RELIABILITYIMPROVEMENTOFDFIG-BASEDWINDENERGYCONVERSION SYSTEMSBYREALTIMECONTROL By LinaAdnanAbdullahElhmoud ADISSERTATION Submittedto MichiganStateUniversity inpartialentoftherequirements forthedegreeof ElectricalEngineering-DoctorofPhilosophy 2015 ABSTRACT RELIABILITYIMPROVEMENTOFDFIG-BASEDWINDENERGY CONVERSIONSYSTEMSBYREALTIMECONTROL By LinaAdnanAbdullahElhmoud Reliabilityistheprobabilitythatasystemorcomponentwillsatisfactorilyperformits intendedfunctionundergivenoperatingconditions.Theaveragetimeofsatisfactoryoper- ationofasystemiscalledthemeantimebetweenfailures(MTBF)and,thehighervalueof MTBFindicateshigherreliabilityandviceversa.Nowadays,reliabilityisofgreaterconcern thaninthepastespeciallyforwindturbinessincetheaccesstotheseinstallations incaseoffailuresisbothcostlyandPowersemiconductordevicesareoftenranked asthemostvulnerablecomponentsfromreliabilityperspectiveinapowerconversionsys- tem.Thelifetimepredictionofpowermodulesbasedonmissionisanimportant issue.Furthermore,lifetimemodelingoffuturelargewindturbinesisneededinorderto makereliabilitypredictionsintheearlydesignphase.Byconductingreliabilitypredictionin thedesignphaseamanufacturecanensurethatthenewwindturbineswilloperatewithin designedreliabilitymetricssuchaslifetime. Thisworkpresentsreliabilityanalysisofpowerelectronicconvertersforwindenergy conversionsystems(WECS)basedonsemiconductorpowerlosses.Arealtimecontrol schemeisproposedtomaximizethesystem'slifetimeandtheaccumulatedenergyproduced overthelifetime.Ithasbeenvthroughthereliabilitymodelthatalo basedcontrolcanelyincreasetheMTBFandlifetimeofthepowermodules.The fundamentalcausetoachievehigherMTBFliesinthereductionofthenumberofthermal cycles. Thekeyelementinapowerconversionsystemisthepowersemiconductordevice,which operatesasapowerswitch.Theimprovementinpowersemiconductordevicesisthecritical drivingforcebehindtheimprovedperformance,,reducedsizeandweightofpower conversionsystems.Asthepowerdensityandswitchingfrequencyincrease,thermalanal- ysisofpowerelectronicsystembecomesimperative.Theanalysisprovidesinformationon semiconductordevicerating,reliability,andlifetimecalculation. Thepowerthroughputofthestate-of-the-artWECSthatisequippedwithmaximum powerpointcontrolalgorithmsissubjectedtowindspeedwhichmaycause tthermalcyclingoftheIGBTinpowerconverterandinturnleadtoreductionin lifetime.Toaddressthisreliabilityissue,areal-timecontrolschemebasedonthereliability modelofthesystemisproposed.Inthisworkadoublyfedinductiongeneratorisutilizedas ademonstrationsystemtoprovetheenessoftheproposedmethod.Averagemodel ofthree-phaseconverterhasbeenadoptedforthermalmodelingandlifetimeestimation. Alobasedcontrollawisutilizedtomodifythepowercommandfromconven- tionalWECScontroloutput.Theresultantreliabilityperformanceofthesystemhasbeen tlyimprovedasevidencedbythesimulationresults. ACKNOWLEDGMENTS MyLord,increasemeinknowledge.Quran(20:114) "OAllah,Iaskyouforknowledgewhichisbandsustenancewhichisgood,and deedswhichareacceptable."hadith-ThegreatestprophetMuhammad Alhamdulilah .ThisthesishasbeencompletedwiththeblessingoftheMercifuland Almight,AllahS.W.T.Ipraisehimforprovidingmeopportunityandgrantingmethe capabilitytoproceedsuccessfully. Atthismomentofaccomplishment,ofallIpayhomagetomysupervisorProf. BingsenWang,whogivesmehopewhenIthoughthopewasgone.Mysupervisorhas constantlyforcedmetoremainfocusedonachievingmygoals.Hisobservation,comments, immenseknowledge,invaluableideashelpedmetoestablishtheoveralldirectionofthe researchandtomoveforwardwithinvestigationindepth.Ihavelearnedagreatdealfrom hisuniqueperspectiveonresearchandhisexpectationsofexcellence.Thisworkwouldnot havebeenpossiblewithouthim.Itisagreathonortoworkunderhissupervision.Iowe himsomuch. IexpressmysinceregratitudetomyPhDcommitteemembers,especiallyprof.Hassan Khalil,forhiskindhelp,generousadviceandsupport.MysincereappreciationistoProf. GuomingZhoandProf.EliasStrangasfortheirhelpfulcommentsandsuggestionsduring myprogressthesispresentation. Lifewouldnothavebeenascolorfulwithoutmanygoodpeoplewhohavehelpedme duringmystudyinMSU,especially,FamilyResourceCenter(FRC),forInternational StudentsandScholars(OISS)andElectricalandComputerEngineeringDepartment MysincerethankstoallmyfriendswhoIcannotmentiontheirnamesinthislimitedspace. iv Wordsfailmetoexpressmyappreciationtomybelovedfamilyfortheirencouragement, unconditionallysupport,andgenerouscare.Theyarealwaysbesidemeduringthehappy andhardmomentstopushme,motivatemeandliftingmeuphillthisphaseoflife. Toallofthese,Igratefullyacknowledgemydeepindebtedness. v TABLEOFCONTENTS LISTOFTABLES .................................... ix LISTOFFIGURES ................................... x Chapter1Introduction ................................ 1 1.1Background....................................1 1.1.1Whywindischosenforstudy?......................1 1.1.2WhyReliability?.............................4 1.1.3WhyWindReliability?..........................5 1.1.4WhyWindPlantReliabilityModeling?.................6 1.2LiteratureReview.................................7 1.3MotivationoftheWork..............................8 1.4ProblemStatement................................11 1.5ScopeoftheWork................................12 1.6OrganizationofTheThesis...........................12 Chapter2AnalysisandMethodsinWindReliability ............. 14 2.1Introduction....................................14 2.2ReliabilityPrediction...............................14 2.2.1EmpiricalPredictionApproach.....................18 2.2.2PhysicsofFailurePredictionTechniques................20 2.2.3Test/FieldData..............................21 2.2.4SystemReliabilityAssessmentPrediction................21 2.2.5SimilarItem/CircuitPrediction.....................22 2.2.6PredictionbyOperationalTranslation.................22 2.3FaultTreeAnalysis(FTA)............................23 2.4ReliabilityBlockDiagram(RBD)........................24 2.5FailureModesandAnalysis(FMEA)..................25 2.6TheAnalyticalMethods.............................26 2.7SimulationMethods................................27 Chapter3ReliabilityatWindEnergyConversionLevel ........... 30 3.1ModelingWindTurbineReliability.......................30 3.2LifeCurve.....................................35 3.2.1TheShapeParameter .........................36 3.2.1.1Earlyfailure < 1.......................36 3.2.1.2Constantfailurerate =1..................36 3.2.1.3Deterioration > 1......................37 3.3ElectricalandElectronicComponentsReliability...............39 vi 3.3.1ElectricalComponents..........................39 3.3.2ElectronicComponents..........................40 3.4ReliabilityDesigninWECS...........................44 3.4.1DesignforElectricalReliability.....................47 3.4.2DesignforMechanicalReliability....................49 3.4.3DesignforPowerElectronicReliability.................52 3.5Severitys..............................54 Chapter4ReliabilityatWindFarmLevel .................... 56 4.1Introduction....................................56 4.2ReliabilityModelingofWindTurbine......................60 4.2.1WindSource...............................60 4.2.2WindTurbineCharacteristics......................62 4.3ReliabilityCharacteristics............................63 4.4ReliabilityIndices.................................67 4.5FactorsforWindFarmReliabilityAssessment.................69 4.6WindFarm...............................70 4.7TechnicalAspectsofIntegratingWindFarmsintoPowerSystems......72 4.7.1ActivePowerControl...........................73 4.7.2ReactivePowerControl.........................74 4.7.3VoltageFlickers..............................74 4.7.4Harmonics.................................75 4.8WindFarmTopologies..............................76 4.8.1ACTopologies...............................76 4.8.2DCTopologies..............................78 4.9WindFarmLosses................................79 4.10Challenges.....................................80 4.11OperationandMaintenancePlanning......................81 Chapter5Real-TimeOptimizationofThermalCyclingCapabilityofRo- torSideConverterinDFIG-BasedWECS ............. 84 5.1Introduction....................................84 5.2PhysicalSystemModeling............................85 5.2.1WindTurbineCharacteristics......................85 5.2.2Doubly-FedInductionGenerator....................87 5.2.3AveragedModelof (PWM)converter ..................91 5.3ElectrothermalModelingandLifetimeEstimationoftheVoltageSourceCon- verterforWindTurbine.............................92 5.3.1PowerLossesofIGBTintheRSC....................93 5.3.2ThermalModelingTechnique......................96 5.4LifetimePredictionandDesignofReliability..................100 5.4.1wCycleCounting.........................102 5.4.2LifetimeModeling............................104 5.5PrinciplesofFilteringWindTurbinePowerCommandFluctuations.....106 vii Chapter6ConclusionsandFutureworks ..................... 115 6.1Conclusions....................................115 6.2FutureWork....................................116 BIBLIOGRAPHY ................................... 118 viii LISTOFTABLES Table2.1Comparisonoftreliabilitypredictionmethodologiesforelec- tronics...................................16 Table2.2Timeoperationperiodforreliabilitypredictionmethodologies....17 Table3.1SeverityofFailuresModes........................55 Table5.1DFIGElectricalParameters......................91 Table5.2IGBTthermalcharacteristicvalues...................96 Table5.3ThermalElectricalAnalogousQuantities...............98 Table5.4Theenessofthenewstrategy..................110 ix LISTOFFIGURES Figure3.1BathtubeCurve.............................35 Figure4.1WindFarmBlockDiagram........................59 Figure4.2Shadowkerevidencebase.......................75 Figure4.3ACredialsystem.[1]..........................77 Figure4.4ACredialloopsystem.[2].......................77 Figure4.5ACstarsystem.[2]...........................78 Figure5.1Systemunderstudy.[3][4].......................88 Figure5.2VariablesofthreephaseDFIGinstationary,synchronousandrotor referenceframes..............................90 Figure5.3Lifetimeestimationmodelforpowersemiconductordevices.[5]...93 Figure5.4FosterThermalImpedanceBetweentheJunctionTemperatureand CaseLayer.................................97 Figure5.5CauerThermalImpedanceBetweentheJunctionTemperatureand CaseLayer.................................98 Figure5.6StressStraincycles............................104 Figure5.7ofthewindspeedusedinthissimulation............107 Figure5.8PlotsofthepowercommandswithandwithouttheLPF.......109 Figure5.9PlotofthedcbusvoltagethatstaysthesamewithorwithoutLPF.110 Figure5.10RotorwindingterminalvoltagewithoutLPF..............111 Figure5.11LPFresponsecharacteristicusingIIRimpulseresponse,minimum ordermode,singleratetype,Butterworthalgorithm..........111 x Figure5.12IGBTjunctiontemperaturevariationswithoutLPF..........112 Figure5.13IGBTjunctiontemperaturevariationswithLPF............112 Figure5.14Temperaturemeanvalue T m extractedfromwcountingalgo- rithmwithoutLPF............................113 Figure5.15Amplitude, T extractedfromwcountingalgorithmwithout LPF....................................113 Figure5.16Frequencydistributionoftemperaturecyclesbytheirampli- tude T andtemperaturemeanvalue T m extractedfromw countingalgorithmwithoutLPF....................114 xi Abbreviations v s , v r RMSvoltagesforstatorandrotor i s , i r RMScurrentsforstatorandrotor s Slip s , r Fluxlinkagesofstatorandrotor m 1 , m 2 Modulationfunctionsofstator-androtor-sideconverters v a , v b , v c 3-phasesupplyvoltages v d , v q , v , v Supplyvoltagecomponentsindq-and -referenceframes ! e , ! r , ! slip Supply,rotor,andslipangularfrequencies P , Q Activeandreactivepowers e , s Phaseanglesofsupplyvoltagevectorandstatorvector C DClinkcapacitance V dc DClinkvoltage L s , L r Per-phaseinductancesofstatorandrotorwindings L ls , L lr Per-phaseleakageinductancesofstatorandrotorwindings L m Magnetizinginductance R s , R r Per-phaseresistancesofstatorandrotorwindings L , R Inductanceandresistanceofsupply p Numberofpolepairs J Inertiaofmachinewindturbinerotors T e Electromagnetictorque xii Chapter1 Introduction 1.1Background 1.1.1Whywindischosenforstudy? Windgenerationisoneofthemostsuccessfulformsofenergyproductionfromrenewable sourcesintermsofaccumulativeinstalledcapacity.Asthenumberofgridconnectedin- stallationsgrowrapidlyworldwide,thereisaneedtostudythereliabilityoftheseenergy conversionsystemsandfurthertoassesstheirimpactontheoverallsystem. Theimpactcanpotentiallyembodyinmultipleaspects. Environmentalissuesandregulations: Windenergygenerationsystems,incontrasttofossil-fuel-basedsystems,donotpro- ducegreenhousegas(GHG)emissionsthatadverselyclimatechange.Further- more,UNSecretaryGeneralofSustainableEnergystatesthattheworldrecentlypassed 400partspermillionofatmosphericCO 2 ,whichisadequatetopotentiallystimulate warmingof2 Ccomparedwithpre-industrialera[6].Thus,environmentalfriendly windgenerationsystemssquarelyaddresstheregulatoryandpracticalenvironmental concerns. 1 Economicaspects: Investmentinrenewableenergyproductionwillleadtojobcreationinthisrelatively newindustry[6].Inaddition,electricalenergyisthekeytodevelopmentofamodern economy.Itishasacceptablerangeof?levelizedcostofproduction?,whichdepends oncapitalcost,operatingcostsandfuelcosts.Remoteareasthatarenotconnectedto electricitypowergridcanusestand-aloneturbinestoavoidhighcostassociatedwith theinfrastructureoftransmissionlines. Technicalmaturity: Windenergyiswidelyavailableandaccessible.Thewindenergyhasbeenprovenan economicallyviablealternativetofossil-fuelbasedenergy. Smallfootprint: Windsystemproduceslow-levelnoisesandnowasteproduct.Thesmallfootprintof windsystemiscompatiblewithmanylandusesorsmallplotofland,whichmeansthe landbelowcanstillbeutilized.Thisisespeciallyincaseofagricultureareaasfarming canstillcontinue.Nowadaysmoreattentiontowardsmarinewindfarms. Energysafetyandsecurity: Renewableelectricitygenerationmakestheoverallelectricitygenerationsystemless reliantoncoalandnaturalgasandthuslessvulnerabletovolatilityindomesticand globalfuelmarkets. Deferringoffossilfuelproduction: Themoredevelopedskillsindingalternativeenergyresourcessuchaswindenergy decreasedthelevelofinterestofsuperpowersintheoil-producingcountries. 2 Industrialpollution: Windplanthaslessindustrialmishapsthathavetobebroughtundercontrol,un- likeconventionalgenerationplantproducemultipleformsofindustrialpollutionthat includecontaminationofdrinkingwater,airandsoil. Fortheaforementionedreasons,windgenerationsystemtechnologyisoneofthemost promisingrenewableenergytechnologies.Nonetheless,thefastexpansionofthewindpower facessomechallengesthatrequirefocusedresearchattention.Theresearchareasthatad- dressthesechallengesincludewindfarmmodelingforreliabilitystudiesandapplicationof thereliabilityassessmenttechniqueswithmissionanddynamicmodelbroughtinto consideration.Windplants,whichareunlikecoalornaturalgaspowerplants,cannotbe deterministicallyscheduledtodeliverspamountsofpoweratsptimesduesto thestochasticnatureofavailablewindpower.Windpowerplantsgenerateelectricitywhen energyresourceisavailable.Manyelectricitysystemoperatorsseethisvariabilityasbluster tosystemstabilityandreliability.Therearethreefundamentalsolutionstothevariability challenge[7]. Increasingtheilityofelectricitysupplyoptions: Thissolutioninvolvesconstructingwindfarmsthatcanrapidlyadjusttheiroutputby increasinginstalledcapacity.Forinstance,Germanyvelyincreasesitsinstalled capacitythroughcontractualtradingofelectricitywiththeneighboringcountryDen- mark. Demandsidemanagement: Demandsidemanagementemployspricingandotherincentivetoolstoor 3 controlthedemandforelectricity.Increaseddemandycanleadtoreduced peakloadandimprovedcapacityfactorofthesystem. Energystorage: Usingphysicalstorageofelectricityto?smooth?theoutputofvariableelectricity sources.Thereareseveralphysicalstoragetechnologiesunderdevelopment.Energy storagetechnologiesthatcanquicklydeliverenergyincludeenergystorage, batteries,andsuper-capacitors. Themostseriousimplementationbarriersforincreasingwindgenerationarehighcosts associatedwithconstructingnewtransmissionlinesthattypicallycosttwotofourmillion dollarspermile[7],andtheyassociatedwithsitingandpermittingprocesses.More- over,themainenvironmentaldisadvantagesareerosion,movingshadows,interferencewith electromagneticcommunications,impactsonbirds,unsightlystructuresandsomepollution producedduringthemanufacturingprocess. 1.1.2WhyReliability? Sinceitappearedin1800 s ,reliabilityhasbeenafundamentalattributetothesafe operationofanymoderntechnologicalsystem.Thetermreliabilitywascoinedby theEnglishpoetSamuelT.Coleridge,whoalongwithWilliamWordsworthstartedthe EnglishRomanticMovement[8].Reliabilityofasystemistheprobabilitythatthesystem willperformitsintendedtasks.Thisprobabilityisusuallydeterminedasapercentageof time[9].Aprincipalobjectiveofreliabilityanalysisistogainfeedbackforimprovingdesign. Reliabilitystudyofsystemsallowsforoptimizingthemaintenancestrategyinordertoreduce cost. 4 1.1.3WhyWindReliability? Developmentofwindpowergenerationisbltoadjustthestructureofenergy,reduce theenvironmentalpollutionandpressureofenergyimportandexportandpromotethe economicdevelopment[10].Thatleadsustoshedstronglightononeofthemostimportant aspectsinwindenergy,reliability.Thereliabilitycanbeimprovedbychoosingproperdesign spandexercisingstrictcontrolonthemanufacturingprocessorbyusinggood qualitymaterials.Inaddition,preventivemaintenancetechniquesalsoplayanimportant roleinreliabilityimprovement.Theliteraturesurveysuggeststhatwindreliabilityisto increasethewindenergygrowthandcorrespondinglytodecreasethecostofwindenergy. Moreover,improvementofwindsystemreliabilitycanfurtherextendpenetrationlimitsand enhancethereliabilityoftheoverallpowersystem. Reliabilityisconsideredasthescienceoffailures[10]or?probabilityofsuccess?[11].Reli- abilityevaluationandenhancementisanimportantfactorinwindenergy.Consequently,the reliabilitymethodsandproceduresofwindsystemsareofgreatimportanceandwillreceive focusedattentioninthefuturewithincreasinggenerationfromwindresources.Reliability analysisofwindturbineswouldallowtoidentifyweaknessesinpartsandsubassemblies. Sensitivityanalysisbasedonreliabilityevaluationcouldindicatethespreadofunreliability amongpartsandsubassembliesinthewindturbineandarankingofcriticalsubassemblies couldbeachieved.Thereisgreatpotentialformorewindturbinestobeerectedinremote andlocationswhereagreaterwindenergyharvestcanbeachieved.Nonetheless, theaccesstotheseremotelylocatedturbinesformaintenancewillbelimited,whichneces- sitatesaccuratereliabilitypredictions.Reliabilitypredictionsforwindturbineswillhavean importantbearingonthefuturedevelopmentofwindpowerresources. 5 1.1.4WhyWindPlantReliabilityModeling? Thekeyissueindevelopinganysystemingeneral,andwindenergysysteminparticular, istorealizeandunderstandtherequirementsandpurpose.Inwindenergysystem,the requirementsshouldbeconsideredintwoaspects:componentsavailabilityandwindspeed's randomnessandvariability.Theobjectivesaretomaximizewindenergyproductionwhile minimizemaintenanceandreducecostwithoutcompromisingreliability.Inessence,the inquiryamountstowhatisimportantforwindmodeltoperformandwhatisnot. Windenergysystemscanbemodelledattwotlevels:windfarmlevelandwind energygenerationsystemlevel.Atwindfarmlevel,theoverallsystemreliabilityisfocused. Thus,therelationbetweenwindfarmreliabilityandpowersystemreliabilityisstudied. Powersystemreliabilityistoassesstheabilityofpowersystemprovidingenergytocustomers withtheacceptablequalityandquantitywithoutinterruptions.Powersystemreliabilitycan bedividedintotwobasiccategories:systemadequacyandsystemsecurity. SystemAdequacy relatestoexistenceoftfacilitieswithinthesystemto satisfytheloaddemandorsystemoperationalconstraints,consideringsystemcompo- nentsscheduledorunscheduledoutage.Adequacyisalsonamedstaticreliabilitysince itcategorizestheabilityofsystemprovidingenoughpowertocustomersunderstatic conditions.Thisconceptofadequacyconsidersastateincompleteisolationwithout takingtheactualentryordeparturetransitionsascauseofproblems. SystemSecurity meanstheabilityofpowersystemrespondingtodisturbances,such asshortcircuitfaults,orgeneratoroutages,arisingwithinthatsystem.Securityis alsoreferredasdynamicreliability,whichrepresentstheabilityofsystemtosupply powertocustomersunderdynamicconditionswithoutinterruptions. 