnetworkforallMVnetworksinp.u.forallcontrolcases(basevalue500kVA)
For the aRCI and aRACI it is lowered further and eventually reversed due to the 2022
connectedPVPPMswhichinjectreactivecurrenttosupportthevoltage.
ThedisconnectionofthePVPPMsat100mslowersthevoltageofthenetworkwhichin
turn slightly decreases the consumption of reactive power of the LV network. In total the
activepowerandreactivepowergraphsofthetransformersshowthatduringthefaultonly
aRCIandaRACIinjectreactivepowertothenetworkwhileaRACIandnoLVRTinjectactive
poweruntiltheydisconnect.Theoutcomeofthisprocedureisahighervoltageduringfault
for the control modes that inject reactive power and only slight differences for the active
power injection proving the role of the high X/R ratio at busbar R1 of the low voltage
network.
FinallyforthepostfaulttimeframeinFigure4.14allcontrolmodeshavealowerpost
faultvoltageduetotheactivepowerlossofthedisconnectedPVPPMsandinthecaseofno
LVRTthevoltageisslightlylowerduetodoublethelossofactivepowerasinthiscaseallLV
connected PVPPMs disconnect from the network. However the difference is not very
significantasactivepowerlostisreplacedbytheMVnetworkandforbusbarR1thelosses
are not very large due to the short distance from production and high X/R ratio of the
networkatthatbusbar.
The results differ for Figure 4.15, as the postfault voltage is lower in all cases and
furthermorethenoLVRTcasehasavoltagelowerby0.05p.u.relativetotheothercases,
stressing the importance of the loss of active power on the voltage, deep inside the LV
networkthathavealongerdistancefromproductionandalowX/Rratio.
In the ZPM case the postfault voltage stays also at the levels of the noLVRT up until
0.7swhenitrisestotheleveloftheothercontrolmodes.Thereasonforthisbehaviouris
61
thatthevoltagethresholdfordeactivatingZPM(0.9p.u.)isnotreachedatfaultclearance.
HencetheZPMisappliedalsopostfaultandthereforeuntil0.63swhenthevoltagereaches
0.9p.u.PVPPMsarenotexchanginganyactiveorreactivepowerwiththenetworksimilarly
to the behaviour of noLVRT. The outcome of this behaviour is also seen in Figure 4.16,
wheretheactivepowerdrawnbytheMVnetworkishigherpostfaultfortheZPMbutlower
than noLVRT as only the PVPPMs connected to node R15 (where the voltage recovers
slower) are applying the control mode postfault. By contrast, the PVPPMs connected at
busbar R1 are already producing the prefault power at fault clearance as the voltage
thresholdismetatthatbusbar.
ZPM has the lowest voltage during the first 100ms of the fault as it injects no active
power, noLVRT is secondworse as it injects only active and no reactive power and aRACI
raisesthevoltageby0.1p.u.,whichissignificantduringthefaultasitinjectsbothactiveand
reactivepowerwiththesameangleasthenetworkimpedanceinnodeR15.
TheconclusionsofthisanalysisarethatthenoLVRTcausesundervoltagespostfaultin
thelowvoltagenetworkwhicharemoreseveredeeperinsidethenetworkwheretheX/R
ratio is smaller. About one third of the peak load is lost in generation which seems
unacceptablefromasystemperspective.
This behaviour is corrected by all other control modes due to fast postfault
reconnectionofthePVPPMs.WhileZPMallowsforfastreconnectionofPVPPMpostfaultin
principlethisreconnectionmightbedelayedduetodelayedvoltagerecovery.Thiswouldbe
unacceptableforapowersystemwithlowinertia.
Finally aRCI and aRACI are by far the superior control modes for raising the voltage
during fault, as seen in all figures, and have a similar effect to the voltage close to the
transformer, but deeper inside the network aRACI shows better results. Disregarding
detrimentaleffectsonnetworkprotection,theyarethepreferredcontrolmodes.
After presenting the comparison between control modes for the base case, the
correspondingcomparativeresultsforallcontrolmodesinfullreversepowerflow(fRPF)of
thenetworkwillbepresented.
InFigure4.18thevoltageatbusbarR1oftheLVnetworkconnectedtoMVbusbar1is
shownforallcontrolscenarios.Thevoltagestartsat0.97p.u.Asitisdepictedthevoltage
dropsdownto0.41p.u.atthemomentoffaultoccurrenceforallcontrolmodesexceptfor
ZPMwhereitdropsto0.38p.u.Duringthefirst100msofthefaultthereisasteepvoltage
rise that leads the voltage to 0.56 p.u. in aRCI and 0.52p.u. in the noLVRT case. The
differencesbetweencasesaresmallbutasitcanbeseenintheenlargedpartofthefigure,
thestrongestinfluenceonthevoltageduringthefirst100msofthefaultisappliedbyaRCI,
followed by aRACI, ZPM and finally last is noLVRT. After the first 100ms of the fault, the
62
disconnection of the LV connected PVPPMs is triggered resulting in a voltage rise for all
control modes, a rather unintuitive behaviour based on the results of the normal power
flow. In this 50ms time frame voltage support between control modes follows the same
orderasforthefirst100msofthefault.Atfaultclearancethevoltagerisestoahighervalue
thantheprefaultforallcases.Themostsignificantdifferencebetweenprefaultandpost
fault voltage is noticed in the noLVRT case. The rest of the cases have similar postfault
voltages.
1,2
1,0 0.1 s
0.56 p.u.
0,8 0.1 s
0.52 p.u.
0.01 s No LVRT
0,5 0.41 p.u. ZPM
0.01 s aRCI
0.38 p.u. aRACI
0,3
0,0 0,3 0,6 0,9 1,2 [s] 1,5
Figure4.18:Voltagesinp.u.atbusbarR1ofthelowvoltagenetworksconnectedatbusbar1oftheMV
networkswithR/Xratio4.4forfourdifferentcontrolmodesinfRPF
1,2
1,0 0.1s
0.64 p.u.
0,8 0.1 s
0.60 p.u.
0.01 s No LVRT
0,5 0.49 p.u . ZPM
0.01 s aRCI
0.45 p.u. aRACI
0,3
0,0 0,3 0,6 0,9 1,2 [s] 1,5
Figure4.19:Voltageinp.u.atbusbarR1ofthelowvoltagenetworkconnectedatbusbar10oftheMV
networkswithR/Xratio4.4forfourdifferentcontrolmodesinfRPF
63
In Figure 4.19 the voltageof busbar R1of the corresponding LV network connected in
busbar 10 of the MV network is shown. As seen the behaviour is similar to the previous
graph with the only difference of a higher retained voltage during the fault due to the
locationofthisbusbardeeperintotheMVnetworkandfurtherfromthefault.Howeverthis
results in higher postfault voltages values that in the noLVRT case reach the overvoltage
limitof1.1p.u.Thendisconnectionofloadsandproductionistriggeredandthesimulation
atthatpointcouldnotconvergeanymoreasthesystemwasledtoinstability.
InFigure4.20thevoltageprofileatbusbarR15oftheLVnetworkconnectedtobusbar1
oftheMVnetworkispresented.Adifferenceinthebehaviourofthevoltagecurvesrelative
toFigure4.18isnoticed.ThevoltagedropisnotasdeepasinFigure4.18,andthevoltage
riseduringthefaultissignificantlyhigherleadingthevoltagetoavalueof0.64p.u.inthe
aRACIcase.
Also the difference between the controlcases in raising the voltage during the fault is
morepronounced.MoreoverintheZPMcasethevoltageincreaseslessthantheothercases
(0.57p.u.) and in the aRACI the most, contrary to the voltage profile of busbar R1.
Furthermore the disconnection of PVPPMs at 100ms causes a voltage drop instead of an
increase, and postfault the voltage returns to the prefault operating point. The noLVRT
case depicts a slightly lower voltage postfault incomparisonwith the other cases but the
differenceisnotaspronouncedasinFigure4.18.
1,2
1,0 0.1 s
0.64 p.u.
0.1 s
0,8 0.60 p.u.
0.1 s
0.57 p.u.
No LVRT
0,5 0.01 s
0.48 p.u. ZPM
aRCI
0.01 s
0.46 p.u. aRACI
0,3
0,0 0,3 0,6 0,9 1,2 [s] 1,5
Figure4.20:Voltagesinp.u.atbusbarR15ofthelowvoltagenetworksconnectedatbusbar1oftheMV
networkswithR/Xratio0.43forfourdifferentcontrolmodesinfRPF
Theresultspresentedintheprevioussectionwillbeanalysedinthissectionwiththeaid
ofFigure4.21andFigure4.22thatdepicttheactiveandreactivepowerexchangeoverthe
MV/LVtransformerforallcontrolcases.Itisimportanttonoticefortheanalysisthatdueto
64
PVPPMsproducingactivepowerat1p.u.,inorderforthenetworktoenterreversepower
flow,theyalsoconsumereactivepowerprefaultduetotheirdynamicpowerfactorcontrol
as explained in section 3.4.1. This behaviour is apparent also from Figure 4.22 where the
prefaultpointhasaveryhighreactivepowerconsumptionifcomparedwiththebasecase
(Figure4.13).Thebehaviourswillbeanalysedfirstforthefaultperiodandthenforthepost
fault.