6 1.2LiteratureReview Athoroughliteraturereviewsuggeststhattresearchhavebeenfocused onimprovingthereliabilityofwindenergysystems.Thesecanbecategorizedinto twolevels:windfarm(WF)levelandwindenergyconversionsystem(WECS)level.The combinationofthesetwolevelswillbethebuildingblocksforfuturewindenergysystem. Awindturbinesystemiscomposedofmanysubsystemswhichcoverthetopicsofelectrical andelectronicengineering,softwareengineeringandmechanicalengineering.Althoughthere areanumberofstudiesconsideringtheimpactofwindpoweronthereliabilityofalarge powersystem[12],[13],therehavebeenfewarticlesthatconsiderwindturbine[14].A numberofmethodsarenowavailableforthereliabilitypredictionofelectronicsystems andequipment?s[15,16,17].Theyincludephysicsoffailuremethod[18],[19],empirical methods[20],similaritemdatabasedmethod[17],testordatabasedmethod[17], andsystemreliabilityprediction[21],[22].Concerningpowersystemreliabilitytechniques, themodelsproposedintheliteratureareeithersimulativebasedonMonteCarlotechnique [23],oranalyticalbasedonMarkovmethod[24],[25].Thesemethodshaveadvantagesand drawbacksandcanbeverypowerfulwiththeproperapplication.Furthermoremodeling windfarmreliability[26,27,28,29],inadditiontothemodelingofwindturbinereliability [30,31,32],studiesfocusoneconomicbofreliablesystem[33],[34].Moreover,wind farmreliabilityissuesarediscussedin[35],[36],[37],[38].Commonlyadoptedreliability indicesareintroduced[39,40,41].Attentionshavealsobeendevotedtoimprovedpower electronicsystemsintermsofreliability[42,43,44,45]. Theresearchonlifetimepredictionhasbeenmainlyfocusedateitherdevicelevel[46][47], [48],orsystemlevel[49],[50],[3].Reliabilityofpowerelectroniccomponentsisakeyconcern 7 nowadays.Itisstronglydbytheoperatingtemperatureofthesecomponents[51], [52].Switchingfrequencyhasbeenmodthroughthecontroltodecreasethermalcycles [53].Theconvertertopologyhasbeendiscussedwidely.Hencemultileveltopologiestend tosharethestressamongdevicesandthestressoneachsingledevicedependsonthetotal numberofdevicesintheconverter[54][55].Furthermore,faulttolerantarchitectureshave beenproposedtoincreasethelifetimetly[56],[57],[58].Thecoolingsystem designhasplayedatroleinlifetimeoptimization[59],[60],[61].Inaddition conditionmonitoringhasbeenproventobeacostemeansofenhancingreliability andimprovingcustomerserviceinpowerequipment[62][63],[64].Thethermalperformance canbeimprovedbyinjectingproperreactivepowercirculationwithinthewindturbine system,therebythethermalcyclingcanbereducedandthereliabilityofthepowerconvert canbeenhanced[65],[66].Besides,advancedpowerelectronicconverterscanprovidethe meanstocontrolpowerwandensureproperandsecureoperationoffuturenetworksspace here[67].Aginghasbeeninvestigatedbyfocusontheagingofthermalinterfacematerials thataresubjectedtothermalcyclingconditions[68],[69].Theuseofdiscontinuouspulse widthmodulation(DPWM)iscanminimizelossesduetotheelyreducedswitching frequencyandconsequentlycanenhancesystemreliability[70],[71].Improvedpackaging technologyisneededtoimprovereliability[72].Thereforelifetimeschemescanbe asactivethermalcontrolandpassivethermalcontrol. 1.3MotivationoftheWork Animportantobservationabouttheresearchofthewindenergyreliabilityisthatitlags theresearchprogressofreliabilityinmanyotherindustries.Theintentionofthisworkis 8 toprovideanintroductiononthereliabilityofwindenergysystemsatbothwindfarmlevel andwindenergygenerationsystemlevelandreviewtherelatedresearchAlthough itsurelycannotcoveralltheconcepts,thisthesistriestoreachthislaudablegoal.Some importantquestionsthataretobeansweredinclude:Whydowedotheresearchonwind? Whydowedoresearchonreliabilityofwind?Whydoweneedpredictionmodels?What aretheavailablereliabilitymethodsandprocedureareusedinwindturbine?Why windenergy?Adetailedliteraturesurveyisconductedtoinvestigatethevariousavailable reliabilitymethods.Stateoftheartofwindfarmreliabilityisprovided.Anoverviewof reliabilityofpowerelectronicsinwindenergypublicationsisdiscussed.Akeymessageis thatthereexistsgreatspaceforimprovementalthoughalotofprogresshasalreadybeen achieved.Thepaceofactiveresearchfromacademicinstitutionsandwindindustrieswill increaseovertime. Reliabilityhasagreatroleinwindenergy.Manyideasandproposalsthatareonthe researcherstableswillprovideimportantcontextfortheproposedworkinthisthesis.Some importantissueshavebeeninvestigated.Thesurveyistoexplainwhywindreliabilityis studied.Alsothesurveyhasattemptedtoshowthatmoremustbeinvestedasthe predictionprocessbecomesmorecredibleandlessimprecise.Thesurveyhasalsodescribed theattemptswhichhavebeenmadetoimprovingwindfarmandwindenergygeneration systemsreliability.Thesurveystillleftwithmanyopenproblemsviz.gridrequirements, vibrationsensorscollectingdataofgeneratorhealth,windturbineconditionmonitoringand faultdiagnosistechniques.Despitethatwindturbineisoneofmarvelsamonggeneration plant,itisactuallyfallingfarshortintermsofrealtimeresearch. Reliabilityevaluationandenhancementisanimportantfactorinwindenergy.Conse- quently,thereliabilitymethodsandproceduresofwindsystemsareofgreatimportanceand 9 willreceivemoreattentioninthefuturewithincreasinggenerationfromwindenergy.Each reliabilitymethodhasitsownadvantagesanddrawbacks.Allmethodscontaincertainas- sumptionsthatmayormaynotbeThespapproachshouldbemadeavailable tlyearlytoncethereliabilitydesignandselectionofthedesignparametersfor thewindsystem.Theoflargewindfarmsinpowersystemoperationandplanning mustbehighlighted.Inalargewindpowerplant,anoversimmodeloftheplantasa singleturbineisgenerallyinadequate.Averylargewindpowerplantshouldberepresented bygroupsofwindturbinesrepresentedbytheiruniquecharacteristicswithrespecttothe location,thetypeofturbines,thecontrolsetting,andthelineimpedance.Itispreferred tousewindenergyinharmonywithotherformsofenergy.Powerelectronicsformodern windturbineshascapturedtheattentionofresearchersallovertheworld.Itplaysavery importantroleintheintegrationofwindenergysources.Finally,almostalloftheaspects relatedtothewindenergytechnologyarestillunderactiveresearchanddevelopmentsince therearemanyproblemsstill Furtherisrequiredtoimprovereliabilityofwindenergysystems,additionalin- formationshouldbesuppliedforselectinghighriskcomponentsthatneedbothresearch andindustrydevelopments.Upgradescienceandinnovativetechniquesshouldbeimple- mentedinwindenergygenerationsystems.Inadditionmaintenancetechniquesthathave beenprovenandimprovedinotherindustriesshouldbeestablishedinwindenergyindustry. Thequestionthatstillneedstobeinvestigatedistheexactreliabilitymechanismbehind suchobservedbehaviorinwindturbinefailures.Windturbinematerialsandmanufacturing techniques,processesforrepairoperations,collectingmonitoringdatamustbeprovedand improvedbasedonlessonslearnedfromtheandonalongsuccessfulhistoryinindustrial andcommercialapplications. 10 1.4ProblemStatement Reliabilityistheprobabilitythatacomponentwillsatisfactorilyperformitsintendedfunc- tionundergivenoperatingconditions.Theaveragetimeofsatisfactorilyoperationofa systemiscalledthemeantimebetweenfailures(MTBF)andthehighervalueofMTBF indicateshigherreliabilityandviceversa.Nowadays,reliabilityisofgreaterofconcernthan inthepastespeciallyforwindturbinessincetheaccesstowindturbinesin caseoffailuresisbothcostlyanddPowersemiconductordevicesareoftenranked asthemostvulnerablecomponentsinapowerconversionsystemintermsofreliability. ThelifetimepredictionofpowerIGBTmodulesbasedonmissionisanimportant issue.Furthermore,lifetimemodelingoffuturelargewindturbinesisneededinorderto makereliabilitypredictionsaboutthesenewwindturbinesintheearlydesignphase.By conductingreliabilitypredictioninthedesignphaseamanufacturecanensurethatthenew windturbineswilloperatewithindesignedreliabilitymetricssuchaslifetime. Thisworkpresentsreliabilityanalysisofpowerelectronicconvertersforwindenergy generationsystemsbasedonsemiconductorpowerlossesaswellasaimstomaximizesemi- conductorlifetimeusinglo(LPF)basedcontrolschemesinceMTBFwillbe higherthanwithoutThefundamentalcausetoachievehigherMTBFliesinthe reductionofthenumberofthermalcycles. Thekeyelementinapowerconversionsystemisthepowersemiconductordevice,which operatesasapowerswitch.Theimprovementinpowersemiconductordeviceisacritical drivingforcebehindtheimprovedperformance,,sizeandweightofpowercon- versionsystems.Asthepowerdensityandswitchingfrequencyincrease,thermalanalysis ofpowerelectronicsystembecomesimperative.Theanalysisprovidesinformationonsemi 11 conductorrating,reliability,andlifetimecalculation. 1.5ScopeoftheWork Thegoalistodiscussanoverallsurveyaboutreliabilityinwindenergyapplica- tions. ThesecondgoalistocalculatelifetimeanddesignthereliabilityforhighpowerIGBT's inwindpowerapplications. ThethirdgoalistooptimizethedynamicsystemlifetimeforIGBTmoduleinwind energyapplications. 1.6OrganizationofTheThesis Thisthesisisorganizedinthefollowingchapters. InChapter2,somemethodsandtechniquesarediscussedsuchasreliabilityprediction, whichinvolvesphysicsoffailurepredictiontechniques,test/data,systemreliabil- ityassessmentprediction,similaritem/circuitpredictionandpredictionbyoperational translation.Thesemethodsalsoincludefreetreeanalysis,reliabilityblockdiagram, failuremodesandanalysis,theanalyticalmethodsandsimulationmethods. InChapter3,issuesrelatedtoreliabilityanalysisonwindenergygenerationsystem levelarepresented.Moreover,modelingofwindturbinereliability,lifecurvearepre- sented.Reliabilitydesignofwindenergygenerationsystemincludesdesignforelectri- cal,mechanical,andpowerelectronicsubsystems. 12 InChapter4,windfarmreliabilityisstudiedandtheissuesareintroduced.Thetopics includereliabilitymodelingofwindturbines,reliabilityindicesandcharacteristics ofwindfarm,factorsreliabilityassessmentforwindfarm,of wind,technicalaspectsofintegratingwindfarmintopowersystem,windfarm topologies,windfarmlosses,challengesandoperationandmaintenanceplanning. InChapter5,concepts,mathematicalequations,implementationofwindturbinechar- acteristics,doublyfedinductiongenerator(DFIG)andtheaveragemodelforthecon- verterandthephysicalsystemarepresented.Semiconductorpowerlosses,junction temperatureanditsvariationthatdirectlythelifetimeoftheconverteraredis- cussed.Furthermore,lifetimecalculationsandreliabilityestimationarepresented. IssuesrelatedtoanadequatecontrolforsmoothingthepowercommandofDFIGis introduced. InChapter6,asummaryofthemaincontributionsoftheworkandtopicsoffuture workarepresented. 13 Chapter2 AnalysisandMethodsinWind Reliability 2.1Introduction Fortheanalysisofthereliabilityofwindenergysystems,manymethodsandtechniques eitherqualitativeorquantitative,numericaloranalyticalhavebeendeveloped.Thequal- itativetechniquesleadtoidenofweaknessesinthedesignpriortoquantitative approaches,whichareexecutedbyusingseveralsystemlevelfailuremodeandanal- ysis(FMEAs).Further,quantitativeapproachesareusedduringthedesignstage.Someof thesemethodsemployreliabilitypredictiontoolsoffaulttree(FT),reliabilityblockdiagram (RBD),reliabilitygraph(RG).AnalyticalmethodsarebasedonMarkovandsimulation methodsarebasedonMonteCarlo. 2.2ReliabilityPrediction Asthestepofreliabilityprediction,asetoffactuallypossibleconditionsandtheir relatedconsequencesmustbeconsideredbywhichthesystemcanbedesignatedasbeing failed.Then,convertingthequalitativeapproachintoaquantitativeapproachisasecond stage.Thisisobtainedbyprobabilitytheory,andtheassignmentofprobabilitiestoeach 14 ofthestatesthatleadthesystemtofailure.Sometimesthisrathercomplicatedprocedure canbepartiallycircumventedbyastatisticalanalysisofpreviousfailuredata,fromwhich directestimationoffailureprobabilitycanbemade.Thus,thepredictionproblemisnot merelytocharacterizethesysteminsomeway.Italsohastotakeintoconsiderationfurther factorssuchastheexperienceofthedesignteam,theconditionsunderwhichthedesignwas achieved. Reliabilitypredictionsforwindturbineswillhaveanimportantbearingonthefuture developmentofwindpowerresources.Thissectionisconcernedwithunderstandingthe reliabilitypredictionofmodernwindturbinesandpowerelectronicsubsystems.Thismethod isnotonlyapplicabletowindturbinesbutalsoapplicabletoanyrepairablesystem.The mainpurposeistodiscussthepracticalmethodsofpredictinglarge-wind-turbinereliability. Moreresearchwillprovidebothnewinsightsandspforwindturbinegeneration modeldevelopmentandapplication.Reliabilitypredictionisconsideredasaquantitative reliabilityanalysistechnique.Itisusedtopredictthefailurerateofasystembasedon itscomponentsandoperatingconditions.Thistechniqueisalsousedtoverifyprogressin reliabilityengineering. Thereliabilitypredictioncallsforbuildingamathematicalmodelforthesystemunder studyanddesomereliabilitymeasuressuchasexpectedenergynotsupplied(EENS), annualinterruptionfrequency,annualinterruptionduration,meantimetofailure(MTTF). Thenitisfollowedbydevelopingatechniqueforevaluatingthereliabilitymeasures,and comparingpredicteddataagainstexperimentalresults.Thecomparisonsoftmethod- ologiesaretabulatedinTable2.1. Thereliabilitypredictionofelectronicsystemsandequipmentrequiresadequateknowl- edgeandrealizationaboutthecomponents,besidesdeepunderstandingaboutthedesign, 15 Table2.1:Comparisonoftreliabilitypredictionmethodologiesforelectronics. Methods Advantages Disadvantages Empirical Implementationisverysimple sincemodelsarealreadyavail- able. Historicalrecordscanresult ininaccurateestimatefornew components. PhysicsofFail- ure Highlevelofpredictionusing knownfailuremechanismisper- formedtodeterminethewear- outfunction. Highlevelofknowledgeofcom- ponentmaterialsandprocess anddesignscienceisrequired. Hence,itischallengingtoapply sincetoomanydataandparam- etersareneeded,hardtocal- culateastoomanydataarere- quired. SimilarItem Data Fastestwaytoestimateanew product?sreliabilityandittakes placewhenlimiteddesigninfor- mationisknown. Thesimilarproductforevalua- tionmaybesubstantially entfromtheoneunderconsid- eration. Test/FieldData Resultscanbeaccuratelydeter- minedastestsincludeassoci- ateduncertainty. Thedataare toobtainandassess. Operational Translation Handyandapplicationofenvi- ronmentalfactorsfortoughcon- ditions. Shortageofup-to-dateandlim- itednumberoftranslationsce- nariospresentthechallenge. SystemRelia- bilityAssess- ment Combinetheandtest datawithempiricalprediction throughstatisticalanalysis. Itdemandsaugmentedcompu- tation. 16 themanufacturingprocessandtheexpectedoperatingconditions.Thepredictionmodels shouldberelativelysimpleandeasytomaintain,fullyintermsoftheirjobsand requirementswithidenconstraintsontheirapplication.Thechoiceofreliabilitypre- dictionmethodisbasedonexperienceandthenumberofmodesorperiodsofoperationtime thateachmethodise,whichisindicatedbycheckmarksasshowninTable2.2. Table2.2:Timeoperationperiodforreliabilitypredictionmethodologies. Methods < 1 =1 > 1 1 Testordata X X X 2 Systemreliabilityassessment X X X 3 Systemitemdata X X 4 Translation X X 5 Empirical X X 6 Physicsoffailure X Thepredictionmodelsintcategorieshavebeenbasedontheirusage, characteristics,andconditionsforapplications.Eachmodelisdependentonwidelyt setsofphysicalparameterssuchaselectricalstress,environment,quality,andtemperature. Itisbasedontheassumptionthatsystemsfailasaresultoffailuresofcomponentparts, whichfailpartlyasaresultofexposuretoapplicationstress.Again,theselectionofthe methodisafundamentalchoicemadebythedesignengineersanddirectcompanypolicies basedontheapplicationinconsideration.Italsovariesaccordingtotheproductlifecycle andrelatedreliabilitymetrics.Itisdrivenbythecriticalpartsinthesystemtobemodeled andbythesystemrequirements.Thetnessofareliabilitypredictiondependsonhow wellitisdesigned,developed,andapplied.Itcanalsobeassessedbasedonhowwellit matchesthesystem?sspaswellastheeldobservedbehavior.Dependingonthe assumptionsandmethodsused,reliabilitynumberscanvarydramaticallyandpossiblylead tomisapplicationofthesystembeingconsidered. 17 Reliabilitypredictionistiveduetothefollowingaspects[73]: Supplyreliabilitywithaquantitativeforecast. Improvedesignandmanufacturingprocess. Resultinprocessthatmeetsend-userreliability. Createcompetitiveamongdesigns. Highlightproblemsassociatedwithreliabilitysuchasdesignimbalance,sourceofun- reliability. Helpinfeasibilityevaluationtoachievedesignreliabilitylaudablegoals. Predictwarrantycostandmaintenancesupportrequirements. Assessrisksandprovidinginputstoanalysis. Reliabilitypredictionapproachesarewidelyadoptedthroughouttheelectronicsindustry. Theseapproachesareconsideredasayardstickandacriterionforthecomparisonoft typesofequipment.Someoftheseapproacheshavenarrowscope,andsomehavebeen replacedbynewerapproachesorhavebeenmobutmostofthemhavewidespread adoption.Hereanoverviewofthecommonlyusedreliabilitypredictionsmethodologiesis presented. 2.2.1EmpiricalPredictionApproach Empiricalpredictionapproachisbasedonmodelingpast-experiencedataandpresentgood estimationsdataforthesameproducts.Thusempiricalmodelshavebeendevelopedfrom historicalreliabilityrecords,whichareobtainedfromtsourcesandenvironments, 18 eitherfromactiveorlaboratorytests.Therefore,thereliabilitypredictionwillvary asafunctionofthespempiricalpredictionapproach.Someofthefrequentlyutilized empiricalpredictiontechniqueswereinitiallydevelopedformilitaryortelecommunications, butnowtheyhavealsobeenwidelyappliedinmanyotherindustries. MIL-HDBK-217:themilitaryhandbookforthereliabilitypredictionofelectronic equipment. Telcordia(Bellcore):reliabilitypredictionprocedure(RPP)forelectronicequipment TR-332. HRD-5:BritishTelecomhandbookofreliabilitydataforcomponentsusedintelecom- municationsystems. NTTProcedure:NipponTelegraphandTelephoneCorporationstandardreliability tableforsemiconductordevices. CNET-93:FrenchNationalCenterforTelecommunicationsstudies; RDF2000:FrenchTelecommunications. IEC61709:referenceconditionsforfailureratesandstressmodelsforconversion. IEC62380:reliabilitypredictionprogrambasedontheFrenchTelecommunications standardRDF2000. 229B:ChineseMilitaryStandard. SiemensProcedure:Siemensreliabilityandqualityspfailureratesofcom- ponents. 19 PRISM:systemreliabilityassessmentmethodologydevelopedbyRAC. Themainadvantagesofempiricalapproachisthatitservesasgoodperformanceindica- torsoryardstickofldreliability,simpletouseifthemodelsareavailable,anditprovides aectionofactualfailurerates.Ontheothersideitishardtokeepsupportdata, tocollectdatafromtheirsourceseitherapplicationorlaboratorytestsince failureratescandependonthediversityofthesourcesofdata[73]. 2.2.2PhysicsofFailurePredictionTechniques Thetechniquebasedonphysicsoffailuremodelseachfailuremechanismforeachcomponent. Thismethodhasbeenusedindesignstagepriortodevicelife.Bottomupphysicsof failuremethodrequirescomprehensiveknowledgeofthethermal,mechanical,electricaland chemicallifecycleenvironmentaswellasprocessesleadingtofailuresintheinorder toapplyappropriatefailuremodels.Theultimategoalistopredictforacertaincomponent whenendoflifemechanismwilltakeplace.Thecomponentfailurerateisthesumofallthe failureratesduetovariousfactorssuchasthermal,humidity,voltage,andthermalcycling. Thesystemfailurerateisthesumofallthefailurerateofthecomponents.Examplesof suchfailuremodesincludethethermalagingofelectricalcomponents,theonsetofhigh cyclefatiguecracksinstructuressubjecttocyclicloadsorthedeteriorationofsealsleading tolubricantleakageandcontamination. PhysicsoffailurehasprosperouslonghistoryinelectronicreliabilityTypicalad- vantagesofthephysics-of-failurearemodelingofpotentialfailuremechanisms,estimatingof end-of-life,determiningthevariabilityofeachdesignparameter.Moreover,commonfailure modelsthatareeforexistingdesignscanbeelyappliedtonewmaterialsand 20 structures.Onthecontraryitcannotbeusedtoestimatethereliability.Furthermore, properapplicationsofthismethodrequiresdeepunderstandingoffailuremechanismand designprocessanditisnotapplicableforassessingawholesystem. 2.2.3Test/FieldData Themethodbasedondataworksinthethreemodesofoperations.Thereliability outcomecanbepreciselydeterminedincludingtheassociateduncertaintyoftheestimate, butitisamethodsincethedataarenoteasytocollectandorganize.Itisusedto evaluatereliabilityofelectronicequipmentbasedonbothfailureandtime. 