For the fault period ZPM exhibits the lowest voltage dip at the moment of fault
occurrencebothinFigure4.18andFigure4.20.Thatbehaviourcanbeexplainedasthe2022
PVPPMs that perform the ZPM control strategy at the occurrence of the fault drive the
active power production to zero faster than in the noLVRT case as deducted from Figure
4.21.Thelossofactivepowerproductionthatneedstobereplacedfromactivepowerfrom
theMV networkcauses alarger voltage drop dueto the lossesoverthe larger impedance
thatstandsbetweenproductionandload.AlsoZPMshowsthelowestvoltageforthefirst
100msofthefaultinFigure4.20,butonlythesecondlowestafternoLVRTinFigure4.18.
This behaviour is the combined effect of ZPM providing the least active current injection
duringfault(Figure4.21)andthesecondlargestreactivepowerconsumption(Figure4.22).
DuetothehighX/RratiooftheLVnetworkatbusbarR1reactivepowerconsumptionisa
morecrucialfactorinloweringthevoltagethantherisecausedbyactivecurrentinjection
andthereforethenoLVRTthatconsumesthehigherreactivepowerduringthefaulthasthe
lowest voltage profile. At node R15 of the LV network the X/R ratio is below 1 and hence
activepowerismoreinfluentialtothevoltage.Hence,ZPMwhichinjectslessactivepower
thanothercontrolmodesduringthefirst100msofthefaulthasthelowestvoltage(Figure
4.21).
0,6
0,3
0,0
NoLVRT
-0,3 ZPM
aRCI
aRACI
-0,6
0,0 0,3 0,6 0,9 1,2 [s] 1,5
Figure4.21:ActivePowerexchangeoveroneofthe38parallel500kVAMV/LVtransformersconnectedin
busbar1oftheMVnetworkinp.u.forallcontrolcasesinfRPF(basevalue500kVA)
65
For the first 100ms of the fault period the aRCI strategy, which injects the highest
reactive current and the second highest active current shows the best performance at
busbarR1oftheLVnetwork(Figure4.18)thathasahighX/Rratio.AtbusbarR15oftheLV
network (Figure 4.20) that has a low X/R ratio aRACI exhibits the higher voltage rise as it
injectsthemostactivecurrentandthesecondhighestreactivecurrent.
0,5
0,3
0,2
NoLVRT
0,0 ZPM
aRCI
aRACI
-0,1
0,0 0,3 0,6 0,9 1,2 [s] 1,5
Figure4.22:ReactivePowerexchangeoveroneofthe38parallel500kVAMV/LVtransformersconnected
inbusbar1oftheMVnetworkinp.u.forallcontrolcasesinfRPF(basevalue500kVA)
The combined effect of these two phenomena that have opposing effects: the loss of
activepowerleadstoavoltagedropandtheexcessofreactivepowertoavoltageincrease
affectsinadifferentwaybusbarsR1anR15oftheLVnetwork.ThedifferenceinX/Rratio
leadsthevoltageatbusbarR1torise(Figure4.18),duetothehigherinfluenceofreactive
power, and the voltage at node R15 to drop, due to the higher influence of active power
(Figure4.20).
Forthepostfaultperiodthevoltagebehaviourisagainaffecteddifferentlybyactiveand
reactivepower.PostfaultreconnectionofthePVPPMsinallcontrolmodesexceptforthe
noLVRTresultsinlessactivepowerconsumptionfromtheLVnetwork,andcreatesaslightly
highervoltagepostfaultatnodeR15(Figure4.20).However,thereconnectionofPVPPMs
createsahigherreactivepowerconsumptionforallcontrolmodesexceptthenoLVRTthat
66
result,inalowerpostfaultvoltageforthosemodesatbusbarR1(Figure4.18).Ingeneral
thereisanincreaseinthepostfaultvoltagemostlyatbusbarR1duetothe2012PVPPMs
thatdisconnectedandloweredthepostfaultreactivepowerconsumptionascomparedto
theprefaultone.
Therefore the interaction between dynamic power factor control and obligatory
disconnection during fault can lead to overvoltages. Also the prefault voltage drop
compensationbytransformertapchangerscanraisethepostfaultvoltagevalueduetothe
differentvoltageprofileofthenetworkpostfault.
The voltage drop during the fault is also affected by the prefault starting point of the
network. In order to present this effect the ZPM control case was chosen for all three
differentpowerflowsasseeninFigure4.23.
Figure4.23:Voltagesinp.u.atbusbarR1ofthelowvoltagenetworksconnectedatbusbar1oftheMV
networkswithR/Xratio4.4duringZPMforallloadcases
InthisfigurethevoltageatbusbarR1oftheLVnetworkconnectedtoMVbusbar1is
shown.Thevoltageattheoccurrenceofthefaultisdifferentforallloadcasesaswellasthe
depthofthedip.Thesecondplateauappearsasavoltagedecreaseforthebasecase(red
line)andthesRPF(greenline)andasavoltageincreaseforthefRPF(blueline).Postfault
thevoltagestayslowerthantheprefaultforthebasecaseandsRPFbutinthefRPFitstays
significantlyhigher.
InFigure4.24thevoltageatnodeR15oftheLVnetworkconnectedtoMVbusbar1is
shownforallloadcases.Inthisfiguretheprefaultvoltageisdifferentforallloadcasesas
well but not the same way as in Figure 4.23. During the fault the voltage support is more
67
intenseinthefRPFcaseandpostfaultthevoltagestaysatdifferentlevelsinallloadcase.At
0.66sinthebasecasethereisaverticalriseinthevoltage.
Figure4.24:Voltagesinp.u.atnodeR15ofthelowvoltagenetworksconnectedatbusbar1oftheMV
networkswithR/Xratio0.43duringZPMforallloadcases
In order to analyse the results between different load cases, the effect of the tap
changersmustberemoved.Bysubtractingfromtheprefaultvoltageofeveryloadcasethe
voltageatthefaulttheabsolutevoltagedipcanbecomparedinordertoseeinwhichcase
thevoltagedropwastheworseatbusbarR1(Figure4.23).InthefRPFtheabsolutevoltage
dropwas0.568p.u,inthesRPFitwas0.516p.u.andinthebasecaseitwas0.522p.u.
SotheworstvoltagedropwasexperiencedatfRPFwhereattheinstanceofthefaultthe
reactivepowerexchangeoverthetransformerdroppedby0.4p.u.(Figure4.22).Inthebase
casewherethevoltagedropwassmallerthereactivepowerexchangeoverthetransformer
droppedby0.18p.u.(Figure4.17)andinthesRPFbylessthan0.1p.u.(AppendixF).Asthe
shortcircuitvoltageoftheMV/LVtransformeris4.1%,thevoltagedropisproportionalto
the voltage drop caused over the transformer impedance due to the reactive and active
power loss at the instance of the fault. Therefore this graph aids in understanding the
reasoning for the differences in voltage drop between different load cases. The behaviour
postfaultisalreadyexplainedbypreviouscases.
InFigure4.24theimportantthinkstonotearethatprefaultthehighervoltage(1.011
p.u.) is seen for the fRPF, second higher for sRPF (0.975p.u.) and the lowest for the base
case(0.961p.u.).Thisishappeningduetothedifferenceinpowerflowdirectionbetween
cases.InthefRPFandsRPFthepowerflowsfromnodeR15tobusbarR1andthatiswhythe
68
voltageishigherinthesecasesintheR15node.ForthebasecasetheflowisfrombusbarR1
tonodeR15.
Thesecondimportantfindingofthisgraphisthatinthebasecaseat0.66stheimpactof
ZPMdelayedswitchoverfromfaulttonormaloperatingcontrolmodeonthevoltagecanbe
seen.DuetothelowervoltageofthatcaseandthefactthatinZPMduringfaultthereisno
support on the voltage, at fault clearance, the voltage fails to ascend higher than 0.9p.u.
The threshold for returning to normal operating condition is not met and the PVs stay in
faultmodehavingnoactiveorreactivepowerexchangewiththenetwork.Thereforewith
no active power boosting the voltage in the postfault period stays low until at 0.66s the
thresholdismetandthePVsreturntothenormaloperatingstate.Thenthevoltageisraised
significantlyduetotheactivepowerinfeedofthePVs.
In order to demonstrate the effect of delayed active power recovery (dAPR) on the
voltageprofileoftheLVnetworktwocaseswillbecomparedwiththecorrespondingcases
includingdAPR.ThechoiceofcasesaimstoshowcasetheextentoftheinfluenceofdAPR.
For this reason results will be presented for the ZPM case in comparison with ZPM with
dAPRforthebasecaseandtheLVtoHVreversepowerflow.Thentheresultsfromthose
twocaseswillbeanalysed.
Base Case
InFigure4.25thevoltageprofileofbusbarR1oftheLVnetworkconnectedtobusbar10
oftheMVnetworkispresentedforthecasewithZPMandthecasewithZPManddelayed
activepowerrecovery(dAPR)duringanormalpowerflow.ThechoiceofZPMasthecontrol
mode for presenting the effect of dAPR, is based on dAPR addition showing the most
difference in results as active power drop during fault is the greatest between the three
othercontrolmodes.
Asexplainedinsection3.4.1.2,thedAPRisappliedpostfaultand,therefore,theeffect
ofenablingthedAPRcanonlybeseenpostfault.AsseeninFigure4.25thevoltageprofile
showsminordifferencesbetweenthecasewithdAPR(greenline)whichstaysslightlylower
atfaultrestorationfromthecasewithoutdAPR(redline).
69
1,1
0,9
0,7
0,5
ZPM
ZPMwithdAPR
0,3
0,0 0,3 0,6 0,9 1,2 [s ] 1,5
Figure4.25:Voltageinp.u.atbusbarR1ofthelowvoltagenetworkconnectedtobusbar10oftheMV
networkforZPMwithandwithoutdelayedactivepowerrecovery
The case is different in Figure 4.26 where the voltage profile at node R15 inside the
same LV network is shown for both cases. Here the difference in the postfault voltage is
close to 0.05p.u. and theeffectof delaying thereturn to active power lowersthe voltage
postfaultenough tokeep the fault stateof the PVPPMs until 0.6swhenthevoltage rises
abovethethresholdof0.9p.u.
1,1
0,9
0,7
0,5
ZPM
ZPMwithdAPR
0,3
0,0 0,3 0,6 0,9 1,2 [s] 1,5
Figure4.26:Voltageinp.u.atbusbarR15ofthelowvoltagenetworkconnectedtobusbar10oftheMV
networkforZPMwithandwithoutdelayedactivepowerrecovery
70
LV to HV RPF
Thesamecontrolcase,ZPM,ischosentopresenttheperformanceofdAPRinthecase
offullreversepowerflow(fRPF)fromtheLVnetworktotheeHVnetwork.InFigure4.27the
voltageatbusbarR1oftheLVconnectedtobusbar10oftheMVnetworkispresented,for
ZPM (red line) and ZPM with dAPR (green line). The rise of the voltage in this load flow
scenario,postfault,ishigherinthecasewithdAPR.Moreimportantly,whilstinthecaseof
ZPM the voltage stays lower and avoids reaching the high voltage threshold of 1.1p.u., in
thecasewithdAPRthethresholdismetandfurtherdisconnectionsaretriggeredforcingthe
simulationtostop.
1,1
0,9
0,7
0,5
ZPM
ZPMwithdAPR
0,3
0,0 0,3 0,6 0,9 1,2 [s] 1,5
Figure4.27:Voltageinp.u.atbusbarR1ofthelowvoltagenetworkconnectedtobusbar10oftheMV
networkforZPMwithandwithoutdelayedactivepowerrecoveryinfRPF
In Figure 4.28 the voltage profile at busbar R15 inside the LV network connected to
busbar10oftheMVnetworkforreversepowerflowforbothcontrolscenariosispresented.
After fault restoration at 150ms the voltage in the ZPM (red line) case has a higher value
thaninthecasewithdAPR(greenline),butthedifferenceismarginalandmuchlessthanin
Figure4.27.AlsoherethesimulationforZPMwithdAPRendsbefore0.9s.
71
1,1
0,9
0,7
0,5
ZPM
ZPMwithdAPR
0,3
0,0 0,3 0,6 0,9 1,2 [s] 1,5
Figure4.28:Voltageinp.u.atbusbarR15ofthelowvoltagenetworkconnectedtobusbar10oftheMV
networkforZPMwithandwithoutdelayedactivepowerrecoveryinfRPF
Base Case
In Figure4.25 and Figure 4.26 the lower voltage of the case with dAPR is due to the
delayed active power recovery postfault. Only part of the active power of the 2022
connected PVPPMs is injected in the case with dAPR and therefore the rest of the active
powerprovidedbytheMVnetworkcausesalargervoltagedropoverthetransformersand
lines.
ThiseffectislessintenseinbusbarR1,duetothesmalldistancefromtheMVnetwork
andthehighX/Rratioatthatpointofthenetworkthatminimizestheeffectofactivepower
on the voltage. However it is more apparent in node R15 (Figure 4.26) due to the longer
distanceofthisnodefromtheMVnetworkandthelowX/RratioinsidetheLVnetworkthat
increasestheinfluenceofactivepoweronthevoltage.ThustheadditionofdAPRcreatesa
significantdifferenceinthevoltagepostfaultthatincreasestheseverityofthefault.
Itisconcludedthatinthebasecase(normalpowerflow)dAPRcreatesundervoltages
postfaultthatworsenthestateofthenetworkstability.
LV to HV RPF
In Figure 4.27 and Figure 4.28 the results for the influence of dAPR on fRPF are much
differentthaninthebasecase.ThisbehaviourcanbeexplainedbythestateofthePVPPMs
inthefullreversepowerflowcase.InthiscasethePVPPMs,prefault,areproducingactive
power at1p.u. and thereforeconsume reactive power in order to keep a power factorof
0.9 inductive as described in section 3.4.1.1. In the ZPM, active power production during
fault drops to zero and so does reactive power consumption. Hence postfault if active
72
powerrecoveryisdelayedsoisreactivepowerconsumption.Thisbehaviourcreatesadeficit
of active power postfault and an excess of reactive power, since PVPPMs dont consume
reactive power as prefault. The combined effect of these two components raises the
voltageinbusbarR1duetothelowX/Rratiowhichfavourstheeffectofreactivepowerover
activeonthevoltage.Thiseventuallycreatespostfaultovervoltagesthatleadthesystem
tofurtherdisconnectionsofproductionandload.
In Figure 4.28, the effect is reversed as at node R15 the X/R ratio is below 1, and
thereforetheactivepowerdeficithasastrongerinfluenceonthevoltagethantheexcessof
reactive power. The previously described conditions apply as well but the end result is
different, as the voltage in the case with dAPR is lower than the case with instantaneous
active power recovery due to the delay of active power infeed by the 2022 PVPPMs
loweringthevoltage.Howeverduetotheexcessofreactivepoweratthenetworkthiseffect
ismilderandthevoltagedropisbarelynoticeable.
ItisconcludedthatinthefRPFcasetheeffectofAPRiscreatingovervoltageshigherin
the LV networkwhile connection points deep in the LV network slightly reduces the over
voltagescreated.
4.5 Conclusions
Inthischapterthemainsimulationresultsofthisthesiswerepresented,foranumberof
specificsystemconditionsandcontrolmodes.Firstthetestsystemwaspresented,withan
argumentation for the choice of DG capacity that reflect the level of PV penetration for
Germany in 2022. A thorough description of the sensitivity factors that helped build the
studycaseswasgiven.Finallytheresultsforthemostimportantcaseswerepresentedand
analysedindepth.Theconclusionsofthisanalysiswillbepresentedinthenextchapter.
73
74
5.1 Introduction
In this chapter the conclusions of this thesis will be presented. The general
recommendations, based on the findings of the study, regarding the sensitivity of the
voltage dipdepth in LV networks with high penetration of DGwill be presented. Also, the
impact of reverse power flow will be discussed with a focus on the consequences for the
postfault period. Furthermore, the recommendations for different control modes will be
mentioned, as well as the relevance of delayed postfault active power recovery. The
correlation between the results and the German network will be analysed and finally,
recommendationsforfutureresearchwillbeproposed.
5.2 Conclusions
The results presented in this study depend, to a certain extent, on the particular
network topology and are only applicable for the chosen DG penetration levels. Only
balancedtransmissionsystemfaultswerestudied,astheyarethemostseveretypeoffaults.
ThemainconclusionsforthevoltagesensitivityinLVnetworkswithhighpenetrationofPVs
duringtransmissionsystemfaults,willbepresentedinbulletpointsandanalysedbelow.
RetainedvoltageattheLVnetworkishigherascomparedtoothervoltagelevelsduring
fault.HigherpowerinfeedduringfaultfromLVconnectedDGisthereforepossible.
Asseenfromtheresults,theretainedvoltageintheLVnetwork,duringafault,stays0.1
to 0.2p.u. higher as compared to the eHV busbar connected to the distribution network
(BB8).Therefore,currentinfeedduringthefaultresultsinhigherpower,duetothehigher
retainedvoltageattheLVnetwork.Ofcourse,thedipinvoltagealsocontrolstheamountof
currentthattheDGinjects.Hence,thetotalcontributioncanbethesameaboveacertain
voltageinbothlocations.
Thedepthofthevoltagedipissensitivetolossofproductionofbothactiveandreactive
powerattheinstanceofthefault.
Atthefirstmsaftertheoccurrenceofthefaulttheamountofactiveandreactivepower
lost instantaneously has an impact on the voltage drop, as seen especially in the case of
ZPM. In this case, the loss of active and reactive power ismaximized due to the propriety
control strategy resulting in the lowest voltage drop during the fault. Therefore, a good
approachwouldbetotryandkeepthelossofactiveandreactivepoweratfaultoccurrence
aslowaspossibletominimizethevoltagedrop.
75
Voltage support from LV connected DG, during fault, can increase the voltage
significantly.
As seen from the results, the cases with aRCI or aRACI provide strong support to the
voltage during fault. Also, in these cases the voltage profile of the network is higher as
comparedtothecaseswhereallLVconnectedPVdisconnectby0.1p.uormore.Thisfinding
alsoprovestheimportanceofprovidingreactiveandactivecurrentsupporttothenetwork
incaseofafault.