2.2.4SystemReliabilityAssessmentPrediction Thisapproachtoreliabilityanalysisallowssystemarchitectstoanalyzereliabilityofthe systembeforeitisbuilt.Thisapproachinvolvestwosuccessivestages:pre-buildphase andpost-buildphase.Thephaseisbasedonusingconsolidatedreliabilityassessment method,whichtakesintoaccounttheprocessgradingfactors,requirementspart quality,designandmanufacturinginadditiontoinitialprediction,operationalsoft- ware.Thesecondstageconsolidatesthebestestimateswithsystemtestdataandprocess defectdatausingBayesiancombination.TheBayesiantechniqueisastatisticaltheoryfor combiningtheresultsofseparatestatisticaldata.Thetechniquehasbeenproposedasa methodforcombiningtestresultswithpreviousdataorwithsubjectivejudgmentinorder toderivebetterormoreeconomicalreliabilitypredictionsbasedonlimiteddata. 21 2.2.5SimilarItem/CircuitPrediction Thismethodhasbeenusedtocollectdataofthepastexperienceonthesimilarproducts.If thenewproductshowsagoodperformance,thenthedataprovidegoodrecordforcomparison forthenewitem.Duringthetranslationprocessenvironmentandoperatingconditionsmust beconsidered.Themainfeatureofthistechniqueisthatitisveryquicktoestimatethenew productreliabilityanditisavaluableapproachwhenthereisshortcutindesigninformation. Nonetheless,thepossibilitythatthenewproductistfromthesimilaronecouldcause inaccurateprediction,whichconstituteamain. 2.2.6PredictionbyOperationalTranslation Theoperationaltranslationmethodisbasedonempiricalapproachtoestimatingreli- abilityvalue.Itdependsonfactorsreliability,whichincludeallfailurecauses viz.inducedfailuresresultedfromincompetentsystemdesign,imperfectionmanufactur- ing.Althoughitisastraightforwardprocesstoapplythismethod,therecouldbelimited availabilityofbothtranslationscenariosanduptodateempiricaldata[74]. Itisworthmentioningthataccordingto[75],thebestreliabilitypredictioncouldonlybe achievedbyacombineduseofrentmethods,dependingonthedesign,developmentor manufacturingphase,providedthatthesemethodologiesarenotinterchangeableapproaches. Thereliabilitypredictionsareessentialtechniques,whichimprovereliability,andaremathe- maticallysimplebutnotaccurate.Historicaldataplayamajorrole.Theytakeintoaccount thedesignbetweenusingalargenumberoflowfailureratecomponentsversususing alowernumberofhighfailureratecomponents. Furthermore,throughcompetitiveanalysis,aapproximationoftheproduct's 22 inherentreliabilityisobtainedbycomparingMTBFsofcompetingproductsagainstthe onespredictedbymodels.Thehighercomplexityoftheproducttypicallycorrelatestothe higherprobabilityofimperfectionintheInsummary,reliabilitypredictionmethodsare coherentandmathematicallysophisticated,butontheotherhandtheyhavesomelimitations andtheyareinaccurate.Thelimitationsitcanbeovercomewiththeuseofhistoricaldata andrecordsandappropriatecorrectionfactors.Reliabilitypredictionscanachieveadequate accuracyonapracticallevel.Thevalueofthesemethodsisveryhighattheearlyphase, butthevaluedecreasesrapidlyasprototypesbecomeavailablefortesting.Employingthese methodscanincreasethebaselinereliabilityofaproduct.Sinceproductreliabilitydepends mainlyondesignandoperationconditions,othermethodsandtechniquesshouldbeusedto proveandimprovereliability[76]. 2.3FaultTreeAnalysis(FTA) Faulttreesanalysisisanalyticallogictechniquethatcanbeappliedtoanalyzingsystem reliability.Thediagramfollowsatop-downstructureandrepresentsagraphicalmodelof thepathwayswithinasystemthatcanleadtoafailure.Basedonasetofrulesandlogic symbolsfromprobabilitytheoryandBooleanalgebra,thepathwaysinterconnectcontribu- toryeventsandconditionsusingstandardlogicsymbols.Fortheconnectedsysteminthis study,theresultingfaulttreediagramisagraphicalrepresentationofchainofeventsin windsystem.Theprobabilityofthetop-leveleventcanthenbedeterminedbyusingmath- ematicaltechniquesthatarewidelyusedinsystemreliabilityandsafetystudies.Faulttree analysistheabilitytofocusonaneventofimportance,suchasahighlycriticalsafety issue,andsubsequentlytoavoiditsoccurrenceorminimizetheconsequence. 23 2.4ReliabilityBlockDiagram(RBD) Inthisgraphicalanalysistechnique,thesubsystemsorcomponentsareconnectedaccording totheirfunctionorreliabilityrelationship.ThemainadvantageoftheRBDmethodisthat itiseasytoread.InaRBD,thelogicdiagramisarrangedtoindicatewhichcombinations ofcomponentfailuresresultinthefailureofthesystem,orwhichcombinationsofproperly workingcomponentswillkeepthesystemoperational.AblockinRBDrepresentsthe workingphysicalcomponent,andthefailureofthiscomponentisindicatedbytheremoval ofthecorrespondingblock.IfenoughblocksareremovedinanRBDtointerruptthe connectionbetweentheinputandoutputpoints,thesystemfails.Inotherwords,ifthereis atleastonepathconnectinginputandoutputpoints,thesystemisstilloperatingproperly. Generallytwomaintypesofconnection,seriesandparallel,canbeestablishedbetweentwo ormorecomponents[77].SeriesconnectionsrepresentlogicANDofcomponents,andparallel connectionsrepresentlogicOR.Theparallelunitsinthesystemmeansredundancy.Asystem keepsoperatingsuccessfullyuntilnovalidpathfromleftmostnodetorightmostnodecan beformedfromavailableconnections.Typically,theone-linediagramispreferredsinceitis easytounderstandbyengineerswithminimalexperiencewithreliabilityengineering.This makesRBDaneasytooltousefordeterminingthereliabilityofspdesignsandfor comparingmultipledesignvariationstodeterminethepointofdiminishingreturns.Using RBDonecanhandlemostofreliabilitysituationseventhoughithassomelimitations.For exampleitsupportsonlystandardnsbesidesitsinabilitytorepresentsequence dependentfailures.Inaddition,itisnotintuitivetorepresentfailurescausedbyhuman operators,externalevents,andthelike[78]. 24 2.5FailureModesandAnalysis(FMEA) FMEAisasubjectiveanalysistool,usingaqualitativeapproachtoidentifyingpotential failuremodesandtheirinitiatingrisktoeitherthesystemdesignormanufactureorop- erationalphase.Hence,theFMEAdrivesdesignstowardshigherreliability,quality,and enhancedsafety.Itcanalsobeusedtoassessandoptimizemaintenanceplans.Inbriefit isamethodtoimprovereliabilityduringdesignstage[79].FromtheFMEA,anumerical valueisassignedtoindividualfailuremodesinordertohighlightparticularareasofrisk. Themaingoalistoidentifyandlimitoravoidriskwithinadesign.FMEAisusuallycarried outbyateamconsistingofdesignandmaintenancepersonnelwhoseexperienceincludesall thefactorstobeconsideredintheanalysis.ThefactorsincludeseverityS,howthefailure thecapitaloperationofthesystem,occurrenceO,howthefailuremodeistoinitiate, anddetectionD,howlikelyfailureistobedetectedusingcurrentconditionmonitoringand inspectiontechniques.Thesethreemainfactorsareindividuallyratedusinganumerical scale,typicallyrangingfrom1to10.Thesescales,however,canvaryinrangedependingon theFMEAstandardbeingapplied.Theriskprioritynumber(RPN)isthencalculatedas follows: RPN = S O D (2.1) withvaluesbeing S O D = X ( W i x i ) X W i : (2.2) where W isweightvalueoftheexperts, x isrankand i isnumberofexperts[80]. 25 2.6TheAnalyticalMethods Markovmodelsthatarealsoknownasstatespacediagramsorstategraphsprovidevarious measuresofasystemincludingavailability,MTTR.Markovmodelmaybecomeexcessively complexdependingonthedimensionsofthestatespace.Thismethodconsiderswindspeed, failureandrepairratesofwindturbinesaswellasloaddemandsforshort-termandlong- termreliabilitycalculationandcomparison.ofchangeininitialnumberofworking windturbinesandrepaircrewcanbeinvestigated.Sotheanalyticalmethodsrepresent thesystembyenumeratingpotentialincidents.Thecomplexityofcalculationincreases tlywithexpansioninsizeofpowersystem[23].Whenawindfarmiscomposedof hundredsofwindturbinesthataregrid-connected,thecalculationwillbeconfrontedwith greatculties.Ithastobedeterminedwhenthesystemisrepresentedbymathematical modelsandwhendirectanalyticalsolutionsareusedtoevaluatereliabilityindices.These techniquesrequiresomedegreeofapproximationseveniftheyrepresentthefastestsolution inalmostanykindofanalysis.Theymightactasablackboxandsomeinternalaspectof themodelmightnotbecompletelyevaluated. Inanalyticalmethodsthesystemisrepresentedbymathematicalmodels,fromwhich directanalyticalsolutionsareobtainedtoevaluatepriorireliabilityindices.TheMarkovian approachissuitableformodelingtheenergyproductionandpoweravailabilityofawind turbine.Inthisapproach,thewindspeedstimevariabilityistakenintoaccountbymeansof windspeedinadiscretenumberofcontiguousclasseswitheachcorresponding toarangeofvalues.Thedurationofeachclassisstatisticallytreatedtopreserveinformation aboutitsdurationandthetransitionratesintoalltheotherclasses.Themajordisadvantage oftheanalyticalapproachisthatitdoesnotconsiderthechronologicalvariationofwind 26 speed.Moreover,theMarkovmodelhasanewtechniquecalledvoterbasedstatereduction (VBSR)techniqueforreducingthenumberofstatesinFMEA.TheVBSRtechniquemakes reliabilityanalysisandassessmentofpowerelectronicsystemsmoresimpleandmeaningful [25]. 2.7SimulationMethods SimulationmethodsbasedonMonteCarlotechniquerequiresunfortunatelylargecompu- tationalburdenduetotheamountofdataprocessing.Althoughthemaindrawbackofa MonteCarlosimulationisusuallyitslongcomputationtime,computationtimemaynot beanissueifthestudiedsystemisnotlarge.Thesimulationmethodhasadvantageof includingallvariablesofthesystemandfeaturingmoreyintheanalysis.AMonte Carlosimulationapproachisbasedonhourlyrandomsimulationtomimictheoperationof agenerationsystem,takingintoaccountthenatureofwindspeed,therandom failuresofthesystem.TheMonteCarlosimulationmustbeoptimizedtoreducerequired timeandsamplenumbers. Thismethodestimatesreliabilityindicesbyrepeatedlysimulatinglargenumberoftrials toreplicatetheoperationofapowersystemandrandombehaviorsinthesystem.This methodtreatstheproblemasaseriesofrealexperiments.Oneofitsadvantagesisthatthe multi-statecomponentscanbeincorporatedintheanalysiswithouttincreasein thecomputingtime[81].TherearetwobasictechniquesusedwhenMonteCarlomethods areappliedtopowersystemreliabilityevaluation,whichareknownasthesequentialand non-sequentialtechniques. MonteCarlosequentialsimulationissuitablefortheanalysisofchoppywindgenerat- 27 ingsources.Themainfeatureisthereliabilityevaluationcombinesthechronological characteristicsofwindsuchasdiurnalandseasonwindspeeds,loadandthe chronologicaltransitionstatesofallthecomponentswithinasystem.Sequentialsim- ulationcanproviderealisticandaccurateresultsrelatedtowindpower[82]. Anon-sequentialMonteCarlomethodhasbeendevelopedtoevaluatethereliabil- ityindicesofinterest.ThisMonteCarlosimulationtheoreticallycouldincorporate anynumberofsystemparametersandstates.Nonetheless,intheestablishednon- sequentialsimulation,onlyhourlyuncorrelatedstatesareconsidered. Incomponentmodelingforreliabilityassessment,MonteCarlosimulationcaninclude multi-statecomponentreliabilitymodelsbyanykindofprobabilitydistribution,not onlytheexponentialoneasinothermethods[83]. MonteCarlosimulationmaybepreferablefor[83]: Modelswithnon-exponentialtimedistributions; Characterizationofpeakingunits; ofdistributionsfunctionofoutputindices; Useoftimedependentonchronologicalissues. SowecansummarizeMonteCarloas[83]: Amongreliabilityassessmentmethods,itisarobustmethodbecausethedetailedmod- elsofprimarygeneratingresourcesavailability,internal/externalgeneratingdispatch andcustomerdemandaresimulated.Alsoitcanbeeasilyintegratedintooperative actionssuchasloadshedding,re-dispatch,reactivepowermanagementandplanning toolslikeoptimization. 28 AnykindofprobabilitydistributionsuchasexponentialoneisallowedinMonteCarlo formodelingtimestooutageandtimestorestorationofthecomponents.Thus,with thismethod,agedcomponents,failurerate,andrestorationtimes,canbeeasilymod- eled. Thehighcorrelationbetweendetailedmodelingandreliabilityassessmentofenergy limitedsystemsleadstomorefactualstudieswithlowriskofusingtheresultswhen decisionmakingisrequired. Inthecasesofenergylimitedsystems,iftheoperatorsworkonmaximumcapacityand exceedthecomponentsusefullife,theneedforthesystemupgradingisveryessential, inordertohaveadetailedmodelinthisseveremodeofoperation. 29 Chapter3 ReliabilityatWindEnergy ConversionLevel 3.1ModelingWindTurbineReliability Becauseofeconomicimportance,reliabilitytheorycategorizesthesystemsintorepairable andnon-repairablesystems.Themajorreliabilityindexofasystemisfailurerate,orto bemoreprecise,instantaneousfailurerate.Inreliability,\rate"isalwaysaconditionalrate, Forthecaseofnon-repairablesystemcomponentshavenotfailedorfailedonlyinwhich casethe\hazardrate"isthesuitabletermtoreliabilitymeasurement.Otherwise,for repairablesystems,itiscommonlyassumedthattheconditionofthecomponentscanbe restoredtofullfunctionalperformance,inwhichcasetheterm\failurerate"willbeusedto avoidconfusion.Thesubjectofreliabilitycanthenbetreatedbythetheoriesofprobability andstatistics. Repairablewindturbinesystemmodelinghasshedlightonreliabilityaspect.Earlier studiesandresultsinthisusuallyconsiderthatasystemaftereachrepairisassame asnewforperfectrepairorassameasoldforminimalrepair.Thesetwoassumptionsare foundverylimitedusesinpracticalwindturbineapplicationssincemostrepairactivities mayrealisticallyresultinacomplicatedonethatisinbetween.Nowadays,researchersstart tofocusmoreonthistypeofrepairablesystemswhererepairactionsdonotbringthesystem 30 toanas-same-as-newsituationbutratherbringthestateofafailedsystemtoalevelthat issomewherebetweennewandthestatuspriortofailure. Mostimportantmodelsofwindturbinesarebasedoneitherhomogeneouspoissonprocess (HPP)orpowerlawprocess(PLP).Fromthispointtmathematicalmodelsrises suchaspointprocess,Poissonprocesses,homogeneousPoissonprocess(HPP)andnon- homogeneousPoissonprocess(NHPP)[30][31][32]. Pointprocessisastochasticprocessdescribingtheoccurrenceofeventsintime.In studyingwindturbinereliabilitytheeventsarefailuresandtheindexisasetoftimes orasetofvariablesexpressingthelifeofobjects.Thetimebetweenfailuresarenot independentandidenticallydistributed(IID). Poissonprocessesisapointprocessthatsatisesthefollowing: { Settheobservationbeginningperiodat t =0,thenumberoffailuresis N (0)=0. { Thenumberofperiodsareindependent, 8 a 1 For > 1,theintensityfunctionincreaseswithtimeandPLPdescribesreliabilitydeterio- rationorwearout.Practically,ithasnotyetbeenencounteredinwindturbines,probably owingtotheirrelativelyyoungage.Furthermore,ifthereliabilityofawindturbinereduces dramatically,itwillbetakenoutofservicebeforethedeteriorationphasecanbedetected. Themajorwear-outfailurecausesarethefollowing: 37 Aging; Wear; Degradationinstrength; Fatigue; Creep; Corrosion; Mechanical,electricalandchemicaldeterioration; Replacementoffailedpartsbypartiallyagedones; Shortdesignedinlife. Thefailurerateisobtainedbythenumberoffailuresperturbineperyear,whichiscalculated by[32] = K P k =1 n i;k N i T i 8760 (3.10) where N i isthefailurerateandobtainedasthenumberoffailuresperturbineperyear. n i;k isthenumberoffailuresinsubassembly k duringinterval i . T i isthenumberofturbinesinpopulationatinterval i . 38 3.3ElectricalandElectronicComponentsReliability Duetolowaccessibility,theelectricalandelectroniccomponentsofwindturbines requirehighreliabilitytomaketheWTGmeetavailability.Itisworthmentioningthat theelectricalandelectronicfailuresaremoretoidentifybeforefailuretakesplace. Andtheworkingconditionsforwindpowerplantelectricalandelectronicequipment?sare tfromthoseworkinginotherindustrialconditions.Evaluationofelectricaland electroniccomponentsreliabilitymustconsideroperationconditions,maintenancerecords, andfailurerecords. 3.3.1ElectricalComponents Insidethenacelle,aWECSiscomposedofasuggestedmachinetypeconnectedtoahigh- frequencystepupvoltagethree-phasetransformersatisfythattransformersoperatingat highfrequencycanreduceitsvolumeandweightthatcanbeeasilyabletointothe nacelleofwindturbine[87].Inaddition,thetransformerwithhighfrequencyprovidesa galvanicisolationbetweengridandthegenerator.AcompletepowerelectronicWECSis necessary.Reliabilityassessmentofwindturbelectricalcomponentsshouldconsider bothphysicalmodelandmeasurementuncertainties.Thedeteriorationcriteriaforelectrical componentsmustbeproposedduetooperationaltemperaturenswithinwindtur- bines.Inadditiontoenvironmentfactors,temperature,humidity,turbulencestressdirectly reliabilityassessment.Themainrootcauseofelectricalfailuresisthetypicalvolt- ageirregularitiesandelectricalstress.Themechanicalstressesofwindturbinemachinealso playatrole.SometimestheyarealsobypoorpowerqualityfromIGBT convertorsinwindturbines.Amongalltypesoffailures,theelectricalonesarethemost 39 toidentifybeforefailures.Appropriatemaintenanceandotherpredictivetechniques areconsideredasakeyforincreasingmachinelifeandcandrivethemajorimprovements towardzeromaintenanceandreducingcost.Herearesomeofelectricalwindingsfailuresin windturbinegenerators[88]: Rotorbanding; Conductivewedges; Coolingsystemfailures; Rotorleaddamage; Under-designedmaterialsandsystems; Catastrophicfailureduetosurges; Contaminationissues. Turbineelectricalandelectroniccomponentsrequiremonitoring,control,reporting,routine maintenanceandtestingtomanagetheirfailureandincreasetheirreliability.Thefailure ratesofelectricalandelectronicsubassembliesarehigherthanthemechanicalsubassemblies, buttheirdowntimearelowerthanthemechanicalones.Thisisduethattheelectricaland electroniconeshavehighexchangeability.ThusthereisaneedtoachievehigherMTBF ofthepowerelectroniccomponentsforareliabledesignofthepowerelectronicsystemsin WECS. 3.3.2ElectronicComponents PowersemiconductordevicessuchasIGBTsarewidelyusedinelectronicapplicationsas wellasWECSelectronicmodules.WECScomponentsuppliersandproducerserIGBT 40 asasolutiontohandletheincreasedvoltagepeakswithinthegeneratorthatmightharm thecomponents.Further,IGBTsareusedinWECScontrolsystemsaswell.TheIGBTs workinaharshenvironment,wheretemperaturedrastically.Themission forIGBTscanbecategorizedasoperational,mechanicalandenvironmentaltemperature loadings.Thevoltagesstressiscreatedbypowerelectronicconvertersinthesameunitor theneighboringunitswithinthesamewindfarmorotherwindfarmsinpowersystemplant. Ingeneralrootcausesisnotaneasytask.Powerelectronicisabletochangethe basiccharacteristicofwindturbinefrombeinganenergysourcetobeanactiveelectrical powersource[89].Thereliabilitydesignofpowerelectronicconvertersinwindsystems mustconsidertherequirementsofminimumpowerlossforconverter,maximumoperation reliabilityandminimumcapitalcostofthesystem[33].Thethreemostimportantcausesof failureofsemiconductordevicesarevoltagestress,cosmicrays,andthermalcycling[90]. Althoughtherearesomeproblemsraisedwithconvertersinwindturbinessuchas[91]: Convertershasaconsiderablefailurerate,hencelossofpowerproduction. Atlowpowerlevels,convertershavelow. DuetoPWM,converterscauseharmonicvoltageonthegrid. Powerelectronicconverterofvarioustopologiessuchasmatrix,modularandmultilevelcon- vertersplayatroleinwindturbines.Theyenableimprovedoutputwaveform, reducedharmoniccontentcomparedtostandardtwo-levelconverters,higherpowerrating, andlowerstressacrosstheswitches.Nonetheless,amultileveltopologyhasnotbeenwidely adoptedinwindturbinesduetocomplexityfeatureandnegativereliabilityimpact.Fur- thermore,redundantmultilevelconverterswillbemoreexpensive[91].Relatedtomatrix convertersithasseveralfeaturesinwindturbinesviz.all-siliconbasedconverter,noDC-link, 41 highpowerdensityandcompactdesign.Thedisadvantageofmatrixconvertersisthehigh costduetohighnumberofswitches[91].Modularconverterstypicallycompriseanumber ofconvertermodulesconnectedinseriesorparallel.Forthetopologyofmodularconvert- ers,themainadvantageisthatthebasicmodulesareinterchangeableandrable, therebyprovidingaredundantandreliableconvertersystem[91]. Increasingavailabilityofpowerelectronicconverterstypicallyinvolves[79]: Improvementofthereliabilityofacomponent; Conditionmonitoringappliedtothesystem; Conditionmonitoringappliedtothecomponents; Prognosisappliedtothecomponents; Redundancyandfaulttolerance. Whenthesystemisequippedwithredundancy,thecomponentwillbeisolatedand theredundantcomponenttakesovertheoperationincaseoffailure.Redundantstrategy triestokeepupoperationwhenthefailuretakesplace.Prognosisappliedtothecomponents enablesthepredictionoftheremainingusefullifeforthecomponentsuntilreachingthe wear-outperiod.Althoughthecomponentlivesintheperiodofahigherfailureprobability duetothedeteriorationmode,anexacttimeforthefailurepointcannotbepredicted. Conditionmonitoring(CM)isareal-timemeasurementoftheconditionofacompo- nent.Ifitdriftsawayfromthehealthycondition,anappropriateactionwillbetaken[92]. Thistechniquemonitorstheoperatingcharacteristicofpowerelectroniccomponentsinwind turbines.Itisconsideredasoneofthecostemeansofimprovingreliabilityandcus- tomerserviceinpowerequipment?s[92].Itisappliedtothecomponentsthatareclosely 42 relatedtothedeteriorationfailuresinthewear-outperiodofthecomponents.Itenables todecreasethenumberofextrinsicfailuresinthecomponents.Themainfeatureisthat changeofthemonitoredcharacteristiccanbeusedtoschedulemaintenancebeforefailure occurs.Conditionmonitoringtechniquesareutilizedinsomeaspectssuchasvibrationanal- ysis,oilanalysis,thermography,strainmeasurements,acousticmonitoring,electrical processparameters,visualinspection,performancemonitoringandself-diagnosticsensors. Themainproblemisassociatedwithconditionmonitoringtechniquesistherequirement oflargenumberofsensorsamongtheofwindturbines,whichmakesthistechnique complexandexpensivetoimplement.Nonetheless,theapplicationofconditionmonitoring topowerelectronicsystemsisveryimportantbecause[92]: Inunpredictablefailuresuchascatastrophicaccidentorunscheduledmaintenance,the useofCMinpowerelectronicbecomesmoreessential. Comprehensiveknowledgeaboutthefailureinwindturbinesinadditiontoimprove- mentinsensorsandsignalprocessingleadstoepowerelectronicconverterCM systems.Applyingtheconceptofconditionmonitoringtopowerelectronicsischalleng- ingissue,andmustbeaddressedinasurveypaper[92].Structurehealthmonitoring (SHM)techniquesareunclear.Particularlythevibrationmonitoringonthewindtur- binetowerhasnotbeenovercomeyet[93].Someofthecomponentsofwindturbine thatneedtobemonitoredarefaultsduetoimbalance,wear,fatigueandimpending cracksinrotorblades,bearings,shafts,gearbox,generator,yawandthepitchangle mechanism[93]. Diagnosisistoidentifytherootcauseofthatfaultifithasoccurred[92].Diagnosisen- compassesdetecting,isolating,classifying,andanalyzingfaults[94].Prognosisassessesthe 43 currenthealthlevelofacomponent,andpredictsthehealthofthecomponentatsomepoint inthefuture[92].ThemostimportantaspectforconnectingaWECSwiththeelectric gridrequiresapowerelectronicconverterthatallowsvariablespeedoperation,reducesme- chanicalstress,andincreasesreliability.Zhuetal.baseonavailablestatisticsthat around40%ofsystemfailuresisrelatedtowindturbinesconvertersreliabilitycausedby indelicatedesignorindecorousoperationorimpropercontrol[95]. 3.4ReliabilityDesigninWECS Nowadays,presentingfavorablecircumstanceslikelyresultsinorshowssignsofsuccessin powerelectronicofwindturbines.Powersemiconductorsarenormallyusedinapplications wherehighreliabilityisamust.Longlifecycleofpowerelectronic,warrantycostsreduction, andminiaturizationelectroniccomponent,inadditiontoaccurate,inexpensive,tand lesstimeconsumingreliabilitytestmethods,areneededforpowerelectroniccomponents amongofwindturbines.Somemethodsortestingprotocolshavebeentriedto beintegratedtohelpwindfarmengineersandoperatorsrevealandpredictreliabilityissue inspeedyandtway,suchas: Highlyacceleratedlifetest(HALT)revealsweakpointsandpossiblefailuremodesin shorttestingtime,say2-5daysandprovidesinformationaboutproductoperationin aharshenvironment.Thistestcanrapidlyweaknessesunderaccelerated stressconditions.ThemainfeaturesofHALTincludesavingtimeandmoneydue toeasyapplications.Thistesttechniqueisbestsuitableforapplicationduringearly engineeringdevelopment. Acceleratedlifetesting(ALT)isaprocessofdeterminingthereliabilityofanelectronic 44 componentsoveraofwindturbinesinashortperiodoftimebyaccelerating theuseenvironment.ALTisalsosuitablefordominantfailuremechanisms. Itisusuallyperformedonindividualassemblylevelratherthansubsystemlevelor systemlevel.Itispreferredtousewhenthereiswear-outmechanisminvolved.Italso predictsthelifeofthepowerelectronicandelectricalcomponentsinthewind turbinesandallowstheoperatorstoprioritizecomponentimprovements.Thefailures forALTaretime-dependent.Lifemodelsareutilizedtocorrelatetheproducttestand applicationloadconditionstothefailuresitesandlocalstressconditionsandthento relatetheselocalstressconditionstodamagebasedontheidenfailuremechanisms andmaterialproperties.Themaindisadvantageisthatisneedstlymoretest timeandsamplesalthoughitprovidesinformationaboutthelifeexpectationsinthe testconditions.Thislifeinformationisusefultoevaluatedesignchanges,warranty issues,andlifecyclecosts.Thistesttechniqueisbestsuitableforapplicationbefore productionrelease[95]. Environmentalandfunctionaltestingistoguaranteetopperformanceinclimaticcon- ditions.Powerelectronicconvertersinwindfarmmusthavehighqualitytowork perfectlywhereverandwhenever.Extremeenvironmentalconditionscanhaveanex- tremeonoverallfunctions.Converterslifecycleandtestsproveits reliabilityandtoidentifyweaknessesandinitiateitsimprovementsatanearlystage. Thistestvthatacomponentiscapableofoperatingunderthetestconditions foracertainperiodoftime.Itcannotbeusedtoquantifyreliabilityparameters,such asfailurerate,MTBF,failuredistributions,performancedegradation,etc. 45 Rootcauseanalysisrelatedtoelectricalandelectronicsubassembliesinwindturbines havebeenconductedin[96].Theanalysesincludecomponentthermalaging,thermalme- chanicalcyclingfatiguemechanismsintheconductionandinsulationmaterialsofelectrome- chanicalcomponents,andthermalmechanicalfatiguestressexperiencedbypackagingma- terialsoftheintegratedpowermodules(IPM)inthepowerelectronic.Inadditiontoroot causespresentedin[97],viz.stressstrengthinterferenceduringoperation,occurrenceof randomloadshigherthanexpected,occurrenceofrandomstrengthslowerthanexpected, tsafetymargins,humanerrorsinusage,thissurveysuggestedsomeprotocolsof reducingtheofthesefailurecausessuchasequipmentfailurebeginwithacomplete understandingoftheequipment.Ithasbeenfoundthatthemostimportantfactor componentreliabilityisthedegreetowhichthemanufacturerisabletofabricatedefect-free components[97].Thein-depthknowledgeofoperationandmaintenancepracticeandrefer- encestoequipmentdesignandengineeringpracticesmakeitpossibletodevelopanoverall understandingoftherootcauseandtotlydirecttheinvestigationofanelectricaland electronicsystem. Reliabilitystrategiesinvolveastructuredapproachtoidentifyingcriticalequipmentand systems.ThiscouldincludeaFMEAtoidentifythecriticalcomponentorsystemfailure modes.AMTTForMTBFmodelcanalsobeusedtoderiveaprobabilisticreliability model.gtheappropriatemaintenanceregimenandreplacementstrategiesbasedon thatcriticalitydeterminationisalsopartofawell-designedreliabilitymechanism.When doneproperly,windturbinesleadstolaudabletarget,optimalreliability.Atvariouslevels, rootcausesanalysiscanassistdesign,operation,feasibilitystudies,scheduling,budgeting, andrevenueallocation. 46 Properlyintegratedanalysiscanbemodularizedatanydesiredlevel,resultinginuseful informationtoaiddecisionmakingaboutfailureinelectricalandelectroniccomponents. Factorsoperationsreliabilityareanalyzedtoevaluatethecauseofbreakdownand energyproductionstoppageorreduction.Suchfactorsincludereliability,performance,and adequacyofstructures,equipment,controlsystems,andoperatingandmaintenancepro- cedures.Detailedanalysisofcriticalstructuralsystemsandprocessequipmentmustbe performed.Inaddition,factorsproductqualityareanalyzedtoreducethepoten- tialofelectricalandelectroniccomponentsbeingmanufacturedoutsidespcation.These includeprocesscontrolranges,statisticalsampling,andqualityassurancetestingandrelia- bilitytestingmethods.Workexecutionincludestheidenofworktobeperformed, aswellastheplanning,scheduling,andperformanceofthatwork.Thisispartofquality controlscenario.Continuousimprovementistheprocessbywhichanorganizationlearns fromtheperformanceofeachintheprocessandappliesthatknowledgetoimprovee- nessandthrougheachprocesscycle.Itincludesproperworkcloseoutprocedure, aswellasacomprehensivecorrectiveactionprograminreliabilitycommunity. 3.4.1DesignforElectricalReliability Windturbinesaresubjectedtottypesoffailures.Thusitisnecessarytoidentify whichkindoffailurescanbefoundintherealworldofwindplant.Eachcomponentinthe windfarmwilleventuallyfailassumingthatithasbeeninserviceforalongtime.Deep understandingofthetypesoffailuresandtheiroccurrencefrequencyinaofturbines isimportantissueinreliabilitycommunity.Windturbinelifetimehasbeenrelatedtomany factorsviz.electricalequipmentfailureincorrectinstallation,inaccurateconnectionbetween systems,subsystems,componentsinwindturbine,faults,erroneousgroundingsystemetc.. 47 Moreover,humanerrorscantakeplaceatanyinstantinlifecycleofwindturbinebe- ginningwiththestepsofdesign.Humanerrorscanbeintovariouscategories viz.design,installation,assembly,inspection,operatingandmaintenance[98].Thefailure rateisafunctionofthermalstress,electricalstress,devicegeometry,constructionstress, corrosioncracking,dielectricbreakdown,defectiveconductortracks,allofwhichcanleadto failures.Theemphasishasbeenputonreliabilitydesignprocedurewithmoreattentionto failurerates,failuremode,andfailuredataformanufacturersandsupplierstoidentifyweak designpointsandfailurecommonwithinthesystemandalsotoidentifyfaultstages generation,discoveryandcorrection.Asmentionedearlier,sincewindturbineiscomplex electromechanicalsystem,severaltechniquesviz.conditionmonitoringandfaultdiagnosis andprognosistechniques,inadditiontoredundancymechanismandfaulttolerant,have beenattemptedtoincreasewindfarmreliability,increasingenergyavailabilityandlifetime serviceofthewindfarmswhilereducedowntimeandmaintenancecost. Fromreliabilitypointofview,itispreferredtohandleanyfaultofwindturbineswithout mechanicalvibrationsensors,whichisveryattractiveforconditionmonitoringsystemsto collectdataabouthealthysituationofwindturbines.Furthermore,itistoinstall sensorsespeciallyinwindregimeinadditiontothereliabilityissuesofthesensors.Onthe otherhand,electricalsignatureanalysisismorereliable,lessexpensiveandspeedytodetect windturbinesfaults.Consequentlystateobserversmustberesortedtoobtaintherequired informationusingonlymeasuredvoltagesandcurrentsatWECSterminals.Conditionmon- itoringtechniquesaredearoundvibrationanalysis,oilanalysis,thermography,strain measurements,acousticmonitoring,electricalprocessparameters,visualinspection, performancemonitoringandself-diagnosticsensors[99]. WithregardtoreliabilityWECSconvertersissues,faultdiagnosisandfaulttolerantoper- 48 ationcapabilityhavebeenstudiedin[95].Inaddition,gridevent/gridfaultridethough,and singlethreadversusmultithreadneedscomprehensiveinvestigation.Shetal.presentan tmethodtoimprovethereliabilityandavailabilityoftheconvertersystembyhaving atleastoneindependentredundantconverterwhichguaranteesthesystemoperationincase ofaconverterfailure[100].Althoughtheredundantconverterswillincreasethesystem?s cost,volume,andweight,theproposedcost-rateminimizationmodelaimstosimultane- ouslydeterminetheoptimalallocationofredundantconvertersandtheoptimalnumberof theconvertersthatareallowedtofailbeforesendingamaintenancetothe platform.thetotaldowntimeishighlybyallelectricalcomponents.such astransformers,cables,generators. 3.4.2DesignforMechanicalReliability Thegeneratorbearingsandgearboxinwindturbinesdrivetrainareconsideredthemost fragilecomponents.Oneofthemostimportantwindturbinefailuresisattributedtogearbox relatedissues.Gearboxisamechanicaldevicecapableoftransferringtorqueloadsfroma primarymovertoarotaryoutput,typicallyintrelationshipsofangularvelocity andtorque.Inwindturbines,thegearboxconnectsthelow-speedshaftandthegenerator. Therefore,itsgearratioisgenerallydictatedbytherequirementofthegeneratorandthe angularvelocityoftheturbinerotor[101].Gearboxfailurerateisstillhigh,whichisaround 20%ofdowntimeofwindturbine[101].Evenifthefailureofgearboxisanuisanceone,it needsacranetohandle,replacement,greasing.Cranerental,labor,economiclosseswilllead toexpensiveoperations.Thegearboxhavetoworkinrandomloadingconditions.Many factorssuchasfrictionleadtoovertemperature.Themechanicalstressescauseshaftcrack, toothbreakage,shatteringandinworstcasesdamagethetower.Windturbinecondition 49 monitoringisawitnessthatgearboxfaultdiagnosisisnoteasytaskanditisimportant toenhancewindturbinesystem?sreliability.Eachcomponentingearboxcangenerate vibrationsignalsthatareweakandhardtodetectduringearlystages. Thevibrationanalysisisthemostknowntechnologyappliedforconditionmonitoring, especiallyforrotatingequipmentinwindturbinesuchasgearbox.Thevibrationmeasure- mentisconductedtoidentifythehealthysituationofwindturbinesusingsensororobservers thatarespreadoverthewindturbine.Operatorscanmeasureacceleration.In-depthun- derstandingofthevibrationcanbeachievedbyanalyzingthedatafromsensors.Frequency spectrumisevaluatedusingspectralanalysisalgorithmsbasedonfastFouriertransform (FFT),whichcanprovideuswithcriticalinformationregardingthevibrationbeinghealthy ornot. Anotheroneofthemainfailuresinwindturbinesistransformers,whichcauselong downtime.Thereplacementinsuchfailurecaseisveryexpensive,hardandconsumestime especiallyinharshweather.Thusinsomecaseitispreferabletobeeliminated.Nonetheless, thelimitsofcostandvoltageratingofthepowerelectronicconverterandincreasedlosses incablesandtransmissionleavesolid-statetransformersgreatspaceforimprovementbefore widelyspreadadoption.Threelevelconvertershavebeenrecommendedtoconnecttoa networkwithoutatransformer[102]. Duetothelongdistancebetweentheofwindturbinesandthelargeareaofwind farms,thereliabilityoftheentirewindfarmisstronglyimpactedbythereliabilityofthe cables.Cablelifecanbedividedintostages;manufacture,storage,installation,services,and recovery.Tomaintaincablesinservice,itisclearthattroubleshootingandrepairprocedures havetobeestablishedwithreactiverepairandreplacementandproactivereplacement. Itisimportanttoinsurethatthesecablesusequalitycompoundsandconsistentlymeet 50 thespandrequirements.Failurescouldoccurwhilethecableis handled,installed,andoperatedwithinspTheeandreliableoperation ofinfrastructurecablesandredundanciessupporttheavailabilityofpowerduringfailurecase. Thereisessentialneedforareliablewaytoterminateacablesuchthatitcanwithstand thelonglifeserviceinthemarineenvironmentwherethemechanicalandenvironmental conditionsareunfriendly.Itisworthnotingthattherepairstrategiesplayanimportant roleinidentifyingoverallavailability[103].Inaddition,low-frequencyelectricalnoiseis recognizedasaverysensitivemeasureofthequalityandreliabilityofelectricalandelectronic components[104]. Howtodesignagainstfailure[102]? Conditionmonitoring; Diagnosisandprognosis; Redundancyandfaulttolerant; Collectingdata; Windfarmtopologyandarchitecture; Fieldexperience; Operatingenvironmentconditions; Choosingcomponents. Thedecisionofcomponentchoiceismorecomplicatedandisdrivenbyanumberof variedfactors.Properdecisionwillprovideexcellentsupportandsolutionstomeet 51 windfarmneeds,takeadvantageoflowercosts,increasesy,highquality.Choosing componentsappeartobeadauntingtaskwithmanytmodelsandfeaturesavailable. Theoverallpurposeforthereliabilityanalysisisto: Completedescriptionaboutfailureanditsimpactoncomponents,subsystemsand systemslevelsandrevokeandpreventunacceptableimpacts; Buildarigidsafeandreliablesystem; Provideinformationinordertodevelopsystemsandsubsystemarchitectureandvali- dationdesign; Fault-toleranceorgracefuldegradationmechanismsviz.redundancyorbackupsystems atvariouslevelsofwindturbinegeneratortoenableittocontinueoperatingproperly intheeventofthefailure.Ifitsoperatingqualitydecreasesatall,thedecreaseis proportionaltotheseverityofthefailure.Fault-toleranceisparticularlysought-after inhighavailabilityofwindturbines. 3.4.3DesignforPowerElectronicReliability Themaintaskofpowerelectronicsistohandleenergywbetweenplayers.Forexample powerconverterprovidesandhightinterconnectionbetweenplayersonsmart grid,generation,energystoragesystems,transmission,distributionsandloads.Allthese playersmustgivegridsecurityandsafetythehighestlevelofpriority.Threeimportant issuesofconcerninusingapowerelectronicsystemarereliability,,andcost[105]. Powerelectronicsformodernwindturbineshasexperiencedadramaticevolutioninwind industry.Thissectionfocusesonstateoftheartmainissuesofintroducingpowerelectronics 52 inmodernwindgenerationsystems.Theapplicationsofpowerelectronicsinwindturbine generationsystemsaregreatlyimprovingwindturbinebehaviorandperformance. Nowadaysthereliabilityofpowerelectronicsisanimportanttopicforresearchers.Re- liabilityisespeciallyimportantforwindturbinesasthesizeandthenumberof installedWTsincreases.Thecostofrepairingandthevalueofthelostenergywhenfail- ureoccurscanbehighandsometimesdisastrousforwindfarmowners.Theavailability ofmodernonshoreWTsisaround95%to99%[106].Asmentionedearlier,thegoalof WECSindustryistoachieveextremelyhighreliabilitywithzeromaintenance.Veryhigh reliabilityisalsothemainpriorityforWTsbecauseitconsumestime,money, torepairthem.WECSmanufacturersneedtoconsiderthereliabilityissuewhentheydesign newpowerconvertersforturbines.Bydesigningthepowerconvertertakingintoaccount reliability,theycanguaranteethatthepowerconverterswilllastlongenough[107][108], [44],[109],[110],[106],[111]. Windfarmsthatcanproducelargeamountsofpowerareestablishedbyinstallingand utilizingmanywindturbines.Itbecomesincreasinglyimportanttodeveloptechniquesthat arepracticallyusefulforpowerelectronicindustries.Thereliabilityofthewindfarmwhose variablespeedcapabilitiesareachievedthroughtheuseofanadvancedpowerelectronic converter.Gridpowerqualitywithvariablespeedwindturbinesmodelingandsimulation techniqueswerepresented[112],[113].Ananalyticalreviewoftstand-alonewind energyconversionsystemsbasedonpossiblegeneratortypesavailableinwindmarkethas beenreportedintheliterature[114].Theoverviewisconcentratedonthevariable-speed turbines.Geared-driveturbinesusinginductiongeneratorsandgearless-driveturbinesusing synchronousgeneratorswereconsidered.Themorestudiesnowareondevelopingrealtime electroniccontrollerssuchasaDSP-andFPGAwithhigh-speedcommunicationinterfaces. 53 Thus,itispossibletomonitor,store,andtransferalargenumberofinternalvariablesthat canbesentonlinetolocalorremotehostsinordertotakenewsetpointsofdt generationunits[115],[116].Powerelectronicdevicesandconverterstopologiessuchas back-to-back(BTB)connectedtwo-andmulti-levelvoltagesourceconverters(VSCs),BTB currentsourceconverters(CSCs)and,matrixconverters[44][117].Stateofartregarding powerelectronictopologiesandwindenergyconversionsystemssuchasPMSG,DFIG,IG andSGarediscussedandsomeofthepossiblecontrolstrategiesaretoucheduponinorderto capturemaximumenergytransferfromthewindturbinetothegrid[107].Controltechniques forwindturbinesincludebasiccontroltargets,activedampingcontrolandsensorlesscontrol [118],[119]. 3.5Severity Thereliabilitycommunityratestheimpactofsecurityissuesfoundwindenergygeneration systemsusingafour-pointscale:minor,marginal,critical,andcatastrophic.Thisseverity scaleprovidesaprioritizedriskassessmenttohelpunderstandandscheduleupgradesto windturbinesystems.Thescaletakesintoaccountthepotentialriskbasedonatechnical analysisofthefailureonsystemandsubsystemlevels. 