Active current injection during fault has a more profound effect than reactive current
injectioninsupportingthevoltage,deepinsidetheLVnetwork.
TheaRACIcontrolstrategyusedshowedimprovedvoltagesupportascomparedtothe
aRCIstrategywhenapplieddeepinsidetheLVnetwork.Thereasonforthehighinfluenceof
activepoweronthevoltageisduetothelowX/RratiooftheLVnetwork.Therefore,itis
safetoassumethatthistypeofsupportisrelevantfortheDGconnecteddeepinsidetheLV
networkandshouldbepreferredovertheaRCI.
Reactivecurrentinjection,duringfault,performsbetterinsupportingthevoltagecloser
totheMV/LVtransformeroftheLVnetwork.
TheaRCIcontrolstrategyproposed,achievedthebestvoltagesupportascomparedto
the rest of the strategies when applied close to the transformer of the LV network. The
reasonforthehighinfluenceofreactivepoweronthevoltageisduetothehighX/Rratioof
theLVnetworkclosetotheMV/LVtransformer.Therefore,thistypeofsupportiseffective
fortheDGsconnectedclosetotheMV/LVtransformer.
DuringnormalpowerfloworMVtoHVreversepowerflow,disconnectionofPVsinthe
eventofatransmissionsystemfaultcreateasignificantvoltagedrop,duringthefault,
throughoutthenetwork.
Thelossofactivepowerduringnormalpowerfloworpartialreversepowerflowfrom
MVtoHVlevel,createsavoltagedipwhichismoreintenseinsidetheLVnetworkthanin
otherbusbarsofthesystem.Thisvoltagedipiscreatedasanimminentresultofthelossof
DG active power at the disconnection moment of DG. The active power lost from LV
connectedDGisnowprovidedbythehighervoltagelevelresultinginalowervoltage,dueto
an increased voltage drop over the lines and transformers of the network. This effect is
stronger inside the LV network, as the distance is longer and the lines have a higher
resistance resulting in a larger voltage drop. The second voltage plateau should be
consideredinapossibleFRTcurvetospecifyifthevoltagedipisdeeperattheoccurrenceof
thefaultoratthedisconnectionoftheDG.Ofcourse,thiswilldependontheamountofLV
connectedDG.
76
ThedisconnectionofPVsinthecaseoffullreversepowerflowcreatesadeficitofactive
power and an excess of reactive power, as LV connected PVs were consuming reactive
powerprefault.Thefinaloutcomeisavoltageriseatthepointsofthenetworkwithahigh
X/RratioandavoltagedipatthepointsofthenetworkwithalowX/Rratio.
ImpactofX/RratiooftheLVnetworkisinfluentialwhenperformingvoltagesupport.
As seen from the results, the impedance angle of the network at the DG connection
pointsplaysanimportantroleintheinfluenceofreactiveandactivepoweronthevoltage.
Therefore,itshouldbetakenintoaccountwhenperformingvoltagesupport.
PostfaultreconnectionofthePVsbenefitsthevoltageprofilesignificantlyinallcases.
PostfaultundervoltagesinthebasecaseandinMVtoHVreversepowerflowarelower
ifcurrentpracticesarefollowedandPVsdonotreconnectimmediatelytothenetwork.In
theLVtoHVreversepowerflow,postfaultovervoltagesarehigherifcurrentpracticesare
followedandPVsdontreconnectimmediatelytothenetwork.
Inthecaseofreversepowerflowtheresultswereverydifferentthanintheotherload
flowcases,forallcontrolmodes.Theresultsfromreversepowerflowarealsolinkedwith
thePVsproducingmaximumactivepowerandtherefore,consumingreactivepowerduring
theprefaultperiod.TheresultsforthefRPFwouldchangeifthePVssteadystatedynamic
powerfactorcontrolwasdifferent.
Thejoinedeffectoflossofactivepowerandlossofreactivepowerconsumptionisthe
maintriggerforthedifferenceinresults.Thevoltagesupportduringfaultbooststhevoltage
higherinthereversepowerflowcase.Postfaultthevoltagestayssignificantlyhigherthan
prefault due to the excess of reactive power in the network, caused by the PV
disconnection.
The impact of reverse power flow is, in essence, the impact of the PVPPM prefault
reactivepowerconsumptionthatcreatespostfaultovervoltagesandaltersthebehaviour
ofthenetworkduringthefault.
Also, another factor that plays a role in the postfault behaviour of the voltage is the
effectofthetapchangersofthetransformers.Prefaulttapchangersaresettocompensate
for the voltage dropover the lines, but due to the different load flow state ofthe system
postfaulttheirsettingsarefurtherboostingtheovervoltageofthenetwork.
The network fault control modes used in this study were compared with the current
standards for disconnection of the low voltage connected DGs. Their performance varied,
butingeneralshowedimprovementregardingvoltagesupportandsystemvoltagestability.
Intheeventofasystemfault,theapplicationofcurrentstandardswouldleadtoalossof
77
4284MW of active power infeed, representing 3570% of LV peak load. These values
wouldbehalvedbyapplyinganyofthecontrolmodespresented.
Specifically,thefirstcontrolmodeused,ZPM,ascomparedwiththenoLVRT,decreased
thevoltagefurtherattheoccurrenceofthefaultandduringthefaultsupportedthevoltage
slightly less in the base case and MV to HV reverse power flow. On the positive side, it
supported the voltage more during LV to HV reverse power flow close to the MV/LV
transformer, and postfault kept the voltage closer to the prefault value. It is an
improvementwithrespecttothenoLVRTmodeonlyregardingthepostfaultbehaviourof
thesystemvoltage.DuringfaultitperformsworsethanthenoLVRTmodeinsupportingthe
voltage, with few exceptions. Also, it can lead to delayed voltage recovery, as seen in the
resultsfornodeR15,deepinsidetheLVnetworkinthebasecase.
The second mode used, aRCI, as compared with the noLVRT, helps in supporting the
voltageprofileduringfaultinallcases.Thevoltagesupportduringfaultwasalways0.05p.u.
ormore,higherthanthenoLVRTmodeand0.07p.u.higherthaninZPM.Theinfluenceof
aRCI on the voltage was stronger closer to the MV/LV transformer due to the higher X/R
ratio of the network. This fault mode also improved the voltage postfault in all load flow
scenarios better than the noLVRT mode. In total the aRCI control strategy would be
beneficialforthenetworkintheeventofatransmissionsystemfault,ifitwasappliedtothe
DGsconnectedclosertotheMV/LVtransformerinthelowvoltagenetwork.
Finally,theaRACIstrategyshowedimprovedvoltagesupportduringthefaultwithvalues
similar to the aRCI case. In the case of low voltage connected DG, deep inside the LV
networktheperformancewasevenbetterthanaRCIastheX/Rratioofthenetwork,atthat
point, was below 1 and active current injection was more influential to the voltage than
reactive current. Postfault, as in all control cases, the voltage was benefited by the
reconnection tothe networkof the 2022 connected PVPPMs.This controlmode would be
verybeneficialtothenetwork,ifappliedtotheDGconnecteddeeperinsidetheLVnetwork.
However, the possibility of interference with network protection should be investigated,
eventhoughtheliteraturesurveydoneinthisthesissuggeststhatitisnotanissue.
Theadditionofdelayedactivepowerrecoverypostfault(dAPR)wasinvestigatedforthe
postfaultbehaviourofthevoltageinallcontrolcases.DuetothenatureofaRCIandaRACI
that kept the same, if not higher, active current infeed to the network during fault, the
application of dAPR was rarely activated, in those control cases. This behaviour can be
explainedthroughtheconditionforperformingdAPR:activecurrentinfeedofDGatfault
restorationmustbelowerthantheprefaultone.
InthecaseofZPMhoweverthisrequirementwasmetandthebehaviourofdAPRwas
analysed. As seen from the results of section 4.4.5, dAPR is mostly inhibiting, postfault,
restorationofthevoltage.InthebasecaseandMVtoHVreversepowerflow,thegeneral
postfault undervoltage is worsened with the addition of dAPR. DAPR lowers the active
powerinfeedofDGatfaultrestorationand,consequently,increasesthepowerextracted
78
fromtheMVnetworkandthevoltagedropitinduces.ThisbehaviourcanleadtoPVPPMs
stayinginfaultmodelongerthanwithoutdAPR.
In the LV to MV full reverse power flow, the postfault overvoltage is typically raised
even more with the addition of dAPR, as the PVs delay their reactive power consumption
postfaultandtherefore,theexcessofreactivepowerishigherandraisesthevoltagemore.
As a general conclusion, the effect of dAPR in the low voltage network in terms of
voltage stability is negative in all cases as it leads to postfault undervoltages or over
voltages. From a voltage stability point of view it should not be included in future LV
connectedDG.
ThemodellingstructureofthetestsystemusedinthisstudywasbasedontheGerman
network and the behaviour of installed DG was modelled based on the grid codes of
Germany. However someaspects of themodelling are not 100% representative of typical
German distribution networks. The major difference between this study and the German
networkisthatinstalledloaddataweretakenfromtheCigrbenchmarkreportsratherthan
from statistical data ofthe German network. The reason was the unavailabilityof German
loaddataduringthemodellingphase.However,mostbenchmarkreportsusedarebasedon
existing German networks and consequently, certain resemblance for the load data is
expected.