54 Table3.1:SeverityofFailuresModes. Description Category Catastrophic i Afailuremodethatcausesdeath,WTGloss orsevereenvironmentaldamage. Critical ii Afailuremodethatcausessevereinjury,se- vereoccupationalillness,majorWTGoren- vironmentaldamage. Marginal iii Afailuremodethatcausesminorinjury,occu- pationalillness,minorWTGandenvironment damage,ormissiondegradation. Minor iv Lessthanminorinjury,occupationalillnessor lessthanminorWTGorenvironmentaldam- age 55 Chapter4 ReliabilityatWindFarmLevel 4.1Introduction Windgenerationsystemtechnologyisstillthemostpromisingoneamongtherenewable energytechnologies.Thefastexpansionofthewindpowerfacessomeproblemsthatrequire focusedreliabilitystudies.Theyandrevenueofwindfarmsisadversely bypoorsystemreliability,andhence,highmaintenancecosts.Theoflowreliability onturbinedowntimehasbeenseenmostacuteduringthemovetowindfarms. Thewindenergyexpansionanditspenetrationintothepowersystemisanimportantdriver necessitatingtheconsiderationofwindfarmreliabilitywherewindfarmisconsiderasapower source.ForexampleinDenmarkitisnolongerpossibletoignorewindfarmreliabilityin overallsystemreliabilitystudies.TheapproachesevaluatingWFbothplanningand operationofthepowersystemasawhole.Thepotentialofwindfarmistremendous. Windsystemresearcher?sseekandinvestigatesolutionswhileconsideringthenewtech- nologiesavailableforreliabilityinwindfarm.Themainreasonisthatthenumberofrandom variablesandsystemcomplexitiesgreatlyincreasewhenrenewableenergysourcesareadded tothesystem.Moreover,detailedhourlyloadmodelsconsideringtareasandzones ofthesystemarebecomingaconcerntomanyplanners.Newcomputationalmodelsand toolshavetobedevelopedtodealwiththesenewtime-dependentvariables.Toidentify thecriticalfailuremodesatthecomponent,subsystemandsystemscalewithinwindenergy 56 basedontheanalysisofavailablelongtermoperationaldataandfaultrecordsloggedbysu- pervisorycontroldataacquisition(SCADA)andconditionmonitoring,largedatarecorded fromwindfarmsoperatingintlocationsaroundtheworld.Atypicalwindfarm consistsofafewtohundredsofturbinesinstalledinarraysperpendiculartotheprevailing winddirection.Theturbinessitontowersupto300feetormoretotakeadvantageofless turbulenceofairw.Theseparationdistancebetweeneachwindturbineunitandtheother istypicallyaroundeighttimestherotordiametertominimizethewakeLarge-wind farmsistypicallyconnectedtothegridattransmissionlevel. Thetransmissionsystemoperatorhastofocusontheimpactofwindfarmonpower systemintermsofstabilityandquality.understandingofwindfarmcontrolcapabilities willleadstotheimportantknowledgeofstrategythatcanbeusedtosimulatethedynamic interactionsbetweenawindfarmandapowersystem.Thepowercollectionsysteminthe windfarmincludetheelectricmachines,windmodelandaerodynamicmodelsforthewind turbine,transformers,transmissionlinesand,thegrid.windfarmsaremore expensivethanonshorewindfarmsinbothinstallationandmaintenance. Thepoweroutputofeachwindturbinegenerator(WTG)dependsonthewindwith inherentvariabilityandonthenon-linearcontrolcharacteristicrelatingthepoweroutputto thewindspeed.Theoutputpowerofwindfarmalsodependsontheavailabilityoftheenergy conversiondevicesorpowerelectronicdevices.Reliabilityassessmentisusedtoaccountfor theofthefailureandrepairprocessesofwindturbinegeneratorsandtoevaluatethe performanceoftheintermsofavailablepoweroutput. Itiswellknownthatmanylargewindfarmsarebeinginstalledaroundtheworldfor bulkpowergeneration.Largewindfarmsareusuallyinstalledinlocationswithgoodwind resources,andconnectedtoapowersystemthroughtransmissionlines.Itisincreasingly 57 importanttoassesstheadequacyofthetransmissionfacilityrequiredtodeliverwindpower fromwindfarmstotheloads.Thewindfarmpowergenerationmodelcanbefurther motoincorporatetheofthetransmissionlineconnectingthewindfarmtothe bulkelectricsystemthroughtransformers.Thereliabilityofpowersystemdecreaseswith increasingsystemload.Theoftielineunavailabilityonthereliabilitylevelofwind powerdeliverysystemisrelativelysmallcomparedwiththewindspeedonwindturbine generators. IntegrationofWFswithdistributionandtransmissionnetworksraisesquestionsregard- ingthereliabilityoftheoverallsystemandhowthesystemreliabilityisbyfactors suchasthewindregime,levelofpenetration,loadmodel.Thewindfarmreliabilitymodel shouldaccountforthevariableanduncertainnatureofsupplyandfailureandrepaircharac- teristicsofwindenergyconversionsystems,inadditiontothecorrelationofWECSoutput withloadrequirements.Itispreferredtoincludetheloadmoapproachforboth thepeakloadandbaseloadinordertocomputethesystemrisk.Thecoincidenceofload andwindspeedpatternhighlythereliabilityofthesystem[120].Hence,ap- propriatepoliciesshouldbeconsideredtoenhancethereliability.Translatingthisgeneral insightintopracticalpoliciesisverycult.Researcherscouldoutlinereliabilityassuring measuresfromthetechnicalaspectstofacilitatepolicymaking. AwindfarmmaybemodeledasablockdiagramasshowninFigure4.1.Themodel consistsoffourmainparts: Blocka :Themodelinputisspeedrecordsdata,whichcanbeobtainedeitherfrom windspeedmeasurementsorfromtheresultsofsimulation.Incaseofsequential analysis,aMonteCarlosimulationrequiresawindspeedtimeseriesforeachsample, 58 Figure4.1WindFarmBlockDiagram. whereastheanalyticalmodelneedsasetofstatisticalinformation. Blockb :thetermavailabilityistheprobabilitythatthesystemisoperatingappro- priately[121].Itisworthmentioningthattheoperationandmaintenancestrategy ofthesystemisthemainvariancebetweenreliabilityandavailability.Thesystemis consideredhighlyreliableifitsfailurefrequencyisexcessivelylow.Whenafailure, theterminationoftheabilitytoperformarequiredfunctionofasystem[121],takes place,thenavailabilitybecomesverylowifthereisnomaintenanceorrepairactionis performed.Themainthreepartsofthisblockarewindturbine,internaltransmission cablesandconnectorstoshore[37].Eachcomponentcanbeupordown. Blockc :lumpedboth(a)and(b)blockstohavetheoutputpowerofthewindfarm. Blockd :dependingonthegeneratedoutputpower,someindicestheterm reliability.Examplesofmeaningfulindicesarecapacityfactor(CF),lossofloadex- pectation(LOLE),lossofenergyexpectation(LOEE),andexpectedloadnotsupplied (ELNS). Thegeneratingunitsaredividedintotwomaingroups:theconventionalandtheuncon- ventionalunit,theconventionalcanbecontrolled,scheduledandrepresentedbyatwo-state 59 model upanddown orbyamultistatemodelthatincludesderatedstates.Whiletheun- schedulednonconventionalunits,i.e.WECScanberepresentedbyseveralpartial-output states,whilethenumberofstatesdependonthetypeofwinddataavailable,thenature ofthewindregime,thecharacteristicsofthewindturbine,availabilityofcomputational timeandthedesiredaccuracy[122].Thegenerationsystemconsistsofcomponentsviz. aerodynamiccomponentsofbladeangleandmechanicalcouplingthroughshaft,generator componentofgeneratorexcitationsystem,generatorwinding,circuitbreakercomponents ofswitchgearandrelayprotectionandstepuptransformer.Itisthereforethereliabilityof generationsystemthatwillbeevenwhenanyoneofthecomponent?sreliabilityis atrisk.Thus,reliabilityisbasedontheincrementalreliabilityofitsparts.Thereliabilityof thedistributionsystemdependsonthereliabilityofindividualandcollectivelycomponents, whichfollowsaPoissondistributionthatisafunctionofthefailurerateandtheinterruption time.WECSreliabilitywindfarmreliability,consequentlydistributionsystem reliability. 4.2ReliabilityModelingofWindTurbine 4.2.1WindSource Windisundoubtedlythemostpopularsourceofelectricityaroundtheworld.Windenergy ist,random,intermittentandnon-scheduling.Windhasinstantaneous,minute byminute,hourly,diurnalandseasonalvariations.ReliabilityandcostofWECSrequires simulationoflongtermchronologicalwindspeeddataforgeographicalwindfarmlocations. Manystudieshavereportedstatisticaltestsonwindspeedsusingtdistributions,such asWeibull,Rayleigh, r 2 andsoon.Weibulldistributionismostcommonlyusedtomodel 60 windspeeds.Itisversatileandinvolvesascaleparameterandashapeparameterwhich canbeadjustedtosuitethewindregimeunderstudy.Usingthismodel,theprobabilityof windbeingbetweenanytwovaluescanbeeasilycalculated.Weibullmodelforwindspeeds andIEEERTS-79testsystemwasemployedwithPowerWorldSimulator8 : 0.TheWeibull probabilitydensityandthecumulativedistributionfunctionsaregivenasfollows: f w ( )= h i 1 exp (4.1) F w ( )=1 exp (4.2) where isthescaleparameterprovidesinformationabouttheaverageofthewindspeed and istheshapeparameteroftheWeibulldistributionanditprovidesinforma- tionaboutthedeviationofthewindspeedvaluesfromthemeanaswellasthefeatureof probabilitydensityfunction.Thesetwovaluescanbeobtainedbyusinganalyticalmeth- odssuchasmaximumlikelihoodestimator(MLE),methodofmoments(MOM)andleast squaresmethod(LSM).Thereareanumberoftmodelsthatcanbeusedtorepresent windspeedsinpowersystemreliabilitystudies.Forexample,observedwindspeed,mean observedwindspeed,ARMA,MA,normaldistribution,andMarkovchainmodelstosimu- latewindspeedsinageneratingcapacityreliabilitystudy.Thetwindspeedmodels resultintwindspeedprobabilitydistributions.Otherwindparameterssuchaswind frequency,siteairdensity,powerlawindex,environmental,landlawsandsocialaspects shouldalsobetakenintoaccountinareliabilitystudy.Windspeedcansystem reliabilitytly. Itiswellunderstoodthatwindpowergenerationisaninconsistentandintermittent energysource.Theofwindspeeddependsonthelocationofwindfarms.The 61 poweroutputfromthewindturbinealsowhichmakesitnecessarytostudythe ofwindenergyonelectricnetwork.Windpowerpotentialisassessedbywind monitoring,windmappingandcomplexterrainstudies. 4.2.2WindTurbineCharacteristics ThereisanonlinearrelationshipbetweenthepoweroutputoftheWECSandthewind speed.TherelationshipcanbedescribedbytheoperationalparametersoftheWECS.The commonlyusedparametersarethecutin,rated,andcutoutwindspeeds.Thehourlypower outputcanbeobtainedfromthesimulatedhourlywindspeedusing: P t = 8 > > > > > > > > > < > > > > > > > > > : 00 SW t V ci A + B SW t + C SW 2 t P r V ci SW t V r P r V r SW t V co 0 V co SW t where V ci , V r , V co and P r arethecutinspeed,theratedspeed,thecutoutspeed,and theratedpowerofaWTGunit,respectively[123].AWTGproducenopowerifthereisnot enoughwindenergytosupplythegrid.Theconstants A , B , C maybefoundasfunctions of V ci and V r withthefollowingequations[124]: A = 1 ( V ci V r ) 2 V ci ( V ci + V r ) 4 V ci V r ( V ci V r ) 3 2 V r (4.3) B = 1 ( V ci V r ) 2 4( V ci + V r ) ( V ci V r ) 3 2 V r (3 V ci + V r ) (4.4) C = 1 ( V ci V r ) 2 2 4 ( V ci V r ) 3 2 V r (4.5) 62 Thus,thepoweroutputofwindturbinecanbecalculatedfromits?speed-power?curve. Thisrelationisusuallygivenbytheturbinemanufacturer,designatedaspowercurveofthe turbine.TheindividualWECSoutputpowerdependsonthesameprimaryenergysource, thewind.Generally,isnotnecessarytoassumethatalltheturbinesinawindfarmhave similarcharacteristics.Windturbinepoweroutputisafunctionofsomeneglectedvariables suchasairdensity,pressure,temperature[35]. 4.3ReliabilityCharacteristics Asmentionedin[97][125],thekeyreliabilitymetricsinclude: Meanavailability:theaverageavailabilityovertime Failurefrequency:expectednumberoffailuresperunittimeatasptime. Totaldowntime:thetotaldowntimebetweentheindicatedstartandendtime( t 1 to t 2 ). TDT ( t 1 ;t 2 )= Z t 2 t 1 U ( t ) d ( t )(4.6) Expectednumberoffailures:thetotalnumberoffailuresexpectedbetweentheindi- catedstartandendtimes( t 1 to t 2 ) n f ( t 1 ;t 2 )= Z t 2 t 1 v ( t ) d ( t )(4.7) Someusefulitionsrelatedtoreliabilitycharacteristicsinclude: 63 { Availabilityistheprobabilityofasystemintheoperatingstateatsome timeintothefuture; { Meantimetofailure; { MeantimetorepairordowntimeMTTR:theaveragetimeforasubassembly toberecoveredforanyfailure; { MeandowntimeMDT:totalnumberofhoursduringwhichtheturbinewasnot operationali.eincludesallthetimeneededtorestoretheWECStoanoperating condition; { TimetorepairTTR:actualnumberofhourscompletingtherepair,excluding logisticsassociatedwithrepairactionsuchashavingthecomponentdeliveredto siteorarrangingthetechnicians?time; { Meantimebetweenfailuresorreliability. MTBF = MTTF + MTTR = 1 + 1 (4.8) Theaverageperiodbetweenunplannedstoppagesofasubassembly,intheeventthata failurecannotberepairedimmediately,theremaybealogisticdelaytime(LDT)tocarry outtherepair.ThentheMTBF: MTBF = MTTF + MTTR + LDT = 1 + 1 + LDT (4.9) where: FailureRate; = 1 MTBF (4.10) 64 RepairRate; = 1 MTTR (4.11) A = MTBF MTTR MTBF = 1 (4.12) ThepoweroutputofWFcanbedeterminedfromthewindspeedandwindturbine availability.Itiswellknownthatwhenthewindspeediseitherlowerthanthecut-inspeed orhigherthanthecut-outspeed,theoutputpoweroftheturbinewillbezero.forexample, mostWTGsaretakenoutofservicewheneverthewindspeedoverpassesordropsbelow certainvalues.Ifthewindspeedisinbetween,ratedpowerwillbegenerated.Thegenerated powercanbesplitstates,thenumberofwhichisarbitraryanddependsontherequired accuracyofthemodel.Thenthetimeseriesoftheoutputpowerisobtained. ttypesofwindturbineswithtratedcapacitiesandparametersareused inwindfarms.Notallofwindturbinesinawindfarmareundertheofthe samewindspeedmagnitudeanddirectionatthesametime.Windenergyproducedcanbe intothreecategories:energyproducedbyeachwindturbineunitsduringlifetime operation,energydispatchedtothepowersystemgridalsoduringlifetimeoperation,and energylossesduetobothnormaloperationandfailures.Indeed,powerproductionofwind turbinesmaybeinterruptedbecauseoffailure,theavailabilitystatusofwindturbineshould beconsideredwhenWFoutputisdetermined.Inreliabilityanalysisofawindfarmsystemis thateachindividualunitdependsonthesameenergysource,thewind.WECSforcedoutage rate(FOR)isastheunitunavailability,thatisafactorindicatesinaprobabilistic fashion,thedegreetowhichmechanicalorelectricalfailureswillmodifythemachine'spower output[126][124].TheturbineFORhasslightonreliabilityindexes,suchasLOLEand LOLFindices[120].Forthisaim,WECSctuatingoutputpowerispresentedbyt approaches.Thereafter,themeantimetofailMTTFandthemeantimetorepairMTTR 65 areevaluated.Afterward,severalapproachesusedeitheranalyticalorsimulationmethods usedtoestimatethereliabilityindicesoftheWECSsystem.Bymeasuringtheavailability ofeachwindturbinescomponentandbyrecordingthefailurehistoryofaunitsimilarto theoneunderstudyFromtheselifehistoryrecords,somestatisticalinformationareusually extrapolated.However,ifpriorityisaccuracythenitbecomesnecessarytomeasure thisfailurehistoryforaperiodthatislongenoughinordertohaveatnumberof samples.Thenwehave[127]: Unavailability ( FOR )= + (4.13) Availability = + (4.14) Thismodelisdirectlyapplicabletoageneratingunitwhichiseitheroperatingorforced outofservice.Alsowehave[127]: MTTF = 1 (4.15) MTTR = 1 (4.16) MTBF = MTTF + MTTR (4.17) where: :Repairratein repair=year FOR :Forcedoutagerate. MTTF :Meantimetofailure. MTTR :Meantimetorepair. 66 MTBF :Meantimebetweenfailures. Itobservedthatmuchofthesophisticatedcalculationsforbothwindvariationsandforced outagesareimportantforsmallnumberofwind.Theofforcedoutageswilldiminish asthenumberofturbinesincreases.Atsomepointitmaybepossibletoignoreitcompletely withonlyasmallresultanterror.Inwindfarmitwillbeassumedthatcapacityvariations areonlyduetowindspeedvariations. 4.4ReliabilityIndices OptimumselectionofWTGtypeisfocusedonsomeindices.Thegridoperatorsmostly reliabilityduringpeakloadhoursorhighpricehourssinceinjectionwindpower intothegridinpeakperiodbringsmoretotheWFownersalthoughitcompromises systemsecurity.Theremustbecoincidencebetweensystemloadpatternandwindpower regime[128].Systemsecurityissuesassociatedtotheintegrationofwindpowerhavebeen mostlyimportantforthetransmissionsystempperator?s(TSO).Researchesmustdevelop requirementsandmethodsthatenablewindturbinestowithstandcriticalsystemevents.The capacitymodelofthepowersystemlumpedtogetherbothWFandconventionalgeneration. Studyingreliabilityevaluationisraisingsomenumericalreliabilityindices.Thereliability testsystem(RBTS)isusedtoexaminetheofttypesofwindturbineson theseindices.windplaysatrolewhenstudyingWFbyseasonalRBTS. Forexampleinseasonswhichwindspeedisquitegood,theloaddemandislow,andonthe otherhandinseasonswhenloaddemandishigh,windspeedislow.Thus,unfortunately annualpeakloadtakesplaceinseasonswhenwindspeedislow.Reliabilityanalysismust shedlightforRBTSthatconsidersseasonal-hourlypatternofwindspeed.TheRBTScanbe 67 usedintmodelsbecauseitimprovesassessmentofwindfarmreliabilityevaluation [40].Someofthebasicindicesingeneratingsystemadequacyassessmentare: Lossofloadexpectation(LOLE):istheexpectednumberofdaysorhoursinasp period[40]. Lossofenergyexpectation(LOEE);istheexpectedunsuppliedenergyduetogenerat- inginadequacymorecomplexindexandisacompositeofthefrequency,durationand magnitudeofloadloss[40]. Expectedloadnotsupplied(ELNS):expectedvalueoftheload(powerdemand)not supplied.Hence,itisapowerfromthedimensionalviewpoint[41]. Expectedenergynotsupplied(EENS):expectedvalueoftheenergynotsuppliedby thepowergenerationsystemwithrespecttothatdemandedbytheloads[41]. Lossofloadprobability(LOLP):probabilitythattheloadsarenotsupplied[41]. Theperformanceofthewindfarmwasmeasuredbythereliabilityindices[39] Installedwindpower(IWP):thesumofthenominalpowerofallturbinesofthewind farm. Installedwindenergy(IWE):representstheenergythatcanbeextractedinoneyear fromthewindfarm. IWE = IWE 8760(4.18) Expectedavailablewindenergy(EAWE):amountofenergythatcanbegeneratedin oneyearwithoutconsideringfailureofwindturbines. 68 Expectedgeneratedwindenergy(EGWE):amountofenergythatcanbegeneratedin oneyear,consideringthefailureofwindturbines. Windgenerationavailabilityfactor(WGAF):similartotheloadfactorofconventional plantsbutconsideringtheofthefailureofwindturbines. WGAF = EGWE IWE (4.19) Besides,thecapacityfactorofthewindfarmcanbecalculatedwithoutconsidering theofthefailureofwindturbines,justthewindavailability. FC = EAWE IWE (4.20) Generationratioavailability(GRA):anavailabilityindex,itisproposedtoevaluatethe electricalsystemofwindfarms(OWF).TheGRAistheprobabilitythatatleasta certainpercentofwindpowercouldbetransferredtothegridsystemthroughtheconcerned electricalsystem.TheGRAdoesnotdependontheloaddemandandhasweakercorrelation withthewindspeedincomparisonwithotherreliabilityindices,suchasLOLPandEENS [129]. 4.5FactorsforWindFarmReliabilityAssessment factorsthatncereliabilityassessmentofwindfarminclude[23]: Windspeedrandomnessandvariability; Windturbinetechnologyandtopology; 69 Powercollectiongrid; Gridconnection environment; twindspeedswithintheinstallationsite; Hubheightvariations; Wakeectsandpowerlosses; Correlationofoutputpowerfortwindfarms; ofpenetration. 4.6WindFarm Nowadays,manycoastalstatesaredevelopingplanstobuildwindfarmsinfavorofgoing wherewindsarestrongandlanduseisrelativelyt.Itcanbeestimated thatthecostforeoperationistentimesthecostofthesameoperationforon-shore windturbine[32].Inordertositeanwindfarm,variousissuesneedtobe considered,whichinclude:windspeedsanddirection,waveheights,correlationofwindand wavedata,tidaldata,waterdepths,cableconnections,closenesstoload,seabedgeology, adjacentshorelineuse,dumping,existingshippingmovements,andperditionfromnational andlocalplanningauthorities,again,availabilityofwindrecourses,highspeedandmore uniform,homogeneousandsystematicwaves.Windconditionstendtoimproveasthedis- tancefromshoreincreases.windfarmshavemanyadvantages.Minimalturbulence 70 leadstoincreasedlifetimeofturbinescomponents.Theexistenceofansub- station,wherethevoltagelevelissteppedupbeforebeingtransmittedtoshore,dependson thesizeofthefarmandthevoltagelevelatthepointofcommoncoupling(PCC)whichis consideredasthegatebetweenwindfarmworldandpowersystemworld,thatislocatedon land.Inadditiontocablesbetweenwindturbinesareeasilyidenbyand havenonegativeimpactsonwildlife.Themarineenvironmentcanthefailurerate and(MTTR)whichlargelyincreasesduetotoughweatherconditions.Inaddition,noise levelisnotasbigissueWFasonland. Ontheotherhand,windfarmsrequiremoreareaperMWhthanmanyotherelectricity generationtechnologies.Highcostisassociatedwithconstructionandmaintenance.Some- timestheseabedisfarawayfromthecenterofloads.ceofrepairtimeincreasesaccess timerequiredfortransport,inspection,andrepairduetotransportvehicleandweathercon- ditions.ThusItiscostlytosendcrewsbyhelicopteroraboattorepairfailuresandperform scheduledmaintenancesinthenacelle,Thisworkoutthereareshortageinstudiesre- gardingwindfarmtechnology,suitabletopology,conventionalwindgeneratortype, developedpowerelectronicreliabilitydesign,optimizationelectricalcomponent,protection andearthingsystems,foundationdesign,marinecables,insulations,tofthere- quirementsoftheTSO?srelatedtosystemsecurity.Alloftheseissuesmustbetakeninto accountalongmarineenvironmentssuchas:humidity,corrosion,weather,waves,tide,sea surface,waterintrusion,whicharemorecomplicatedthanonshoreenvironmentbutnodoubt ithasbrilliantfuture. 