Therefore,itisnotsafetoassumethatthisnetworkdescribesindetailthebehaviourof
theGermannetwork,butratherthatitgivestheoutlineofitsresponsetoanetworkfault.
Anotherpointofdifferenceisthatnodynamicloadwasconsideredinthisstudy,onlystatic
load.
NeverthelesssomegeneralconclusionscanbedrawnfortheGermannetworkbasedon
thisstudy:
German grid codes amendment point: Steady state reactive power consumption of
PVPPMs combined with disconnection at fault leads to postfault over voltages in
reversepowerflow.
The LV grid codes of Germany stipulate that during steady state operation of the
network,installedPVPPMsoftheLVnetworkareobligedtoconsumereactivepoweratfull
active power output in order to maintain a power factor of 0.9 inductive in case of
installationslargerthan13.8KVAand0.95incaseofinstallationssmallerthan13.8KVA[10].
This requirement is based on avoiding overvoltages at the PCC of PVPPMs, due to high
active power infeed. Furthermore in the same publication by VDEFNN low voltage
connectedPVPPMsarerequiredtodisconnectfromthenetworkintheeventofanetwork
faultthat,wouldlowerthevoltageatthePCCbelow0.8p.u.[10].
The combination of these two requirements, as seen from the results of this study,
creates postfault overvoltages in the network as the reactive power consumption from
79
PVPPMs is not applied postfault, due to the disconnection. Consequently, the excess of
reactive power in the network creates overvoltages up to 1.1p.u. that lead to unstable
postfault behavior and may disconnect further load and production from the grid. This
behaviorisanoutcomeofthecombinedeffectofthesegridcodesthathinderthenetwork
responseintheeventofafaultduringreversepowerflowanddegradethestabilityofthe
system.
Asseenintheresultschapter,whentherequirementforreactivepowerconsumptionis
withdrawn,postfaultbehaviorofthenetworkisbenefitedandsystemstabilityisfavored.
Therefore, it is recommended that the relevant grid codes should be revised within a
reasonabletimeframeinordertoensurethatthisbehaviorisavoided.Thefinaloutcomeof
this procedure will depend on more factors than voltage stability and it will need further
research and understanding of the benefits and pitfalls that a proposed solution will
introduce. However, it is a finding of this study that the current grid codes hinder system
stability under specific circumstances and that this issue should be investigated and
resolved.
RetainedvoltageatLVnetworknodesishighduringfaultandthatallowsforsupport.
Asalreadyexplainedinthegeneralconclusions,theretainedvoltageintheLVnetwork
duringafaultishighandthatmakestheLVnetworkastrategiclocationforcurrentinjection
during the fault in order to support the voltage. This is an added benefit of using voltage
supportintheLVnetwork.
Postfaultreconnectionbenefitsthevoltagesignificantly.
PostfaultreconnectionoftheDGhasapositiveeffectinthevoltageinallloadcases.In
normal power flow or MV to HV reverse power flow the undervoltage seen postfault is
significantlyraisedincaseofreconnectionofDG.InthecaseofLVtoHVreversepowerflow
the general overvoltage seen postfault is lowered in case of DG, postfault, reconnection
duetoDGsresumingreactivepowerconsumption.Soregardlessofthefaultcontrolmode
followed,postfaultreconnectionhasapositiveeffectonthevoltagestabilityofthenetwork
andshouldbeconsideredforsettingfuturegridcoderequirements.
FRTcriteriaforLVnetworksarerelevantandmayneedtobeadjusteddependingonDG
location.
ThegeneralresultofapplyingFRTcontrolmodesisasignificantvoltagesupportduring
fault and postfault. The feasibility of such an addition should be investigated further, but
the benefit for the voltage profile of the network is significant and therefore, makes FRT
requirements relevant. Specifically, the aRCI strategy shows the best results for voltage
supportforDGconnectedneartheMV/LVtransformerandaRACIforDGconnecteddeeper
insidetheLVnetwork.
80
5.3 Recommendations for Future Research
For future research there are a number of topics that seem interesting for continuing
theworkdoneonthisthesis:
Useofthisnetworkasavalidationmodelforaggregatednetworkstructures.
The network structure created in this study is elaborate and rather complicated in the
attempt made to resemble an actual network with different voltage levels. Of course
creating such a structure is time consuming and also delays simulations and makes the
systemhardertoworkwith.Agoodwayoftakingadvantageofthisworkisthecreationof
anequivalentaggregatedistributionsystemthatcanproveeasiertoworkwithandvalidate
theresultsofthissystemwiththedetailedsystempresentedinthisstudy.Thiswouldalso
overcomethelimitationofthepresentmodelthathasonlyonedistributiongridconnected
ateHVbusbar8.
IncludeloaddatafromGermannetworkandapercentageofdynamicload.
An addition to this studythat would make results even more relevant for theGerman
networkisintroducingloaddataforinstalledloadsthatstemfromstatisticalanalysisofthe
Germannetwork.Alsotheuseofdynamicloadforapercentageoftheinstalledloadwould
ensurehigheraccuracyoftheresultsandwouldalsomakethenetworkmorerealistic.
Performsensitivityonfaultlocation.
Thenumberofsensitivityfactorsisnumerousforthisstudy.Aninterestingchoicewould
be to vary the fault location in order to study the importance of voltage support from LV
connected generators during faults that are very severe or very mild to the distribution
systemconnectedonaspecificbusbaroftheeHVnetwork.
PerformsensitivityonSCpoweroftransmissionsystem.
Short circuit power of the transmission system is very important in assessing the
contributionoftheLVnetworktothefault.Asensitivityanalysisshouldbeperformedwith
differentunitcommitmentateHVleveldependingonthepowerflowcase.
Useofadditionaldistributionsystems.
Anotherinterestingadditionistheuseofmoredistributiongridstructuresliketheone
usedinthisstudyconnectedtodifferenteHVbusbarssothatthetotalcontributioninthe
voltageprofilecanbeexamined.Thiswouldalsomakethesystemmorerealisticbecausein
realityalmostallbusbarshavedistributionnetworksconnectedtothem.Thisoptioncanbe
combined with the use of aggregated distribution networks in order to speedup the
simulationprocess.
FurtherworkonDGmodels.
ThemodellingofDGinthisstudywasverydetailedbutsomesimplificationsneededto
be made like the introduction of the chopper on the dc side of the PV inverter. There is
81
always space for improvement in the modelling of DG in order to ensure more detailed
behaviourthatcouldgiveinsighttotheinteractionsbetweenthenetworkandDG.
Impactondistributionnetworkprotection.
FinallyanissuewiththecontributionofLVconnectedDGduringafaultistheresponse
of protection mechanisms of the DG itself and of the network. As seen in the literature
surveythisissueismostlyseenincaseoffaultswithintherespectiveLVorMVdistribution
networkbutitisworthwhiletoinvestigateinterferenceofDGFRTwithnetworkprotection.
Investigatebehaviourforothertypeoffaults.
Forexampleofdifferentvoltagedipspropagatingthrougha Yd transformersee[67].
BasedonthespecifictransformationbyYdtransformers,theeffectofatwophaseto
earthfault would be the most severe for the voltage dip as the voltage drops below the
lowerboundary(0.8p.u.)fornormaloperationinallthreephases.Nextfollowsthephase
toearthfaults with only two phases below the limit and finally the phasetophasefaults
whereonlyonephasevoltageisbelowthenormaloperatinglimit.Dependingonthetypeof
transformersusedadditionalfaultswillbeprioritizedbasedonseverity.
82
References
[1] Grid integration aspects of large solar PV installations: LVRT capability and reactive
power/voltagesupportrequirements.2011IEEETrondheimPowerTech,Trondheim,19
23June,2011.
[2] E.ONNetzGmbH,GridConnectionRegulationsforHighandExtraHighVoltage,Apr.
2006.
[4] Technische Bedingungen fr den Anschluss und Betrieb von Kundenanlagen an das
Hochspannungsnetz(TABHochspannung),EVDEARN4120,2012.
[6] P.Kundur,Powersystemstabilityandcontrol:NewYork:McGrawHill,Inc,1994.
[8] ThomasAckermann,GranAndersson,andLennartSder,Distributedgeneration:a
definition,ElectricPowerSystemsResearch,vol.57,no.3,pp.195204,2001.
[9] Frank Marten, Jan von Appen, and Martin Braun, German perspectives on future
energysystemchallengesbyphotovoltaic:HowmuchPVcanbeintegrated?Smart
PV Grid Integration, Fraunhofer Institute for Wind Energy and Energy
SystemTechnology,Mar.2013.
[10] Generators in the low voltage distribution network. Application guide for generating
plantsconnectiontoandparalleloperationwiththelowvoltagenetwork,VDEARN
4105,2011.
[11] 1547 IEEE Standard for Interconnecting Distributed Resources with Electric Power
Systems,2003.
[12] entsoe,ENTSOENetworkCodeforRequirementsforGridConnectionapplicabletoall
Generators. Available: https://www.entsoe.eu/resources/network
codes/requirementsforgenerators/(2012,Aug.18).