71 4.7TechnicalAspectsofIntegratingWindFarmsinto PowerSystems large-scalewindfarmwillhavetimpactontraditionalpowergrid.Sothereisa needtoexplorepowergridandwindturbinepower,securityandstabilityissues. ThemainaimofthissectionistotheoftheWECSpenetrationlevel.The designerdecisionistomatchWTGandwindfarmfromonesidewithpowersystemfrom anotherside.Thedecisionisnotonlydependonreliabilitylevelbutalsodependsonthe economicparametersofthesystem.Asaresult,thetotallife-cyclecostsinstallationcosts, theoperationalandmaintenancecosts,playsimportantroletoachievehighlevelofreliability. Duetotheburgeoningandhighpenetrationlevelofthewindpowertechnology,windfarms areaskedtomeetgridrequirements.Consequently,somegridcodeshavebeenedor mo Largewindfarmsmaycontributetpowertothegridsandplayanimportant roleinpowersystemoperationandcontrol.Thetermwindenergy\penetration"refersto thefractionofenergyproducedbywindgenerationunitscomparedwiththetotalavailable generationcapacity.Inrealitythereisnogenerallyaccepted\maximum"levelofwindpene- tration.Thelimitdependsonmanyfactorssuchas:theexistinggeneratingplantsinservice, pricingpolicy,capacityforstorageunits,systemplanning,loadforecasting,powerelectronic equipmentsanddemandmanagement,etc.Integratingwindenergyintopowersystemgrid canbechallengingduetoafewmainissuesofgridrequirementsthatwindturbineshaveto meet. 72 4.7.1ActivePowerControl Activepowercontrolisrequiredinordertolimitoverproductionofwindpowerthatcan causegridinstabilities.Sinceelectricpowervariesascubeofwindspeed,ratevariationcon- trolrequiresoperationatreducedoutputpowerlevelsduringconditionswhenwindspeed varies.Thewindfarmscanbecontrolledtoprovidefrequencyresponse,withinthelim- itationsoftheavailablewindpower.Activepowercontrolisusedtocontrolthesystem frequencybychangingthepowerinjectedintothegrid.Thereliableoperationandbalance onthetransmissionsystemismaintainedbyuseofthesystemfrequency.Frequencyinthe powersystemisanindicatorofthebalancebetweenproductionandconsumptionorbetween supplyanddemand.Forthenormalpowersystemoperation,thefrequencyisclosetoits nominalvalue.Windturbinetechnologiesallowitsparticipationinfrequencyregulation. Forbothcases,controlsforprovidingresponseagainsthighfrequencyareiveandare enhancedbytheintermittencyofthewind.Onthecontrary,controlsforprovidinglow frequencyreservesmaybenegativelyimpactedbywindvariations.Oneofthebestwaysto improvethesituationistoelyutilizethespeedofresponseofenergystoragetech- nologysuchasbatteries,pumpstorage,fuelcellsandwheel.Fastfrequencyresponse servicescanstabilizethegridandcanprovidemorereliablereservestolowfrequencyevents. Inconventionalgeneratingunits,thereliableoperationoffrequencycontrolisnormallypro- videdbyfastautonomouscontrolofindividualturbinegeneratorsthroughgovernorcontrol, automaticgenerationcontrol(AGC). 73 4.7.2ReactivePowerControl Asmentionedearlier,whenwindspeedsgobeloworexceedcertainvaluesorsystemsgo undercertainseveredisturbances,windfarmcouldlosegridsynchronism.Consequently bothrealpowerandreactivepowerarelost.Reactivepowerisusedinordertokeepthe voltagewithintherequiredlimitsandavoidvoltagecollapse.Windgenerationshouldalso contributetovoltageregulationinthesystem.Oneofthebigchallengesisthatduetohigh penetrationlevelofwindfarminpowersystemthedisconnectingofwindfarmduringshort circuitsfaultsorgridinstabilitypresents,abigriskforthesystemstability.Therefore,the provisionandplanningofreactivepowerbecomeextremelyimportant.Newregulations requirethatwindfarmsremainconnectedduringalinevoltagefaultandparticipatein recoveryfromthefault.Sincewindturbinesareresponsibletoimproveloadperformance, theymustbeabletosupplyreactivepower.Traditionalapproachestomanagingreactive powerarenolongeracceptable[105].Thereactivepowercanbecontrolledinwindturbines withPWMconverters.Variousdynamicreactivedevicessuchasstaticvarcompensatoror synchronouscondenser,areusedforgridconsolidationwheredynamicvoltagesupportis required. 4.7.3VoltageFlickers Inwindpowerplantsvoltageishighlyexpected[29][130].Lackofpowerquality causesaconsumerinconvenienceandcomplaints.Windfarmsconnectedtoweakpower systemgridhavepronouncedvoltageker,whichisembodiedbytheofvoltage amplitudeduetowindgustingnatureandtowershadowisasourceofpowerat thefrequencyatwhichtherotorbladespassthetower.Alsovariationofactivepowerand 74 reactivepoweratthePCCwiththegridgivesrisetoker.Thewindturbineoperating pointandthe Q-P characteristicofthegeneratordeterminethepointofminimumker emission. Thekerphenomenonisevaluatedbythekermeter,describedinIEC61000 4 15 [131].Highcapacitywindturbinescauseseriouskerproblem,whichleadstoshortthe lifespanofwindfarmelectricalandelectronicequipments.Thekerisby thegridstrengthand X=R ratioofgridinternalimpedance.Therotatingwindturbine bladesthroughconstrainedcancastshadowstothewindowsofneighbouringproperties,a phenomenonknownasshadowker,whichhasbeenusedbysomewindfarmopponent groupsasareasonforimpedingnewdevelopmentsasshowninFigure4.2[132]. Figure4.2Shadowkerevidencebase. 4.7.4Harmonics Harmonicsisoneofthepowerqualityissuesofwindfarms.Knowledgeofthewindfarm harmonicpollutionbehaviorisfundamentaltostudytheofthesefarmsonnetwork harmonicdistortion.Theassessmentofthewindfarmharmonicspectrumandtheanalysis oftheofoperatingpointinWECSareveryimportant.Thecombinationofwind 75 turbinesandpowerelectronicsexcitevoltagedistortioninnetworks[133]duetothelow switchingfrequencyofhighpowerconverters,controlsystemimperfection,generatorsand transformersnonlinearities[134].HarmonicsmeasurementprocedureforindividualWECSis describedinIEC61400 21[135].Theharmoniccurrentssummationofindividualwindfarm unitsareusedforharmonicemissionevaluationinwindfarms[135].Itisworthmentioning thattherearefewstudiesthatapplybothprobabilisticanddeterministictechniquesto actualwindmeasurements. 4.8WindFarmTopologies Thewindfarmischaracterizedbyaspecinumberofturbinessimultaneouslyconnected thepowersystemwithoutcing.Turbinesinterconnectionwithinawindfarmisbased onACorDCtopologies.Asuitabletopologyandvoltagelevelofthenetworkshould bechosenforoptimal,reliability,maintainabilityandcosteness.Thetotal powerproducedbywindfarmasawhole,thepowerproducedbyeachindividualwind turbine,andthedistancetoshoreandtheindividualdistancebetweeneachwindturbine arethemainfactorsthatareconsideredindesigningwindfarmtopology. 4.8.1ACTopologies 1.ACradialconnection:Theradialconnectionsystemiswidelyspreadanditiscon- sideredasthesimplestwayofconnectionwhereanumberofturbinesareconnected tothesamefeeder.Subsequently,manytfeedersareconnectedtogetherata substationthatcollectspowerfromtheentirefarm.Thesystemvoltageisstepped upandtransmittedtothegrid.Themainfeaturesarelowcostandsimplecontrol. 76 Ontheotherhand,highlossesandlackofredundancy,poorreliabilityarethemain disadvantages. Figure4.3ACredialsystem.[1] 2.ACsingle-sidedring:Anumberofturbinesareconnectedtothesamefeederfrom thebothends,whichmeansonemorecableisconnectedfromthelastturbineina row.Thiscableshouldbeabletocarrythecumulativepower.Althoughitismore expensive,thesingle-sidedringtopologywillhavelesslossesandbemorereliable. 3.ACradialloopconnection:Inthisconnection,thepowergeneratedinafaultedfeeder maybesuppliedbytherestoffeeders.Themainfeaturesarehighdegreeofreliability andlowlosses,althoughitisconsideredhavinghighcost.Thecontrolsystemcould becomplicateddependingonnumberofswitchesandtheirlocations. Figure4.4ACredialloopsystem.[2] 4.ACstar/clusterconnection:Eachturbineisconnectedtothepointofinterconnection 77 ofstarpointinthistopology.Theadvantagesiscableratingisequaltoturbinecurrent ratingconnectedtoit.Themainprosarehighlevelofsecurityduetocomplexcontrol andprotectionforeacharm.Eachwindturbinehasitsownconnectiontotheplatform. Conclusionaroundcostcannotbedrawnunlesseachcaseisstudiedseparately.In somecasesofstarconnectionsmultipleclustersarerequiredinordertoreducecabling constructioncost. Figure4.5ACstarsystem.[2] 4.8.2DCTopologies ThecoreideaofDCtopologywindfarmsistoraisethevoltagetoconnectdirectlyto transmissionlineswithoutconverterstationbyseriesconnectionofWECS.Thereason behindsizingtheratingsofDCwindturbinesistogainfeaturesintermsof ofDCandreducelossesthrougheliminatinglargeconverterstationsBecauseit isattainableformaintenanceandrepairsandithashighlysecureconnection[102].The commonDCtopologiesincludethefollowing: 1.DCradialconnection:Thistopologythat,theturbineoutputpoweriseither MVDCorACandistransmittedasHVDC. 2.DCseries/daisychainconnection:Inthiswindfarmconnection,theaccumulative 78 DCisobtainedbyseriesconnectionofturbineoutputsinordertoincreasevoltageto transmissionlevel. 3.DCseries-parallelconnection:Inthistopology,windturbinesareconnectedin daisychainconnection.Inaddition,acertainnumberofseriesconnectionsarecon- nectedinparallelinordertoincreasewindfarmcapacitypower. Basedonexperience,theradialsystemconnectionisthebestamongtheothersand mostpopularduetorelativelowlossesandcostwithoutreliability[37].Ifthe priorityiscost,thenthewindfarmownermaychoosedcseriestopology.ACredial topologyoutperformsothersintermsofisbetter[136].Asageneralcomparisonbetweendc andacoptions,moreadvantagesaregainedbyselectingadcseriestopology,whichisbased onbothreliabilityandcost.Thecostcalculationincludesbothequipmentsandinstallation andneglectedotherfactorssuchastaxes,interestratesandsoon. 4.9WindFarmLosses Lossesingeneraldependonloadwcalculation.inwindfarms,thelosseswillvarywith windspeed[137]andcanbedividedintotwomaingroups;turbinelosses,andcollection andtransmissionsystemlosses.Theturbinelossesinturncanbedividedintogenerator, converterandtransformerlosseswhilethecollectionsystemlossesconsistsofohmiclosses. Thetransmissionlossesincludetheconverterandtransformerlossesalongwiththe transmissioncablelosses.Electricalandcontrolsystemlossesplayarelevantroleinthe assessmentofoutputpower,dependingonsizeanddesignofthewindfarm.Thesetwo elementsreducethetotaloutputofthewindfarmandtheymightbeincludedinacomplete model.Thesimplestsolutionusesancotwhichdependsonwinddirection, 79 numberofwindturbines,theirspatialarrangementandpowercollectiongriddesign.Ifplan- nersofenergylimitedsystemswanttoexploitcurrentsystemstotheirmaximumcapacity andextendtheuseofsystemcomponentsbeyondtheirusefullife,theymustbeawareof lossescalculationssincecablesinsulationswhichreliability.Thusadetailedmodeling isrequiredbecausetheresultsobtainedusingadonearemoreconservativeand couldindicatetheneedofsystemupgradingbeforenecessary. Factorsthechoiceofvoltagelevelare: 1.Operationalcosts; 2.Capitalcosts; 3.Voltageinsulation; 4.Converters; 5.Networkingfactorsandcablelengths. 4.10Challenges Theintegrationofwindpowerfarmintothegridfacessomechallengesfrombothoperation andsecurityaspectssuchthat: 1.Trippingintransmissionlinesbecauseoffaultssinglephaseandthreephases; 2.Varyingwindspeedcauseoutputvoltageons; 3.Lackofcontrolvoltageregulation; 4.Inabilitytocontrolactivepowerproduction; 80 5.Shortageofreactivepowercontrol; 6.Maximumpowertransferwherepowerelectronicsthatplayimportantrolemaylimit thepowerextraction.Atthesametime,thepowerelectronicshelpsprotectthesource fromsuddenloadchanges; 7.Powerquality 8.Harmonicdistortionmustberemediedbycontrolalgorithmsuchasresonant 9.Lackofhistoricalrecordsdataforwindresources. 4.11OperationandMaintenancePlanning InMIL-STD-721maintainabilityisasthemeasureoftheabilityofanitemtobe retainedinorrestoredtospconditionwhenmaintenanceisperformedbypersonnel havingspskilllevels,usingprescribedproceduresandresources,ateachprescribed levelofmaintenanceandrepair.Recentlymanystudiesaredevotedtoidentifytheoptimal OMstrategiesthatwillovercomethehighcostofunexpectedfailures.Themaintenance costsamounttoaround30%ofthetotalenergygenerationcost[136].Therefore,researchers mustshedmorelightonmaintenanceproceduresanddevelopments.Generally, OMplanningmightbeintofourcategories:correctivemaintenance,preventive maintenance,predictivemaintenance,andproactivemaintenance. CorrectiveOM(COM)isperformedafterthefailureeventhasbeenobserved.Itis referredasreactive,breakdown,repair,orrun-to-failure(RTF)mainte- nance.Itmayincludeoneorsomeorallofthisgrouplocalization,isolation,disassem- bly,exchange,reassembly,alignment,checkout.Themainprosarelowmaintenance 81 costduringoperationandmaximumlifeuseofcomponentswhilethemainconsare consequentialdamgescauseextensivedowntimes. PreventiveOM(POM)isimplementedwhilethefailureeventisnotobserved.Itmay includeoneorsomeorallofthisgroupregularlyscheduledinspection,adjustments, cleaning,lubrication,partsreplacement,calibration,andrepairofcomponentsand equipment.Themainprosaredowntimeisnotlong,scheduledprocess,whilethe mainconsareexpensive. Predictivemaintenanceorcondition-based,maintenanceactivitiescompriseequipment testsbasedontheuseofon-lineandsensorsandtests.Itusesprimarily non-intrusivetestingtechniques,visualinspection,andperformancedatatoassess machinerycondition.Datacollectedareusedinoneoffollowingwaystodetermine theconditionoftheequipmentandidentifytheprecursorsoffailure.Themethods ofanalysisincludetrendanalysis,patternrecognition,datacomparison,testsagainst limitsandranges,correlationofmultipletechnologies,statisticalprocessanalysis. ProactiveMaintenanceisdesignedtoextendthelifeofmachineryassupposedto: { Doingrepairswhenoftennothingisbroken; { Accommodatingfailureasroutineandnormal; { Preemptingcrisisfailuremaintenance. Further,POMmightbeperformedbasedonusageage,periodicallyscheduled,condition basedandriskbasedmaintenancestrategies.TodetermineanoptimalOMstrategy,the objectivefunctionsshouldbedeterminedbyminimizationormaximizationduringtheservice lifeortimehorizon,subjecttothemodellimitations.Objectivefunctionstobe 82 minimizedmightbeedbasedoncostsordowntimes,whereasobjectivefunctionstobe maximizedcouldbebasedon(boravailabilities. 83 Chapter5 Real-TimeOptimizationofThermal CyclingCapabilityofRotorSide ConverterinDFIG-BasedWECS 5.1Introduction Reliabilityistheprobabilitythatacomponentwillsatisfactorilyperformitsintendedfunc- tionundergivenoperatingconditions[138].Theaveragetimeofsatisfactorilyoperationofa systemiscalledthemeantimebetweenfailures(MTBF)[138]andthehighervalueofMTBF indicateshigherreliabilityandviceaversa.Nowadays,reliabilityismoreofconcernthan inthepastespeciallyforwindturbinessincetheaccesstowindturbinesin caseoffailuresisbothcostlyand[3].Powersemiconductordevicesareoftenranked asthemostfragilecomponentsinapowerconversionsystem[139].Thelifetimeprediction ofpowerIGBTmodulesbasedonmissionisanimportantissue.Furthermore,life- timemodelingoffuturelargewindturbinesisneededinordertomakereliabilitypredictions aboutthesenewwindturbinesintheearlydesignphase.Byconductingreliabilitypredic- tioninthedesignphaseamanufacturecanensurethatthenewwindturbineswilloperate withindesignedreliabilitymetricssuchaslifetime. 84 Thisworkpresentsreliabilityanalysisofpowerelectronicconvertersforwindenergygen- erationsystemsbasedonsemiconductorpowerlossesaswellasaimstomaximizesemicon- ductorlifetimeusinglo(LPF)basedcontrolschemesinceMTBFwillbehigher thanwithThefundamentalcausetoachievehigherMTBFliesinthereductionofto reducethenumberofthermalcycles. Thekeyelementinapowerconversionsystemisthepowersemiconductordevice,which operatesasapowerswitch.Theimprovementinpowersemiconductordeviceisdrivinga criticalforcebehindtheimprovedperformance,cy,sizeandweightofpowerconversion systems.Asthepowerdensityandswitchingfrequencyincrease,thermalanalysisofpower electronicsystembecomesimperative.Theanalysisprovidesinformationonsemiconductor rating,reliability,andlifetimecalculation.AcomprehensivethermalmodelforpowerIGBT isdevelopedinthreesteps[140]: i )thepowerlossesarecalculated[141],[142]; ii )thejunction temperaturesareevaluated[143],[51];[144],[145],[52],and iii )thelifetimeisestimated [49],[50]. 5.2PhysicalSystemModeling 5.2.1WindTurbineCharacteristics Mappingamissioninwindpowerapplicationsinvolvesmultipletimeconstants.The timeconstantrangefromshortterminelectricalcomponents,tomediumterminmechanical components,tolongterminwindspeedandmuchlongerterminsystemreliability.Wind turbinescapturepowerfromthewindbymeansofaerodynamicallydesignedbladesand convertittorotatingmechanicalpower.Thenumberofbladesistypicallythreeinamodern windturbine.Formulti-MWwindturbines,therotationalspeedistypically10 15 rpm 85 .Themostweightcientwaytoconvertthelowspeed,hightorquemechanicalpower toelectricalpoweristouseagearboxandastandardgeneratorwithapowerelectronic interface.Theenergyproducedfromthewindturbinesconvertinganddependsonwind speeds.Atlowwindspeedsaround1 3 m=s thewindturbinewillnotfunction.At\cut inwindspeed"2 : 5 5 m=s thewindturbineswillstart.Thewindspeedrangeofabout 12 15 m=s iscalledthe\nominalorratedwindspeed",withwindturbinesworkingon theirfullrange.Athighwindspeedsoverthecutoutspeed25 m=s ,thewindturbinewill bestoppedduetopotentialdamagetothewindturbinesbladesandtowerstructure.The outputpower P m isdependentonthepowercot C p .Itisgivenby[146]: P m = 1 2 ˆR 2 ˛ 3 C p ( )(5.1) andthetipspeedratioisas: = ! t R ˛ (5.2) where ˆ isspairdensity; R isradiusoftheturbineblade, ˛ isthewindspeed. ! t is turbinerotationalangular; C p isthecotofpowerconversionand isthepitchangle. Thepowercocient C p ( )isfurtherexpressedas[146]: C p ( )= c 1 c 2 1 c 3 c 4 e c 5 1 + c 6 (5.3) wherethecots c 1 through c 6 dependontheshapeofthebladesanditsaerodynamic performanceofwindturbine.And 1 isas: 86 1 1 = 1 +0 : 08 0 : 035 3 +1 (5.4) 5.2.2Doubly-FedInductionGenerator TheDFIGisconsideredasoneofthemostpopulartopologiesappliedinWECS.Itsmain advantageistoadjustthespeedoflargesystemwithpowerconvertersofathirdoffull rating.ThisisbecauseitsRSCoperatesunderslipfrequency.Onlyneedstosupportslip powertotheoverallsystem.Thesemiconductorpowerdevicessuchasinsulatedgatebipolar transistors(IGBTs)intheRSCandgridsidecontrol(GSC)aresusceptibletopowercycling caused.DFIGallowsforcontrolsfromrotorsideaswellasstatorside. Thewindturbinerotorisconnectedtothegeneratorviagearbox.TherotoroftheDFIG isfedbyback-to-backvoltagesourceconverter(VSC).Thegeneratorspeedcorresponding toratedwindspeedcanbesetatanypointbythechoiceofgearboxratio.Thebacktoback allowsforindependentcontroloftheRSCandtheGSCduetothedecoupling providedbythedc-linkcapacitor.TheRSCoperatesthespeedandtorqueofthegenerator whiletheGSCcontrolstheactiveandreactivepowerinjectedtothegrid.Anwind farmequippedwithpowerelectronicconverterscanperformbothactiveandreactivepower control.Itoperatesthewindturbinesinvariablespeedtomaximizetheenergycaptured. Besidesitreducethemechanicalstressandacousticalnoise. TheoverallcontrolstructureofaDFIGconsistofcascadedcontrolloops:innercontrol loopandoutercontrolloopasshowninFig.5.1.Theinnerloopprovidesadequatedecou- plingbetweenactiveandreactivepowers.Thedesignedcontrollawsaretotrackthepower commandofinthewindgenerator.Theouterloop,theerrorsignalsfeedingitsPIcontrollers 87 Figure5.1Systemunderstudy.[3][4] 88 areobtainedbysubtractingthereferencepowers P s ref or Q s ref fromtheiractualvalues P s or Q s .ThefollowingequationsdescribethedynamicmodelofDFIG[147]: v ds = r s i ds + ds dt ! qs (5.5) v qs = r s i qs + qs dt + ! ds (5.6) v dr = r r i dr + dr dt ( ! ! r ) qr (5.7) v qr = r r i qr + qr dt +( ! ! r ) dr (5.8) ds = L s i ds + L m i dr (5.9) qs = L s i qs + L m i qr (5.10) dr = L m i ds + L r i dr (5.11) qr = L m i qs + L r i qr (5.12) T e = 3 2 pL m ( i qs i dr i ds i qr )(5.13) where: L s = L ls + L m (5.14) L r = L lr + L m (5.15) TheparametersofthesystemunderanalysisarelistedinTable5.1. ThemodelofDFIGhasbeenbuiltfromscratch.Figure5.2,showthe DFIG performance instationary,synchronousandrotorreferenceframes. 89 (a) V ds & V qs insynchronousreferenceframe. (b) Vabc s insynchronousreferenceframe. (c) Vqr & Vdr inrotorreferenceframe. (d) Vabc s inrotorreferenceframe. (e) I ( abc ) s & I ( abc ) r inrotorreferenceframe. (f) V qr & V dr instationaryreferenceframe. Figure5.2VariablesofthreephaseDFIGinstationary,synchronousandrotorreference frames. 90 Table5.1:DFIGElectricalParameters Power, P n 7 : 5 KW StatorVoltage, V n 415 V Frequency, f n 50 Hz Statorwindingresistance R s 1 : 06 Statorwindinginductance L s 0 : 2065 H Rotorwindingresistance R r 0 : 8 Rotorwindinginductance L r 0 : 081 H Magnetizinginductance L m 0 : 0644 H Inertia J 7 : 5 kgm 2 Polepairs 3 Ratedspeed 970 rpm 5.2.3AveragedModelof (PWM)converter Theaveragedmodelofthree-phaseback-to-backPWMconverteriswidelyadoptedinprac- ticalengineeringbecauseitissuitableandtforsimulationofcontrolsystem.Itis mainlyusedtoreplacetheswitchingcircuitsinsimulationenvironmenttoreducethecompu- tationalcomplexity.Thesmallsignalmodelforcontrollerdesignareconventionallyderived fromtheaveragemodel.Inthemeantime,theaveragedmodelissuitableforevaluationof powerlossesofaconverter. Oneofthecommonlyadoptedapproachtoformulatingtheaveragemodelofthesystemis toreplacetheinstantaneouslyswitchedregionswithinaveragemodels.Theaveragemodel oftheswitchesistypicallyrepresentedbythecontrolledcurrentandvoltagesourcewith thedutyratioormodulationfunctionasacontrolinput.Inaveragedcircuitmodel,the complexityisreducedandleadstofastertime-domainsimulationwhileadequateaccuracy ofthesystemdynamicsisstillmaintained.Furthermore,over-modulation,saturation amongstothernon-linearitiescanalsobeproperlymodeled.Althoughtheaveragemodelis 91 notsuitedforanalysisofswitchingfrequencyripple,Itissuitableforevaluationofthepower lossesofthesystemundervariousloadingconditions. 