83
[14] FNN, VDE|FNNRoadmap identifiziert weiteren Entwicklungsbedarf zum Umbau der
Netze:Pressrelease.Berlin,2012.
[15] entsoe,BriefingNotetoallinterestedpartiesonthestatusoftheNetworkCodeon
RequirementsforGenerators(2013,Apr.02).
[16] Requirements for the connection of microgenerators in parallel with public low
voltagedistributionnetworks,prEN50438:201208,2010.
[17] Requirements for the connection of generators above 16 A per phase Part 1:
ConnectiontotheLVdistributionsystem,CLC/FprTS505491:2011,2012.
[19] Ecofys, Facilitation of Renewables. Work Package 3 Final Report, in All Island TSO
FacilitationofRenewablesStudies:EirGrid,2010.
[21] J.C.Boemer,K.Burges,P.Zolotarev,J.Lehner,P.Wajant,M.Frst,R.Brohm,andT.
Kumm,OverviewofGermanGridIssuesandRetrofitofGermanPhotovoltaicPower
Plants for the Prevention of Frequency Stability Problems in Abnormal System
Conditions of the ENTSOE Region Continental Europe, in Proc. 1st International
WorkshoponIntegrationofSolarPowerintoPower,Aarhus,Denmark,2011.
[24] M.BaranandI.ElMarkabi,Adaptiveovercurrentprotectionfordistributionfeeders
withdistributedgenerators,inPowerSystemsConferenceandExposition,2004.IEEE
PES,pp.715719,2004.
[25] P.S.GeorgilakisandN.D.Hatziargyriou,OptimalDistributedGenerationPlacement
in Power Distribution Networks: Models, Methods, and Future Research, Power
Systems, IEEE Transactions on, vol. PP, no. 99, pp. 19,
http://ieeexplore.ieee.org/stamp/stamp.jsp?arnumber=6418071,2013.
84
[26] J. Morren and S. de Haan, Impact of distributed generation units with power
electronicconvertersondistributionnetworkprotection,inDevelopmentsinPower
System Protection, 2008. DPSP 2008. IET 9th International Conference on, pp. 664
669,2008.
[28] VDN, TransmissionCode 2007. Network and System Rules of the German
TransmissionSystemOperators,VerbandderNetzbetreiberVDNe.V.beimVDEW
(German Association of Network Operators, since beginning of 2008 part of BDEW),
Aug.2007.
[30] R.Walling,DERImpactsonBulkGridDynamicPerformanceARetrospectiveViewofa
PrescientThoughtExercise,InverterBasedGenerationInterconnectionRequirements
PJM, April 11, 2012, http://mydocs.epri.com/docs/publicmeetingmaterials/417
2012/05BDERImpactsonBulkGridDynamicPerformanceReighWalling.pdf.
[31] E.Camm,M.Edds,C.Murray,W.Peter,andP.Valverde,CurrentRenewableEnergy
Generator Technical Interconnection Requirements, in 2012 IEEE PES General
Meeting,2226July2011,SanDiego,USA2012IEEEPESGeneralMeeting,SanDiego,
USA,2012.
[34] I.ErlichandH.Brakelmann,IntegrationofWindPowerintotheGermanHighVoltage
TransmissionGrid,inPowerEngineeringSocietyGeneralMeeting,2007.IEEE,pp.1
8,2007.
[35] Bauer P, System aspects of solar photovoltaic power generators, Lecture script,
DelftUniversityofTechnology,May2012.
[37] TUBraunschweig,"PVplantconfigurations",Jan.2013.
85
[38] D. Molina, Progress report on the development and validation of DIgSILENT
PowerFactorymodelofthe39Busbenchmarksystemforstabilitycontrols:Addressed
toIEEETaskForceonBenchmarkSystemsforStabilityControls,Sep.2012.
[39] Evangelos Galinas, Sensitivity Analysis for Expansion of Transmission Systems with
HighAmountsofRenewables,MasterThesis,TUDelft,Delft,2013.
[40] J.C.Boemer,A.vanderMeer,B.G.Rawn,R.L.Hendriks,A.R.Ciupuliga,M.Gibescu,
W. L. Kling, and J. A. Ferreira, Fault Ridethrough Requirements for Onshore Wind
Power Plants in Europe: the Needs of the Power System, in 2011 IEEE PES General
Meeting,2428July2011,Detroit,USA2011IEEEPESGeneralMeeting,Detroit,USA,
2011.
[41] A.Yazdani,A.R.DiFazio,H.Ghoddami,M.Russo,M.Kazerani,J.Jatskevich,K.Strunz,
S.Leva,andJ.A.Martinez,ModelingGuidelinesandaBenchmarkforPowerSystem
Simulation Studies of ThreePhase SingleStage Photovoltaic Systems, IEEE
TransactionsonPowerDelivery,vol.26,no.2,pp.12471264,2011.
[42] R.F.ArrittandR.C.Dugan,SR,TheIEEE8500nodetestfeeder,inTransmissionand
DistributionConferenceandExposition,2010IEEEPES,pp.16,2010.
[43] F.K.O.W.P.B.FalkSchaller,ModellierungrealittsnaherzuknftigerReferenznetze
im Verteilnetzsektor zur berprfung der Elektroenergiequalitt, in Internationaler
ETGKongress2011(ETGFB130):UmsetzungskonzeptenachhaltigerEnergiesysteme
Erzeugung, Netze, Verbrauch, Vortrge Fachtagungen 1 bis 5Internationaler ETG
Kongress,Wrzburg,2011.
[46] K. Strunz, Developing benchmark models for studying the integration of distributed
energy resources, in Power Engineering Society General Meeting, 2006. IEEE, pp. 2
pp,2006.
[47] CIGRE Task Force C6.04.02, Benchmark Systems for Network Integration of
Renewable and Distributed Energy Resources: Technical Brochure (excerpts for
EuropeanHV,MV,andLVbenchmarksystems),,May.2013.
[48] 50Hertz Transmission GmbH, Amprion GmbH, TenneT TSO GmbH, and TransnetBW
GmbH, Netzentwicklungsplan Strom 2012 (Network Development Plan Electricity
2012):2.berarbeiteterEntwurfderbertragungsnetzbetreiber(2ndreviseddraftof
thetransmissionsystemoperators),May.2012.
[49] K.Skaloumpakas,ScalingthePVTemplate,Dec.2013.
86
[50] DIgSILENT GmbH, DIgSILENT PowerFactory: TechRef Overhead Lines Models:
DIgSILENTGmbH,2009.
[51] Ludvig Lindstr, Evaluating impact on ampacity according to IEC60287 regarding
thermally unfavourable placement of power cables: Masters Degree Project, KTH,
Stockholm,Sweden,2011.
[53] DIgSILENTGmbH,DFIGTemplate:Manual,Gomaringen,May.2011.
[54] DIgSILENTGmbH,FullyRatedWTGTemplate:Manual,Gomaringen,May.2011.
[55] Hua Bai Student Member IEEE, Pei Zhang Senior Member IEEE, and Venkataramana
AjjarapuFellowIEEE,ANovelParameterIdentificationApproachviaHybridLearning
forAggregateLoadModeling,
[56] Ins Romero Navarro, Dynamic Power System Load Estimation of Parameters from
OperationalData,PhD,LundUniversity,2005.
[57] KPSchneider,JCFuller,FKTuffner,andRSingh,EvaluationofConservationVoltage
Reduction(CVR)onaNationalLevel,Jul.2010.
[58] KeYao, Modeling power system load using intelligent methods, Master Thesis,
KANSASSTATEUNIVERSITY,Manhattan,Kansas,2011.
[60] S.M.I.H.Guo,M.I.K.Rudion,M.I.H.Abildgaard,anda.Z.A.S.S.M.I.P.Komarnicki,
Parameter Estimation of Dynamic Load Model Using Field Measurement Data
PerformedbyOLTCOperation,2008.
[61] A.MaitraMemberIEEE,A.GaikwadMemberIEEE,andP.ZhangMemberIEEE,Using
SystemDisturbanceMeasurementDatatoDevelopImprovedLoadModels,2008.
[62] DIgSILENT GmbH, DIgSILENT PowerFactory: General Load Model: DIgSILENT GmbH,
2008.
[63] Von Gunter Buse, Richard Bopp, and Harald Junkermann Kassel, Das Verhalten der
NetzleistungbeiveranderlicherNetzspannung,Elektrizitatswirtchaft,1983.
[65] DIgSILENTGmbH,GeneralLoadModel,2008.
[66] M.H.J.Bollen,UnderstandingPowerQualityProblems:WileyIEEEPress,2000.
87
[67] M.AungandJ.Milanovic,TheInfluenceofTransformerWindingConnectionsonthe
PropagationofVoltageSags,IEEETransactionsonPowerDelivery,vol.21,no.1,pp.
262269,
http://ieeexplore.ieee.org/ielx5/61/33198/01564208.pdf?tp=&arnumber=1564208&is
number=33198,2006.
88
Appendix A IEEE 39Bus, 10Machine New England Test System
InthisappendixthedetailsofmodellingtheeHVnetworkbasedon[38]arepresented.
89
90
Appendix B Typical German SubTransmission 110 kV Ring
System
InthisappendixthedetailsofmodellingtheHVnetworkbasedon[40]arepresented.