5.3ElectrothermalModelingandLifetimeEstimation oftheVoltageSourceConverterforWindTurbine Thepowerlossmodel[148]hasbeensuccessfullybuilttoperformthethermalanalysis.The modelisbasedonlookuptablethatincludesboththeswitchingandconductionlosses.The modelparametersofthethermalnetworkareextractedfromthedatasheetABBHiPack IGBTModule5SNE0800M170100.Theexactmodelingapproachdescribesaconvertersas atimevaryingsystem.Itistypicallynotapplicableforcontroldesignpurpose,becausethey aretoanalyzeandimpracticalforthesimulationoftherelativelylargesystems.In addition,simulationtimestepshavetobemuchsmallerthanswitchingperiod,whichleads toaverycomputationallyexpensivesimulation.Thus,anaveragedmodelismoresuitable forelectrothermalsimulation.Itcanbeusedtocalculatethesemiconductorlossesatany outputcurrentwaveform.Inaddition,itcanbeparameterizedwithconventionaldatasheet information.Forcalculatingtheinstantaneousjunctiontemperatureofthesemiconductor deviceunderjunctionsattloadconditionsathermalmodeloftheinvertersub-system isnecessary.Astraightforwardapproachistouseanetworkofthermalcapacitance C th and thermalresistance R th .Suchnetworksarereadilyimplementedinvariousprogrammingor simulationenvironments. TheprocedureforcalculatingtheestimatedlifetimeforsemiconductorsdevicesinWECS isillustratedinFig.5.3.Lifetimeestimationstartsfrompowerlossescalculationwhichis basedonthelookuptablemethod.AnequivalentRCnetworkmodelhasbeenbuiltto 92 performthethermalanalysis.Thejunctiontemperaturesandthermalcyclecountscanbe accordinglydetermined.ThelifetimeissubsequentlypredictedaccordingtoMiner'srule [149].Therealtimesimulationenvironmentdictatestherequirementsforthemodels.Easy implementationonthesoftwareplatformSimulink/Matlabandfastcalculationtime. Figure5.3Lifetimeestimationmodelforpowersemiconductordevices.[5] 5.3.1PowerLossesofIGBTintheRSC Therearethreetypesoflossesinpowerdevices:conductionlosses,switchinglosses,and gatelosses.Increasingtheswitchingfrequencywillincreaseswitchinglosses.Switching lossestypicallydominatetotalpowerlossesinhighswitching-frequencyPWMconverters. Thegatelossesisleftoutasitistascomparedtotheswitchingandconduction lossesofthemainpowercircuit.Thecurrentandvoltagetransientwaveformsarenotalways available,whicharenecessaryisaprerequisiteforcomputingthepowerlosses.Therefore, simulationandanalyticallysolutionsbecomeimportant. tapproachesforthecalculationofswitchinglosseshavebeenpublishedinlitera- ture.Ifthenecessaryparameterswithtemperaturedependenciesareknown.theswitching lossescanbeestimatedbypiecewiseapproximationofthetransientcurrentandvoltage waveforminthesemiconductors[142],[148].Therearenumerousothermethodsusedtoes- timatetheswitchinglosses,suchasexponentialorpolynomialapproximationfunctionswith 93 parametersempiricallyobtainedorextractedfromdatasheet. Themethodusedinthisstudyisbasedonlookuptablesusuallythemanufacturerof powersemiconductordevicesprovidethetransientthermalimpedancecurveortableforthe IGBTanddiodeinsideamodule. P IGBT = P cond + P sw (5.16) InpowerconvertersIGBT,theconductionlossesdependsonthreeparameters:thecur- rentthroughthedevice I c ,theon-stagevoltageacrossit V CE ,andthejunctiontemperature T j .UsingtheDCcharacteristicsextractedfromexperimentorsimulation,theon-stage voltage V CE canbeexpressedasafunctionofthecurrentthroughthedevice I c ,whichleads totheconductionlossesexpressedas: P cond = f ( I c ;T j )(5.17) Forswitchinglosses,itisassumedandvthattheswitchingenergylosseslinearly dependsontheswitchedcurrent.ThedependencyoftheswitchinglossesontheDClink voltage V DC andthejunctiontemperature T j ,canbeextractedfromexperimentaldataor simulationresults.Ingeneral,theempiricalrelationbetweentheswitchinglossesandthree parameterscanbefoundbycurvetingmethods[51],[52]. P sw = f ( I;V DC ;T j )(5.18) Afastsimulationmodelforestimatingpowerlossesofathreephaseconverterhasbeen proposedinthispaperbasedonaveragemodel.Largersimulationtimestepsallowfor 94 powerlossesandthermalperformanceofanconvertertobepredictedoverlongperiods time.Thissimulationmethodologybringstogetheraccuratemodelsoftheelectricalsystems performance.Thespeedupisobtainedbysimplifyingtherepresentationofthreephase converteratthesystemmodelingstageusinglargetimestepof100s. LossesinIGBTs: Turnonloss: Pre-switchingvalueofthevoltageacrossthedevice,postswitchingvalueofthecurrent wingintothedevice,andthejunctiontemperatureareusedtodeterminetheenergy losswiththehelpof3Dlookuptable,thisenergyisconvertedintoapowerpulsewhich isinjectedintoathermalnetwork. Turnloss: Pre-switchingvalueofcurrentwingintothedevice,postswitchingvalueofthe voltageacrossthedevice,andthejunctiontemperatureareusedtodeterminethe energylosswiththehelpofa3-Dlokuptable.Thisenergyisconvertedintoapower pulsewhichisinjectedintothethermalnetwork. Conductionloss: Valueofthecurrent( I c )winginthedeviceanditsjunctiontemperaturedetermine whatwouldbethesaturationvoltage( V ce )acrosstheIGBTusinga2-Dlook-uptable. The V ce isthenmultipliedby I c toobtainthelosseswhichareinjectedintothethermal network. 95 5.3.2ThermalModelingTechnique Inliteraturereview,erentapproacheshavebeenusedtoperformthermalanalysis.They includeanalytical,numericalandbehavioralmodels.Inparticularapplications,theIGBT modulesinpowerconverteraremountedtoheatsinksthatareairorliquidcooled.Con- ductionamongtlayersofmaterialsisthemainmechanismofheattransferalthough theheatgeneratedinsidetheIGBTmodulecanalsodissipatedbyconvectionorradiation. Analyticalmethodshavelongbeenusedinthermalanalysistopredicttheoperatingtem- peratureofsemiconductordevices.Thesemethodsprovidebetterphysicalinsightbyuse ofphysicalmodel.Variousassumptionsfortheboundaryandinitialconditionsaremade inordertosolvetheheatconductionequation[145].Numericalmethodsarealternatively employedforthermalanalysis.Finiteelementmethod(FEM),ormethod (FDM)areusedfordiscretizationofthetialequationforheatconduction.Compu- tationaldynamic(CFD)isusedtosolvetheequationsgoverningtheconservativeof mass,momentumandenergy[144].RCladdernetworksarecommonlyadoptedforthermal analysis,sincetheyarereadilytobeintegratedintoexistingcircuitsimulatorandarecapa- bleofsimulatingbothelectricalandthermalcharacteristicscircuits.RCthermalmodelis selectedinthisstudyforrealtimesimulationduetoitseasyimplementationandreduced computationalcomplexity[143].AnRCthermalnetworkforIGBTisbuiltasshowninFig. 5.4andTable5.2. Table5.2:IGBTthermalcharacteristicvalues. i 1 2 3 4 R i ( K=kW ) 15 : 2 3 : 6 1 : 49 0 : 74 ˝ i ( ms ) 202 20 : 3 2 : 01 0 : 52 OncethepowerdissipatedinIGBTpowerconverterhasbeenobtained,thispowerisused 96 astheexcitercurrentsourceinathermalmodelbuilttoestimatethejunctiontemperature ofoperationalIGBTs.Thethermalmodelwillbeincorporatedinarealtimesimulator.The valuesofthethermalimpedancewouldbeextractedfromthedynamicalthermalimpedance curvefromexperiment,simulation,manufacturers'datasheet.Themodeladoptedinthis paperco-simulatesthethermalandelectricalperformanceofthesystem.Thetemperature ofthedevicevariesaccordingtoreal-timeoperating.Theelectro-thermalmodelhasbeen builtinthreesteps: i )theelectronicmodelofthedeviceisdeveloped; ii )thethermalmodel isbuilt;and iii )acouplingbetweenthetwomodelsisestablished.Thecouplingcanbe implementedbysettingthetemperatureintheelectricalmodelsasthestatevariable.The device'stemperatureisdeterminedbyintegrationineachsimulationstep.Selfheating istakenintoaccountthroughtemperaturedependentparametersthataremobythe device'soperatingtemperature.TheFosternetworkisusuallyreferencedindata-sheets. Figure5.4FosterThermalImpedanceBetweentheJunctionTemperatureandCaseLayer. Theoperatingtemperatureplaysamajorroleforperformanceandreliabilityofsemicon- ductorsdevices.AsaconsequenceItisnotsurprisingthatthesafetymarginorreliability 97 ofasemiconductordevicesdecreasesasthetemperatureincreases.Byuseofthepower lossesascurrentsourcevalueinthecircuit,thejunctionandcasetemperaturecanbede- terminedfromthecorrespondingnodevoltages.Infrequencydomain,thepowerlossesand thetemperaturesarerelatedbythethermalimpedanceexpressedastransferfunctionasin (5.19). Z thermal = T ( s ) P ( s ) (5.19) Figure5.5CauerThermalImpedanceBetweentheJunctionTemperatureandCaseLayer. Table5.3:ThermalElectricalAnalogousQuantities AnalogousQuantities Thermal Electrical ThroughVariable Heattransferrate,P,watts Current, i ,amps AcrossVariable Temperature,T,KorC Voltage, V ,volts DissipationElement Resistance, R th , k=watt Resistance, R ,ohms StorageElement Capacitance, C th , sec:watt=k Capacitance, C ,farads 98 ThethermalRCcircuitsforIGBTanddiodearebuilt,usingthepowerlossesascurrent sourcevalueinthecircuit,thejunctionandcasetemperaturecanbedeterminedforcorre- spondingnodevoltages.Inthefrequencydomain,thepowerlossesandthetemperatures arerelatedbythethermalimpedanceexpressesastransferfunctionasinequation ?? Z th ( j c ) ( t )= X R i (1 e t=˝ i )(5.20) ˝ i = R i :C i (5.21) where P isthepowerlossesand T isthetemperature.Inthesimulink,iftheinputsignal andoutputsignalcanbeexpressedastransferfunctionastheformshowninequation5.23, theoutputcanbeobtainedbyconnectinginputsignaltothetransferfunctionblock: H ( s )= y ( s ) u ( s ) (5.22) = num ( s ) den ( s ) = num (1) S nn 1 + num (2) S nn 2 + ::: + num ( nn ) den (1) S nd 1 + den (2) S nd 2 + ::: + den ( nd ) (5.23) wherennandndarethenumberofnumeratoranddenominatorcocients,respectively, numanddencontainsthecotsofthenumeratoranddenominatorindescending powersof(s),ourgoalistothecotsinequation5.23basedonRCthermalcircuits forIGBTanddiode.Thermalresistor( RC )networkswidelyusedforthermalanalysis,the transientthermalimpedanceisastime t as: Z jc ( t )= T j ( t ) T c ( t ) P = T jc P (5.24) Z jc ( t )= n X i =1 i : 1 exp t ˝ i (5.25) 99 Thetransientthermalimpedancecurveisastepresponsecurvewithzeroinitialcondi- tions.Itiswellknownthatthestepresponseofalinearsystemcontainsthefulldescription ofthesystem.BeforeextractingofthermalRCnetworks's,thecurvemethodisap- pliedtotheexperimentaltransientthermalimpedancedatawhichresultinginseriesof exponentialtermsthatareprovidedbymanufacturerasinequation5.25.Thetransferfunc- tion(inputimpedance)ofthethermalRCnetworkisfoundedbyapplyinglaplacetransform to5.27. Z jc ( C )= n X i =1 i ˝ i s + 1 ˝ i (5.26) Inordertoderivethethermal RC networkparametersvalue,weneedtotransfer5.27inthe followingform: Z jc ( s )= 1 sC th 1 + 1 R th 2 + 1 sC th 2 + ::: + 1 R thn (5.27) 5.4LifetimePredictionandDesignofReliability Whendesigningacontrolstrategyforwindenergyapplications,themaincontrolobjective canbesummarizedas: Maximizeenergyproduction. Maximizelifetimeoperation. Minimizemaintenancecosts. Guaranteesafeturbineoperation. Thesafeturbineoperationismainlyduetowindspeedandensuredbylimitingthe angularspeedoftherotorshaftandbycoercetheoperationtheoperatingrangeofthe 100 turbinetosafelimitwithinthemaximumwindspeed25 m=s .Thismaximumisdetermined bythenoiselevelproducedbyrotatingblades,whichisrelated = ! t R ˛ andtheforcesacting ontheblades,tower,etc.Maximizingtheenergyproductioncanbeachievedbyobtaining theoptimaltipspeedratioforeachwindspeed.Operationandmaintenancecostscanbe minimizedbylimitingthedynamicsloadsactingonthemechanicalcomponents,thiscanbe doneby,forexample,capturingwindgustsintheinertiaoftherotoor.Theimportanceof eachobjectivedependontheoperatingpointofthewindturbine,rangesfromcutinspeed toratedwindspeedtocutoutwindspeed.Allarebeyondthescopeofthisthesisexcept ourmaingoal,maximizelifetimeoperation. ThelifetimeestimationofhighIGBTisestimationthelifeexpectancyofthedevicesun- dercertainoperatingconditions(stresses).FortheIGBTsthestressescanbetemperature, voltage,current,vibration,humidity,cosmicradiationlevel[106].Theaccurateassessment ofthereliabilityissueofwindpowerconverteriscriticalforlifetimeestimationandcost reductionforwindpowerapplications[150].Theconverterisoneofthemostunreliable subsystemsofanelectricalsystemoperatinginaharshenvironment[151].Thecostassoci- atedwithconverterfailureandcorrespondingunscheduledmaintenanceishighinthecase ofapplicationduetolimitedaccessibilityofthesystem.Thesystemreliabilitycan begreatlyimprovedbyreplacingdevicesbeforetheyfail.Therefore,lifetimepredictionof anconverteranditscomponentsbecomesacriticalissue. Adesignfocusedonqualityperformanceandreliabilityisessentialinordertosatisfy theexpectedcustomerrequirements.Inrecentyearstherearevariousmodelsandcounting algorithmstoestimatethelifetimeofanIGBTpowerconverter.Theyinthenumber ofparametersusedtospecifyatemperaturecycle.Basically,thelifetimeestimationof powerconverterestablishesthelinkagebetweenanapplication'stypicalloadingwith 101 IGBTs'splifetimemodel.Severalcyclecountingmethodshavebeendeveloped,which include.Forexample,thelevelcrossingcountingmethod,thepeakcountingmethod,and thesimplerange/meancountingmethod.However,thesemethodscannotcaptureallthe characteristicsneededforaccuratefatigueanalysis. 5.4.1wCycleCounting Countingmethodshaveinitiallybeendevelopedforthestudyoffatiquedamagegenerated inaeronauticalstructures.sincetresultshavebeenobtainedfromtmethods, errorscouldbetakenincalculationsforsomeofthem.Levelcrossingcounting,peakcount- ing,simplerangecountingandwcountingarethemethodswhichareusingstressor deformationranges.Oneofthepreferredmethodsistherainwcountingmethod.The signalmeasured,ingeneral,arandomstress S ( t )isnotonlymadeupofapeakalonebetween twopassagesbyzero,butalsoseveralpeaksappear,whichmakesthedetermination ofthenumberofcyclesabsorbedbythestructure. Thewcyclecountingmethod,whichwasdevelopedin1968byEndoandMat- suishi,isoneofthemostpopularcyclecountingtechniquesusedinfatigueanalysis[152]. Thewcountingmethodwasoriginallytermsasiscalled\PagodaRoofMethod".It canbeexplainedasarandomstressS(t)representingaseriesofroofsontowhichwaterfall withtimebeingtheverticalaxis.Thisalgorithmwasoriginallydevelopedformechanical fatigueanalysis.Herein,itisusedtoextractthenumberandamplitudeofthermalcycles capturedfromrealtimesimulationofthetemperatureRepetitivethermalstresses causedbypowercyclinginducefatigueinWECSandreducetheexpectedlifetimeofthe converters.Althoughtherearevariousapproachesforlifetimemodellingofpowersemicon- ductordevices,thelifetimemodelsprovidedbydevicemanufacturersarefrequentlyused 102 [153,154,155,156]. Thisanalogyisconcludedfromthecomparisonofrainfallingonthepagodaandrunning downtheedgesoftheroof.itcanbesummarizedasfollows[157]: 1.Rotatetheloadinghistory90 o ,thatisverticaltimeaxisdownwardandtheloadtime historyresemblesapagodaroof. 2.imagineawofrainstartingateachsuccessiveextremumpoint. 3.ealoadingreversal(halfcycle)byallowingeachwtocontinuetodripdown theseroofsuntil: Itfallsoppositealargemaximumorsmallerminimumpoint. Itmeetsapreviouswfallingfromabove. Itfallsbelowtheroof. 4.Identifyeachhystersisloop(cycle)bypairingupthesamecountedreversal. Thesignalmeasured,ingeneral,arandomstress S ( t )isnotonlymadeupofapeak alonebetweentwopassagesbyzero,butalsoseveralpeaksappear.whichmaked thedeterminationofthenumberofcyclesabsorbedbythestructure,anexampleofrandom stressdataisshownin5.6.Thecountingofpeaksmakesitpossibletoconstitutea histogramofthepeaksofrandomstresswhichcanbetransformedintoastressspectrum givingthenumberofeventsforlowerthanagivenstressvalue.Thestressspectrumisthus arepresentationofthestatisticaldistributionofthecharacteristicamplitudesoftherandom stressasafunctionoftime. 103 Figure5.6StressStraincycles. 5.4.2LifetimeModeling TheIGBTlifetimepredictionmodelscanbecategorizedintoanalyticalandphysicalmodels. Physicalmodellingrequiresfailureanddeformationmechanismstobepriorlyknown.Itis basedontheknowledgeofstress/straindeformationswithindevicesthatcanbegained eitherbyexperimentsorsimulations[158].Analyticalmodelsdescribethedependenceof thenumberofcyclestofailure N f ontheparametersoftemperaturecyclessuchasamplitude, duration,frequency,meanvalueandmaximumandminimumtemperatures. modelisconsideredoneoftheanalyticalmodelshavebeenpublishedinliterature.The modelisusedinthispaper,takesintoconsiderationonlythetemperatureswing 4 T which hasbeenextractedfromthewcountingalgorithm[159].Thenumber N ofcycles untilacertainpercentageofthemodulefailcanbecalculatedasshownin(5.28)[160]: N = k 1 : T k 2 (5.28) 104 Thetwoparameters k 1 , k 2 respectivelyarescaleparameterandtheexponentparameter ortheshapeparameterthatcontrolshowstrongthetemperaturedependenceis.Bothof themaredevicedependent,andhavetobedeterminedbasedonmeasurements.Where k 1 =8 : 2 10 14 and k 2 =5 : 28[160]. Furthermore,itisnotpossibletocalculatetheexactlifetimeofindividualmodules. Insteadthelifetimemustbeexpressedintermsofthe B 10 lifetime,thatisthenumberof cyclesduringwhich10%ofthetotalnumberofmodulesfails.Theanalyticallifeisperformed bymeansofusingMiner'sRulefordamageaccomulation[161],[162].Thecorrelationbetween temperaturechangesandthedamagedproducedwithintheIGBThastobeand thenlifetimeispresentedasinverseofthetotaldamageaccumulatedwithinapowermodule untilthesuspensionofitsnormalworking.Therefore,missiontransformationinto asequenceofnon-uniformtemperaturecyclesisthemainissueoftheanalyticalapproach. Themainassumptionisthateverytemperaturecycleconsumesacertainfractionofthe IGBT'slifetime.Thetotaldamagecanbeasthesumofallthefractionaldamages overatotalof k blocksasshownin5.29. " n at T 1 N 10% ; T T 1 # + " n at T 2 N 10% ; T T 2 # + ::: + " n at T k N 10% ; T T k # < 100% : (5.29) ThelifetimeoftheIGBTispredictedtobe4818 : 5hoursifrunningattheseloadcondi- tionsduring30sinterval. ThelimitationsofthePalmgrenMinerrulecanbesummarizedasthefollowing: Linearitassumesthatallcyclesofagivenmagnitudedothesameamountofdamage, weathertheyoccurearlyorlateinthelife. Noninteractive(sequenceitassumesthatthepresenceof S 2 etc.doesnot 105 thedamagecausedby S 1 . StressIndependentitassumesthattherulegoverningthedamagecausedby S 1 isthe sameasthatgoverningthedamagedcausedby S 2 .Theassumptionsareknownto befaulty,however,Palmgren-Minerruleisstillusedwidelyintheapplicationofthe fatiquelifeestimates. 