91
Appendix C CIGR Medium Voltage Benchmark System
InthisappendixthedetailsofmodellingtheMVnetworkbasedon[47]arepresented.
Line Node Conductor R'ph X'ph B'ph R'0 X'0 B'0 l Installation
segment from- ID
t [/km] [/km] [S/m] [/km] [/km] [S/km] [km]
1 12 2 0.501 0.716 47.493 0.817 1.598 47.493 2.82 Underground
2 23 2 0.501 0.716 47.493 0.817 1.598 47.493 4.42 Underground
3 34 2 0.501 0.716 47.493 0.817 1.598 47.493 0.61 Underground
4 45 2 0.501 0.716 47.493 0.817 1.598 47.493 0.56 Underground
5 56 2 0.501 0.716 47.493 0.817 1.598 47.493 1.54 Underground
6 67 2 0.501 0.716 47.493 0.817 1.598 47.493 0.24 Underground
7 78 2 0.501 0.716 47.493 0.817 1.598 47.493 1.67 Underground
8 89 2 0.501 0.716 47.493 0.817 1.598 47.493 0.32 Underground
9 910 2 0.501 0.716 47.493 0.817 1.598 47.493 0.77 Underground
10 1011 2 0.501 0.716 47.493 0.817 1.598 47.493 0.33 Underground
11 114 2 0.501 0.716 47.493 0.817 1.598 47.493 0.49 Underground
12 38 2 0.501 0.716 47.493 0.817 1.598 47.493 1.30 Underground
13 1213 1 0.510 0.366 1.280 0.658 1.611 1.280 4.89 Overhead
14 1314 1 0.510 0.366 1.280 0.658 1.611 1.280 2.99 Overhead
15 148 1 0.510 0.366 1.280 0.658 1.611 1.280 2.00 Overhead
TableC.1:LineparametersofMVnetwork
92
Appendix D CIGR Low Voltage Benchmark System
InthisappendixthedetailsofmodellingtheLVnetworkbasedon[47]arepresented.
93
94
Conductor PhaseImpedancematrixafterKronreduction[/km]
ID/Installation A B C
UG1/3ph A 0.287+j0.167 0.121+j0.110 0.125+j0.070
B 0.121+j0.110 0.279+j0.203 0.121+j0.110
C 0.125+j0.070 0.121+j0.110 0.287+j0.167
UG2/3ph A 0.455+j0.204 0.185+j0.146 0.190+j0.107
B 0.185+j0.146 0.444+j0.238 0.185+j0.146
C 0.190+j0.107 0.185+j0.146 0.455+j0.204
UG3/3ph A 1.152+j0.458 0.321+j0.390 0.330+j0.359
B 0.321+j0.390 1.134+j0.477 0.321+j0.390
C 0.330+j0.359 0.321+j0.390 1.152+j0.458
OH1/3ph A 0.616+j0.588 0.131+j0.306 0.141+j0.245
B 0.131+j0.306 0.628+j0.566 0.147+j0.276
C 0.141+j0.245 0.147+j0.276 0.650+j0.527
OH2/3ph A 1.457+j0.728 0.143+j0.417 0.152+j0.367
B 0.143+j0.417 1.469+j0.720 0.159+j0.405
C 0.152+j0.367 0.159+j0.405 1.490+j0.704
OH3/3ph A 2.137+j0.776 0.125+j0.453 0.133+j0.406
B 0.125+j0.453 2.146+j0.771 0.138+j0.447
C 0.133+j0.406 0.138+j0.447 2.163+j0.762
TableD.6:PhaseImpedancematrixafterKronreductionofallthreephaselinesintheLVnetwork
ThetapchangerpositionsoftheHVtoMVandMVtoLVtransformersareshown
inthetablebelowforallloadcases.
96
Appendix F Results
Theresultsforallcasesarepresentedinthefollowingondoublepages.Thestructureof
theplotsisdescribedinthefollowingtwopages.Thesamestructureisusedforallcases.
97
Thevoltageatthevariousbusbars(inp.u.):
MVbusbar12 LVnodeR15
(connectedtoMV
busbar10)
TableF.1:Schematicrepresentationofvoltageresults
MV busbars 08 and 09 are interesting because the WPP of the MV network are
connectedthere.AlsoMVbusbars01,10areimportantbecauseLVnetworksareconnected
onthosebusbars.MVbusbars06,12aretheconnectionpointsofPVsintheMVnetwork.
FinallyalltheLVnodesandbusbarsareconnectionpointsofPVs.
Thepowerexchangeoverthetransformersgraphs:
98
ThepoweroutputoftheSGandDGgraphs:
ThedirectandquadratureaxiscurrentsofPVsystems:
99
Base case NoLVRT
Voltages
Powerexchangeoverthetransformers
100
PowerproductionofDGandSG
DirectandquadratureaxiscurrentsofPVsystems
101
Base case ZPM
Voltages
Powerexchangeoverthetransformers
102
PowerproductionofDGandSG
DirectandquadratureaxiscurrentsofPVsystems
103
Base case aRCI
Voltages
Powerexchangeoverthetransformers
104
PowerproductionofDGandSG
DirectandquadratureaxiscurrentsofPVsystems
105
Base case aRACI
Voltages
Powerexchangeoverthetransformers
106
PowerproductionofDGandSG
DirectandquadratureaxiscurrentsofPVsystems
107
Base case ZPM with delayed APR (3 s period)
Voltages
Powerexchangeoverthetransformers
108
PowerproductionofDGandSG
DirectandquadratureaxiscurrentsofPVsystems
109
Base case aRCI with delayed APR (3 s period)
Voltages
Powerexchangeoverthetransformers
110
PowerproductionofDGandSG
DirectandquadratureaxiscurrentsofPVsystems
111
Base case aRACI with delayed APR (3 s period)
Voltages
Powerexchangeoverthetransformers
112
PowerproductionofDGandSG
DirectandquadratureaxiscurrentsofPVsystems
113
sRPF NoLVRT
Voltages
Powerexchangeoverthetransformers
114
PowerproductionofDGandSG
DirectandquadratureaxiscurrentsofPVsystems
115
sRPF ZPM
Voltages
Powerexchangeoverthetransformers
116
PowerproductionofDGandSG
DirectandquadratureaxiscurrentsofPVsystems
117
sRPF aRCI
Voltages
Powerexchangeoverthetransformers
118
PowerproductionofDGandSG
DirectandquadratureaxiscurrentsofPVsystems
119
sRPF aRACI
Voltages
Powerexchangeoverthetransformers
120
PowerproductionofDGandSG
DirectandquadratureaxiscurrentsofPVsystems
121
sRPF ZPM with delayed APR (3 s period)
Voltages
Powerexchangeoverthetransformers
122
PowerproductionofDGandSG
DirectandquadratureaxiscurrentsofPVsystems
123
sRPF aRCI with delayed APR (3 s period)
Voltages
Powerexchangeoverthetransformers
124
PowerproductionofDGandSG
DirectandquadratureaxiscurrentsofPVsystems
125
sRPF aRACI with delayed APR (3 s period)
Voltages
Powerexchangeoverthetransformers
126
PowerproductionofDGandSG
DirectandquadratureaxiscurrentsofPVsystems
127
fRPF NoLVRT (no convergence after 0.9 s)
Voltages
Powerexchangeoverthetransformers
128
PowerproductionofDGandSG
DirectandquadratureaxiscurrentsofPVsystems
129
fRPF ZPM
Voltages
Powerexchangeoverthetransformers
130
PowerproductionofDGandSG
DirectandquadratureaxiscurrentsofPVsystems
131
fRPF aRCI
Voltages
Powerexchangeoverthetransformers
132
PowerproductionofDGandSG
DirectandquadratureaxiscurrentsofPVsystems
133
fRPF aRACI
Voltages
Powerexchangeoverthetransformers
134
PowerproductionofDGandSG
DirectandquadratureaxiscurrentsofPVsystems
135
fRPF ZPM with delayed APR (3 s period, no convergence after 0.9 s)
Voltages
Powerexchangeoverthetransformers
136
PowerproductionofDGandSG
DirectandquadratureaxiscurrentsofPVsystems
137
fRPF aRCI with delayed APR (3 s period)
Voltages
Powerexchangeoverthetransformers
138
PowerproductionofDGandSG
DirectandquadratureaxiscurrentsofPVsystems
139
fRPF aRACI with delayed APR (3 s period)
Voltages
Powerexchangeoverthetransformers
140
PowerproductionofDGandSG
DirectandquadratureaxiscurrentsofPVsystems
141
Appendix G PowerFactory Lessons Learned
ScalingthePVTemplate
ThroughDSLParameterstocreatetemplatesofdifferentrating
1. ChangesinNetworkParameters
PWMConverter
BasicData/RatedPower
Multiplythe0.0112ofthe10kWbythetimesofscaling
LoadFlow/ReactivePowersetpoint
Usethereactivepowercapabilitylimit(negative)witha
capabilitylimitof0.4324107p.u.
LoadFlow/ReactivePowersetpoint
Usethedesiredvalue
Currentsource
NominalCurrent
Divide active power set point by the product of the dc
voltagewiththemaximummodulecurrentandthenumber
ofparallelmodules.
Upgradelinesandtransformersaccordinglyinsidethetemplate
PQmeasurementblockinsideCompositemodel
Ratingaccordingtouser
1MVAfor10kWandproportionally
2. ChangesinDSLparametersinsidetheDSLModels
DSLPVmodule
Scalenumberofparallelmodulesn tomatchP ,
BasecurrentP ,
, , ,
with P , :BasecurrentofPVsystem
I , : BasecurrentofDCcurrentsource
I , : CurrentofonePVpanel(determinedbyotherPVmodule
parameters)
142
Throughparallelmachinestoupscaletheinstanceofthetemplate
1. ChangesinNetworkParameters
PWMConverterLinesTransformers
ParallelMachines/ParallelLines/ParallelTransformers
Allthiselementsinsidethetemplatetakethemultiplier
inthecorrespondingfield.
CurrentsourceCapacitor
NominalCurrent/Capacitance
Multiplybythenumberofparallelmachines.
PQmeasurementblockinsideCompositemodel
Ratingaccordingtouser
Multiplytheratingbynumberofparallelmachines.
143
Datamanagement
Itishighlyrecommendedtouseaconsistentnamingconventionforallrelevant
elementsrightfromthebeginningoftheprojectsetup.
Thefollowingnamingconventionseemstobeideal:
#eHVnode_#HVnode_#MVnode_#LVnode_#Type_#DATE/#MODE_#COUNT
where
o #eHVnode_#HVnode_#MVnode_#LVnode_indicatestowhichnodein
which(distribution)systemtheelementisconnectedto(nonapplicable
voltagelevelswillbereplacedbytwounderscores(insteadof2digit
number));thisisofspecialimportanceifallvoltagelevelsaremodelled;
exampleeHV08_HV02_MV01_LVR01_TERM_01means
LVterminal01of
LVnetworkconnectedtoMVterminal01of
MVnetworkconnectedtoHVterminal02of
HVnetworkconnectedtoeHVterminal08of
thetransmissionsystem
o Foredgeelementsbothconnectionpointsareincludedinthename.
ExamplewithaMVline:eHV08_HV02_MV0102_LINE
ThismeansthelineisconnectedbetweenMVnodes01and02
o Sincetransformersconnecttwodifferentvoltagelevels,theyhavethe
connectionpointsincludingvoltagelevelintheirname,example:eHV
08_HV02_MV02_T01LV01_T01_TRF
Thismeansthetransformerisconnectedbetweenthe(first)
Terminalatnode02intheMVsystemandthe(first)Terminalat
node01inthecorrespondingLVsystem
o #Type=PVPPM/WPPM(WPPM_DFIG/WPPM_FC)/SGCHP/SGCONV/
STATLD(staticload)/DYNLD(dynamicload)/LINE/TERM/SWITCH/TRF
o #DATE=OLD/NEW/specificyear(ifdesired)
o #MODE=PF100/PF095/PFPOW/LVRT/aRCI(foroldgenerators,this
indicatestheirsettings(PF100/095/POWgivesthepowerfactorsettings
constantorbasedontheactivepower)
o #COUNT=2digitnumberforduplicates
SteadystateShortcircuitcalculations
Shortcircuitduration(breaktime)seemsnottomakeanydifferenceinresults.
S/CcalculationfailstoprovideSHCimpedances(X,R,Z)atFCWTGterminals.
S/CcalculationresultsintohigherSkvaluesatallterminalsifFCWTGisputoutof
operation.
Completemethod
o superpositionstheprefaultvaluesbutisnotastandardisedmethod.
o Doesnotcalculateik
Networkimpedanceangleiscalledphiz
144
ParticularitiesofStaticgenerator(ElmGenstat)
ThesignconventionforPandQofastaticgeneratorisunintuitive.
VDEARN4105usesthepassive(consumeroriented)signconvention(P<0for
generation)asfollows:
Effect SignofQ SignofIq Operation Reactivepower
(P<0) (Id<0) exchangewith
grid
Voltage Q>0(pos.) Iq< 0 Under Inductive
decreases (neg.) excited
Voltage Q<0(neg.) Iq>0 Overexcited capacitive
increase (pos.)
TableG.1:Passivesignconvention
ElmGenstatusestheactive(generatororiented)signconvention(P>0for
generation)asfollows:
Effect SignofQ SignofIq Operation Reactivepower
(P>0) (Id>0) exchangewith
grid
Voltage Q<0(neg.) Iq>0 ./. capacitive
decreases (pos.)
Voltage Q>0(pos.) Iq<0 ./. Inductive
increase (neg.)
TableG.2:Activesignconvention
Substationparticularities
Thesubstationhasasecondvoltagesettingnexttotheoneonthebasicdatatab.
TheadditionalvoltagesettingisnamedNominalVoltage(Unom)andcanbefound
byrightclickingthesubstationinthedatamanager(bygoingintothegrid)andthen
Edit. The nominal voltage can be changed by clicking the Set Nominal Voltage
button.N.B.Ifatransformerisconnectedtothesubstationthisisthevoltagethat
thetransformersees.
Projectcollaboration
Thefollowingfunctioncanbeusedtocopyallexternalreferencesofaprojectinto
it.Pleasebeawarethatthisfunctionalityisstillundertestsothereisapossibilityof
unexpectedresults.Therefore,remembertoexporttheprojectfirstasabackup.
1. Deactivatetheproject.
2. Focus onthe project bymaking sure it is selected in the left hand paneof
thedatamanager.
3. press the input window icon on the data manager and type the following
into the command line which appears: _test/cpex (from copy external
references)
4. Runthecommandthatappears.Note,thattheunderscoreisimportant
145
You should find a reference folder is included in theproject which contains all
the external references and you should find that your project's objects link to the
newfolder.
Source:DIgSILENTSupportansweronAugust15,2013
Userfriendliness
Onceanelementwindowisopen,youcanselectotherelementsofthesameclass
andshowtheirparametersinthepreviouslyopenedwindows.
HotlineQ&A
WhatistherelevanceofLevelisDSLmodels?
Thelevelofthemodelrepresentationisonlyimportantwhenusingorchanging
old models. For new created models the highest level should always be used. For
macros,thisoptiondoesnothaveanyimpact,becausethelevelofthehighestblock
isimportant,i.e.thecontrollerdefinition.Thenassummarythelevelisrelatedwith
theversiononwhichthemodelwascreated,thenwhenthemodelisusedinnewer
versions PF recognized and create the compatibility; the higher the level, the
newertheversionwithwhichwascreated.Ourrecommendationinthisregardisdo
notchangethelevelsettingofamodel,moreevenifthemodelistakenfromother
versionlibraryoritwasamodelbuiltbysomeoneelse.
Howtochangethenumberofticks(major/minor)inVisPlots?
Thereisnoofficialwaytochangethestepsizeoftheaxisofaplot.Butthereisa
trickthatmaybehelpfulalthoughitsresultswillbeappliedtothecompletevirtual
instrumentpanelandnotonlyforoneplot.
FirstyouhavetocreateanewstylebyusingthefunctionStyle>createnew
styleintherightmousebuttonmenu.
ThenewstylecouldbefoundintheprojectfolderSettings>Styles.
Rightclickintheplotandselectstyle>>editstyle.ThiswillforcePowerFactory
tocreatetheaxisobjectsthatyoueditinthedatamanagerinthenextstep.
Find the style in the data manager: you should find a folder in the project:
Settings>Styles>Yourstyle>VisPlot.TheVisPlotfolderincludestheaxisobjects.
If you display the flexible data of the axis by pressing the 'detail mode class
select' icon and select axis objects, you should find a variable "Number of Main
Ticks".Thisvariablecannotbechangeddirectlybecauseitisgreyedout.
Ifyouopenthe"InputWindow"(iconinthedatamanager)youcouldenterthe
command"xAxis:mTicks=N"(whereNisthenumberofmajorticksrequired).
Flagastablealwaysappliesonlowerlevelmodelsinside?
TheoptionAstablealgorithmenabledatadvancedoptionspageofCalculation
ofInitialConditionsmenuappliesforallmodelsused.
146
WhatdoesitmeaninOPF(optimalpowerflow)Whenthemodelisunbounded?
(Howtofindoutwhattodebug?)
This message appearsonly in the DC OPF if theLP has no unique solution (i.e.
infinitevalidsolutions)andthetargetfunctioncanhaveunconstrainedhighvalues.
Thiscanbethecaseifyoudefinednotenoughlimits.
Example:Therearetwogenerators,acheapandanexpensive.TheOPFshould
minimize the cost. The LP will increase the cheap and decrease the expensive. If
thereisnolimitfortheactivepowerthecheapwillgotoinfinityandtheexpensive
toinfinity.Youneedmoregeneratorsand/orloadconstrains.
WhatreasonscantherebethataDSLmodelcannotbeedited?
WhatisaIntVariantelementforDPLfor?(Whatisthedifferencebetweenvariants
andvariations?)
The function IntVariant is used in older versions of the software, where the
changesinthesystemwererecordedina"SystemStage",howevernowthechanges
are stored in Variations and Operations Scenarios depending of the type of
modification.
This change was done in 14.0, so, since you have version 14.1, please ignore the
IntVariantmethods.
147