5.5PrinciplesofFilteringWindTurbinePowerCom- mandFluctuations InWECSbasedonvariablespeeddrives,themodernpowerelectronicconverternormally operateundermaximumpowerpointtracking(MPPT)algorithmsandcapturethemaxi- mallyavailablewindpowerthatissubjecttoatmultipletimescales.Oneof thepivotalparametersforlifetimeestimationofIGBTisthethermalenvironmentandthe numberofthermalcyclesthatthedevicesundergo.AlthoughWECSpowerconvertershave fewmovingpartsinvolvedthatmaymechanicallywearout,theyarevulnerabletoexcessive voltageandcurrentsandtheirvariations.Damagescouldresultfromevenbyvery-short durationshocksabovemaximumratings.Inawelldesignedsystem,thepowerconverter devicesarewellprotectedfromsuchevents. StudieshaveshownthatthepowercyclingoftheIGBTisoneofthedominantfailure mechanisminpowerconverters[49].ALPFhasbeenprovedaneapproachto minimizingnumberofthermalcycles.Thecontrolstrategyismostlyfocusedonwindturbine sideactivepowercontrolbutnotthepowergridside.However,thewindturbine'soutput powerduetowindspeedvariations.Therefore,aLPFisusedtosmoothingthese 106 Thispaperproposesanewrealpowercontrolmethod,inwhichthesmoothingperfor- manceisexamined.ThesimulationresultsshowthataWECSwiththeproposedcontrol methodhassuperiorperformanceforoutputpowerwindturbineswithreducedthermalcycle count.Alowpassisusedtoremovetidalandhigherfrequencyionsfromthe timeseriesdata.Itsometimessuppressessmallerinthetimeseriesplotsthat aredrivenbywindanddensityThisparametersaredeterminedbythenewterm \missingenergy"or\deadenergy"andthenumberofthermalcycles.Simulationresultsshow thattheproposedmethodcanensuretheofsmoothingwindpowerThis controlstrategycanmodifythenumberofthermalcycles.Besides,itextendtheservicelife ofthesemiconductordevices.TheLPFdesignisstronglyapplicationdependent. Figure5.7ofthewindspeedusedinthissimulation. Thispaperpresentsacomparativestudybetweenconventionalcontrollerandproposed 107 controllers.IthasbeenshownthatthenumberofthermalcyclesinIGBTpowerconvert- erscanbetlyreducedbyapplyingLPFtovariablespeedDFIG.Thenumberof thermal-cyclesoftheconverterreducesbyapproximately7timeswiththeproposedLPF controlmethod.Inaddition,thelifetimeincreasesbyapproximatearound22 : 25%.Thether- malstressesoftheIGBTjunctiontemperaturearemuchreducedwhenthecontrolmethod isenabled.Furthermore,thewindspeedsignalsusedisarealworldsignals,whichmake simulationclosetorealityasshowninFig.5.7.Dataonthewindareassumedas aprerequisite.ThesewinddatacanbeprocessedwithamodeloftheWECStoderivethe electricaloperatingconditionsoftheconverter.Thusthepowerdissipationforeachsemi- conductordevicecanbedeterminedandusedtocalculatetherelatedjunctiontemperatures. IndesigningtheLPF,moreattentionshouldbepaidtothekineticenergystoredinthe rotatingmassoftheWECS.whichisgivenby: E = 1 2 J! 2 t (5.30) Theinertiaisthecongregationalinertiaofthewindturbinebladesthatcaptureenergy, andarotorhub,thatconnectsthebladestotheshaft,alongwithpitchmechanismthat assistsintcaptureofenergy.Whenthepowercommandbecomeshigher thantheavailablepower,problemarises.TheLPFtimeconstantisdeterminedbasedon theamountofenergyavailabletobedrawnfromtherotatingassembly.Tooptimizethe valueoftheLPFtimeconstantadetectableamountofmissingenergyordeadenergyhas beenconsidered.Thismissingenergyhasbeenestimatedbythebetweentwo areasunderthetwocurvesinFig.5.8.Thisenergyhasbeendecreasedtherotorspeed ofrotatinggroup.Thatmeans,thevalueoftimeconstantisexamined.Wherehightime 108 Figure5.8PlotsofthepowercommandswithandwithouttheLPF. constantleadstonarrowbandpassSubsequentlymissingenergywillbeincreased. Ontheotherhand,thelowtimeconstantmeansthebandpassfrequencywillbehigh, andthemissingenergywillbedecreased.Inotherwords,comebacktotheactualpower commandwhichmeansdonothingtosuppressthethermalcycles.Theofde- loadingthewindturbinetoovercometheshortageofkineticenergyareevaluated.Large contributionstoinertialresponsearepossible,butvarywithoperatingpoint.Contributions arelimitedtoaboveacertainwindspeedduetorotorunder-speed.De-loadingcanbeused tomaximizethecontributionovertheviablerange.Windvariationsreducethemagnitude ofthecontribution,andmakerotorspeedinstabilitypossibleforsomecontrolreferences. Reducingrotorspeedisconsideredasthemostrouteforloweringnoiseemissioninwind turbines.Inaddition,variablespeedoperationisalsoe,enablingdesignerstoprogram operationforlowerspeedsatnight,whennoisesensitivityisgreatest. 109 Table5.4:Theenessofthenewstrategy. WithoutLPF WithLPF Thetotalkineticenergycaptured(J) 139907 114538 Therevolvinggrouplowsidespeed(rpm) 193 175 Therevolvinggrouphighsidespeed(rpm) 976 884 Lifetimecalculated(years) 0.55 12.79 Numberofthermalcycles 1679.5 229 Figure5.9PlotofthedcbusvoltagethatstaysthesamewithorwithoutLPF. Asamatteroffact,semiconductordevicesaresubjectedtoavarietyoftemperature changesduetomanyfactors;someofthesearecausedbythedevicesthemselves,e.g.by switchingorconductionlosses,converterpowervariation,andsomeofthetemperature changesarecausedbyexternalfactors,e.g.changeofseasonsorreducedcoolingThemag- nitudeofatemperaturechangecanrangefromfractionsofdegreestomorethan100Kas shownin5.12. 110 Figure5.10RotorwindingterminalvoltagewithoutLPF. Figure5.11LPFresponsecharacteristicusingIIRimpulseresponse,minimumordermode, singleratetype,Butterworthalgorithm. 111 Figure5.12IGBTjunctiontemperaturevariationswithoutLPF. Figure5.13IGBTjunctiontemperaturevariationswithLPF. 112 Figure5.14Temperaturemeanvalue T m extractedfromwcountingalgorithmwithout LPF. Figure5.15Amplitude, T extractedfromwcountingalgorithmwithoutLPF. 113 Figure5.16Frequencydistributionoftemperaturecyclesbytheiramplitude T and temperaturemeanvalue T m extractedfromwcountingalgorithmwithoutLPF. 114 Chapter6 ConclusionsandFutureworks 6.1Conclusions Thepredictionofpowercyclinglifetimeforapowerelectronicconverterinrotorsidecontrol inDFIGisexamined.AcomprehensivethermalmodelforthepowerIGBTmodulesused inthree-phaseconverterhasbeenbuilttopredictthedynamicjunctiontemperaturerise underrealoperatingconditions.Thepowerlossmodel,whichisbasedonthelook-uptable methodforcalculatingtheconductionandswitchinglosseshasbeensuccessfullyv withsimulation.AnequivalentRCnetworkmodelisbuilttoperformthethermalanalysis. Lifetimeisestimated. Theanalysisshowsthatthelifetimeisheavilybythermalcycling,andthe behaviorofthesemiconductordevicesandtheirmissionwhichdirectlythe lifetime.Hence,anactivereal-timecontrolmethodisusedtominimizedthepower ationsexperiencedbytheDFIGsystemandsubsequentlytoreducethenumberofthermal cycles.ByusingtheproposedLPFmethod,thestressonthepowerconverterduetother- malcyclinghasbeentlyreducedandtheestimatedlifetimeofthesystemhas substantiallyincreased. Contributionsofthethesisare: Tothebestofmyknowledgethisistheworkonrealtimecontrolschemebased 115 onwindreliabilitymodel. Theoverallmethodologyappliedtowindenergyapplicationsisorigional. ALPFhasbeenprovedaneapproachtominimizenumberofthermalcycles andmaximizelifetime. Tothebestofmyknowledgethisistheworkusingaveragemodeltoreduce complexityandleadstofastertimesimulationtoprovereliabilityinWECS. Tothebestofmyknowledge,thisisthestworkreviewreliabilitymethodsforWECS andWFtogether. Tothebestofmyknowledge,thisistheworkclassifyingreliabilityonwindenergy intoWECSandWF. Tobestofmyknowledge,thisistheworkdiscussingreliabilityoverWECSand WF. 6.2FutureWork Continuetostayclosetoindustry. Includeendusersinfuturework. MoreattentiononGSCreliability. Moreworkonmissingenergyandtheenessofmaximumwindenergy. Windenergygenerationsystemsreliabilityisanopenproblematicissue.Whereis thedeterminationofthemostthermo-electricallystresseddevicesofapowercon- 116 verterisveryimportant.ThevoltageandcurrentofIGBTmodulesmustremain withinthelimitsgivenbytheirmanufactureerunderoperatingconditions.Andthe workingtemperatureanditsvariations(temperatureswing)doesnotexceedcertain maximumvalues.Severalshouldhavebeenaddressedtoimprovethepower devicesruggednessunderoverloadingconditions. WithanincreasingIGBToperatingtemperature,wehavetochoiceseitherraising oftheoutputcurrent,ordecreasingofcoolingcosts.theimpactoftheinteraction betweenthepowermoduleanditscoolingsystemontheinverterreliabilityneedmore attention Conductingmoreresearchonagoodthermalmanagementandtakeintoaccountthe surroundingarea. Enablesmallerandmorecosttinverterdesigns.designingalowlosses,reliable operationathightemperaturepowerelectronicmodules. Workingonpowerelectronicmodulesthatenablereliableoperationathightempera- tureandhighcurrents. StudyingtheswitchingspeedofanIGBTmoduledesignbyfocusingonelectromag- neticsimulations.Findnewsimulationsmethodsthatenablestuningtomeetthe requirementsoflowlossandgoodshortcircuitbehavior. 117 BIBLIOGRAPHY 118 BIBLIOGRAPHY [1]V.G.LucianMihet-Popa [2]H.Bahirat,B.Mork,andH.Hoidalen,\Comparisonofwindfarmtopologiesfor shoreapplications,"pp.1{8,July2012. [3]C.Busca,R.Teodorescu,F.Blaabjerg,S.Munk-Nielsen,L.Helle,T.Abeyasek- era,andP.Rodriguez,\Anoverviewofthereliabilitypredictionrelatedaspectsof highpower f IGBTs g inwindpowerapplications," MicroelectronicsReliability ,vol.51, no.911,pp.1903{1907,2011.Proceedingsofthe22thEuropeanSymposiumonthe f RELIABILITY gf OF gf ELECTRON g DEVICES, f FAILURE gf PHYSICS gf AND g f ANALYSIS g . [4]R.Pena,J.Clare,andG.Asher,\Doublyfedinductiongeneratorusingback-to-back pwmconvertersanditsapplicationtovariable-speedwind-energygeneration," Electric PowerApplications,IEEProceedings- ,vol.143,pp.231{241,May1996. [5]L.GopiReddy,L.Tolbert,B.Ozpineci,andJ.Pinto,walgorithmbasedlife- timeestimationofpowersemiconductorsinutilityapplications," IndustryApplica- tions,IEEETransactionson ,vol.PP,no.99,pp.1{1,2015. [6]L.Disclaimer,\RenewableEnergyPolicyNetworkforthe21Centurykerneldescrip- tion,"2013. [7]P.Komor,\Windandsolarelectricity:callengesandopportunities," PowerCentralon GlobalClimateChange,UniversityofColoradoatBoulder,Colorado ,2009. [8]E.Zio,\Reviewreliabilityengineeringoldproblemsandnewchallenges," Reliability EngineeringandSystemSaftey ,vol.94,pp.125{141,Aug2009. [9]P.M.Z.G.J.W.vanBussel,\Reliability,availabilityandmaintenanceaspectsoflarge- scalewindfarms,aconceptsstudy.," SectionWindEnergy,FacultyCivil EngineeringandGeosciences,DelftUniversityofTechnology,TheNetherlands ,2011. [10]k.Bourouni,\Availabilityassaessmentofareverseosmosisplantcomarisonbetween reliabilityblockdiagramsandfaulttreeanalysismethods," Desalination ,2013. 119 [11]M.Anthony,R.Arno,N.Dowling,andR.Schuerger,\Reliabilityanalysisforpower topumpusingfaulttreeandrbd," IndustryApplications,IEEETransactionson , vol.49,pp.997{1003,March2013. [12]E.Gavanidou,A.Bakirtzis,andP.Dokopoulos,\Aprobabilisticmethodfortheeval- uationoftheperformanceofwind-dieselenergysystems," EnergyConversion,IEEE Transactionson ,vol.7,pp.418{425,Sep1992. [13]A.Bakirtzis,\Aprobabilisticmethodfortheevaluationofthereliabilityofstandalone windenergysystems," EnergyConversion,IEEETransactionson ,vol.7,pp.99{107, Mar1992. [14]M.Aghaebrahimi,M.Mehdizadeh,andH.Na\Anewalgorithmforreliability assessmentofwind-dieselsysteminislandingmodeofoperation,"in PowerEngineer- ing,EnergyandElectricalDrives(POWERENG),2011InternationalConferenceon , pp.1{6,May2011. [15]M.Vintr,\Reliabilityassessmentforcomponentsofcomplexmechanismsandma- chines,"in ProceedingsofTwelfthWorldCongressinMechanismandMachineScience- IFToMM2007 ,2007. [16]D.Nicholls,\Newreliabilitypredictionmethodologyincorporatesandtestexpe- rience," JournaloftheIEST ,vol.50,no.2,pp.5{13,2007. [17]N.B.Fuqua,\Electronicreliabilityprediction,selectedtopicsinassurancerelated technologies," START ,vol.4,no.2,pp.1{6,1997. [18]M.Pecht,\Thereliabilityphysicsapproachtofailurepredictionmodeling," Quality andReliabilityEngineeringInternationalConference ,1990. [19]M.PechtandJ.Gu,\Physics-of-failure-basedprognosticsforelectronicproducts," TransactionsoftheInstituteofMeasurementandControl ,vol.31,no.3-4,pp.309{ 322,2009. [20]M.-M.Handbook,\217fnotice2,"reliabilitypredictionofelectronicequipment", 1995,"tech.rep.,MIL-HDBK-217. [21]W.Denson,\Handbookof217plusreliabilitypredictionmodels-may2006,"tech.rep., ReliabilityInformationAnalysisCenter. [22]PRISM,\Systemreliabilityassessmentsoftwaretool,"tech.rep.,ReliabilityInforma- tionAnalysisCenter1999. 120 [23]N.B.Negra,O.Holmstr˝m,B.Bak-Jensen,andP.S˝rensen,\Windfarmgeneration assessmentforreliabilityanalysisofpowersystems," WindEngineering ,vol.31,no.6, pp.383{400,2007. [24]T.MancoandA.Testa,\Amarkovianapproachtomodelpoweravailabilityofawind turbine,"in PowerTech,2007IEEELausanne ,pp.1256{1261,July2007. [25]A.Ranjbar,P.Shamsi,andB.Fahimi,\Anovelvoter-basedmarkovmodelforreli- abilityassessmentofmulti-portpowerelectronicinterface(mpei),"in VehiclePower andPropulsionConference(VPPC),2011IEEE ,pp.1{6,Sept2011. [26]B.SinghandS.Singh,\Windpowerinterconnectionintothepowersystem:Areview ofgridcoderequirements," TheElectricityJournal ,vol.22,no.5,pp.54{63,2009. [27]A.a,M.Chindris,A.Sumper,G.Gross,andF.Ferrer,\Windturbineoperation inpowersystemsandgridconnectionrequirements," ProceedingsofICREPQ ,vol.5, pp.1{5,2005. [28]J.GarciaandM.Babazadeh,\Controloflargescalewindpowerplants,"in Power andEnergySocietyGeneralMeeting,2012IEEE ,pp.1{8,July2012. [29]I.Erlich,W.Winter,andA.Dittrich,\Advancedgridrequirementsfortheintegration ofwindturbinesintothegermantransmissionsystem,"in PowerEngineeringSociety GeneralMeeting,2006.IEEE ,pp.7pp.{,2006. [30]P.J.Tavner,J.Xiang,andF.Spinato,\Reliabilityanalysisforwindturbines," Wind Energy ,vol.10,no.1,pp.1{18,2007. [31]P.Tavner,J.Xiang,andF.Spinato,\Improvingthereliabilityofwindturbinegener- ationanditsimpactonoveralldistributionnetworkreliability,"pp.1{4,June2005. [32]F.Spinato, Thereliabilityofwindturbines .PhDthesis,UniversityofDurham,2008. [33]R.LuczkowskiandR.Muszynski,\Costandreliabilityorienteddesignofthecon- verterforsmallwindpowerplant,"in ReliabilityinElectromagneticSystems,2007 IETColloquiumon ,pp.1{5,May2007. [34]B.BagenandR.Billinton,\Reliabilitycost/worthassociatedwithwindenergyand energystorageutilizationinelectricpowersystems,"in ProbabilisticMethodsApplied toPowerSystems,2008.PMAPS'08.Proceedingsofthe10thInternationalConference on ,pp.1{7,IEEE,2008. 121 [35]M.HaghifamandM.Omidvar,\Windfarmmodelinginreliabilityassessmentofpower system,"in ProbabilisticMethodsAppliedtoPowerSystems,2006.PMAPS2006. InternationalConferenceon ,pp.1{5,IEEE,2006. [36]X.Yu,Z.Yong-qiang,andD.Ze-jun,\Analysisonreliabilityindicesofwindfarmand itsfactors,"in ElectricityDistribution(CICED),2010ChinaInternational Conferenceon ,pp.1{5,IEEE,2010. [37]O.HolmstromandN.B.Negra,\Surveyofreliabilityoflargewindfarms," in EuropeanWindEnergyConference ,pp.1{10,2007. [38]V.Hamidi,\Reliabilityassessmentofnetworksforconnectionofwind- farms,"in ReliabilityofTransmissionandDistributionNetworks(RTDN2011),IET Conferenceon ,pp.1{6,Nov2011. [39]A.Leite,C.Borges,andD.Falcao,\Probabilisticwindfarmsgenerationmodelfor reliabilitystudiesappliedtobraziliansites," PowerSystems,IEEETransactionson , vol.21,pp.1493{1501,Nov2006. [40]M.-R.HaghifamandS.Soltani,\Reliabilitymodelsforwindfarmsingenerationsystem planning,"in ProbabilisticMethodsAppliedtoPowerSystems(PMAPS),2010IEEE 11thInternationalConferenceon ,pp.436{441,June2010. [41]E.Chiodo,D.Lauria,C.Pisani,andD.Villacci,\Reliabilityaspectsinwindfarmsde- sign,"in PowerElectronics,ElectricalDrives,AutomationandMotion(SPEEDAM), 2012InternationalSymposiumon ,pp.577{581,IEEE,2012. [42]M.AmirabadiandH.A.Toliyat,\Ahighlyreliableconverterforwindpowergeneration application,"in AppliedPowerElectronicsConferenceandExposition(APEC),2013 Twenty-EighthAnnualIEEE ,pp.1117{1123,IEEE,2013. [43]J.A.Baroudi,V.Dinavahi,andA.M.Knight,\Areviewofpowerconvertertopologies forwindgenerators," RenewableEnergy ,vol.32,no.14,pp.2369{2385,2007. [44]A.GarcesandM.Molinas,\Astudyofinareducedmatrixconverterfor windfarms," IndustrialElectronics,IEEETransactionson ,vol.59,no.1, pp.184{193,2012. [45]Z.Q.ZhuandJ.Hu,\Electricalmachinesandpower-electronicsystemsforhigh-power windenergygenerationapplications," Compel ,vol.32,no.1,pp.7{33,2013. 122 [46]M.Held,P.Jacob,G.Nicoletti,P.Scacco,andM.-H.Poech,\Fastpowercycling testofigbtmodulesintractionapplication,"in PowerElectronicsandDriveSystems, 1997.Proceedings.,1997InternationalConferenceon ,vol.1,pp.425{430vol.1,May 1997. [47]V.Smet,F.Forest,J.Huselstein,A.Rashed,andF.Richardeau,\Evaluationof v ce monitoringasareal-timemethodtoestimateagingofbondwire-igbtmodulesstressed bypowercycling," IndustrialElectronics,IEEETransactionson ,vol.60,pp.2760{ 2770,July2013. [48]A.LindemannandG.Strauch,\Propertiesofdirectaluminumbondedsubstratesfor powersemiconductorcomponents," PowerElectronics,IEEETransactionson ,vol.22, pp.384{391,March2007. [49]M.Ciappa,F.Carbognani,andW.Fichtner,\Lifetimepredictionanddesignofreli- abilitytestsforhigh-powerdevicesinautomotiveapplications," DeviceandMaterials Reliability,IEEETransactionson ,vol.3,pp.191{196,Dec2003. [50]L.Wei,R.Kerkman,R.Lukaszewski,H.Lu,andZ.Yuan,\Analysisofigbtpowercy- clingcapabilitiesusedindoublyfedinductiongeneratorwindpowersystem," Industry Applications,IEEETransactionson ,vol.47,pp.1794{1801,July2011. [51]S.Azuma,M.Kimata,M.Seto,X.Jiang,H.Lu,D.Xu,andL.Huang,\Research onthepowerlossandjunctiontemperatureofpowersemiconductordevicesforin- verter," VehicleElectronicsConference,1999.(IVEC'99)ProceedingsoftheIEEE International ,pp.183{187vol.1,1999. [52]T.Reimann,R.Krummer,U.Franke,S.Petzoldt,andL.Lorenz,\Realtimecal- culationofthechiptemperatureofpowermodulesinpwminvertersusinga16bit microcontroller," PowerSemiconductorDevicesandICs,2000.Proceedings.The12th InternationalSymposiumon ,pp.127{130,2000. [53]M.WeckertandJ.Roth-Stielow,\Lifetimeasacontrolvariableinpowerelectronic systems,"in Emobility-ElectricalPowerTrain,2010 ,pp.1{6,Nov2010. [54]F.Blaabjerg,K.Ma,andD.Zhou,\Powerelectronicsandreliabilityinrenewable energysystems,"in IndustrialElectronics(ISIE),2012IEEEInternationalSymposium on ,pp.19{30,May2012. [55]F.Blaabjerg,M.Liserre,andK.Ma,\Powerelectronicsconvertersforwindturbine systems," IndustryApplications,IEEETransactionson ,vol.48,pp.708{719,March 2012. 123 [56]B.Mecrow,A.Jack,J.Haylock,andJ.Coles,\Fault-tolerantpermanentmagnet machinedrives," IEEProceedings-ElectricPowerApplications ,vol.143,no.6,pp.437{ 442,1996. [57]L.Yazhou,\Studiesonwindfarmintegrationintopowersystem[j]," Automationof ElectricPowerSystems ,vol.8,p.017,2003. [58]N.Ertugrul,W.Soong,G.Dostal,andD.Saxon,\Faulttolerantmotordrivesystem withredundancyforcriticalapplications,"in PowerElectronicsSpecialistsConference, 2002.pesc02.2002IEEE33rdAnnual ,vol.3,pp.1457{1462,IEEE,2002. [59]R.J.McGlen,R.Jachuck,andS.Lin,\Integratedthermalmanagementtechniquesfor highpowerelectronicdevices," AppliedThermalEngineering ,vol.24,no.8,pp.1143{ 1156,2004. [60]L.E.Bell,\Cooling,heating,generatingpower,andrecoveringwasteheatwithther- moelectricsystems," Science ,vol.321,no.5895,pp.1457{1461,2008. [61]U.K.Mishra,P.Parikh,Y.-F.Wu, etal. ,\Algan/ganhemts-anoverviewofdevice operationandapplications," PROCEEDINGS-IEEE ,vol.90,no.6,pp.1022{1031, 2002. [62]S.Yang,D.Xiang,A.Bryant,P.Mawby,L.Ran,andP.Tavner,\Conditionmonitoring fordevicereliabilityinpowerelectronicconverters:areview," PowerElectronics,IEEE Transactionson ,vol.25,no.11,pp.2734{2752,2010. [63]E.James,M.Tudor,S.Beeby,N.Harris,P.Glynne-Jones,J.Ross,andN.White,\An investigationofself-poweredsystemsforconditionmonitoringapplications," Sensors andActuatorsA:Physical ,vol.110,no.1,pp.171{176,2004. [64]B.Lu,Y.Li,X.Wu,andZ.Yang,\Areviewofrecentadvancesinwindturbine conditionmonitoringandfaultdiagnosis,"in PowerElectronicsandMachinesinWind Applications,2009.PEMWA2009.IEEE ,pp.1{7,IEEE,2009. [65]D.Zhou,F.Blaabjerg,M.Lau,andM.Tonnes,\Thermalbehavioroptimizationin multi-mwwindpowerconverterbyreactivepowercirculation," IndustryApplications, IEEETransactionson ,vol.50,pp.433{440,Jan2014. [66]N.Baars,C.Wijnands,andJ.Duarte,\Reductionofthermalcyclingtoincreasethe lifetimeofmosfetmotordrives,"in IndustrialElectronicsSociety,IECON2014-40th AnnualConferenceoftheIEEE ,pp.1740{1746,Oct2014. 124 [67]S.Bifaretti,P.Zanchetta,A.Watson,L.Tarisciotti,andJ.C.Clare,\Advancedpower electronicconversionandcontrolsystemforuniversalandpowermanagement," SmartGrid,IEEETransactionson ,vol.2,no.2,pp.231{243,2011. [68]J.Ousten,Z.Khatir,andL.Menager,\Studyofthermalinterfacesagingforpower electronicsapplications,"in SemiconductorThermalMeasurementandManagement Symposium(SEMI-THERM),201127thAnnualIEEE ,pp.10{17,March2011. [69]B.Ray,\Temperatureduringhumidageingoninterfacesofglassandcarbon ersreinforcedepoxycomposites," JournalofColloidandInterfaceScience ,vol.298, no.1,pp.111{117,2006. [70]Y.Wu,M.A.Knight,andR.McMahon,\Comparisonoftheofcontin- uousanddiscontinuouspwmschemesonpowerlossesofvoltage-sourcedinvertersfor inductionmotordrives," PowerElectronics,IEEETransactionson ,vol.26,pp.182{ 191,Jan2011. [71]J.Kolar,H.Ertl,andF.C.Zach,\Inofthemodulationmethodonthecon- ductionandswitchinglossesofapwmconvertersystem," IndustryApplications,IEEE Transactionson ,vol.27,pp.1063{1075,Nov1991. [72]A.Hensler,J.Lutz,M.Thoben,andK.Guth,\Firstpowercyclingresultsofimproved packagingtechnologiesforhybridelectricalvehicleapplications,"in IntegratedPower ElectronicsSystems(CIPS),20106thInternationalConferenceon ,pp.1{5,March 2010. [73]A.GoelandR.Graves,\Electronicsystemreliability:collatingpredictionmodels," DeviceandMaterialsReliability,IEEETransactionson ,vol.6,pp.258{265,June 2006. [74]H.Arabian,Y.Feng,andP.Tavner,\Deliverabled1.1{reportonpreviouspubli- cationsdealingwithwindturbinereliability,"tech.rep.,TechnicalReport(Project Deliverable),EUFP7ProjectRELIAWIND212966. [75]B.Foucher,J.Boulli,B.Meslet,andD.Das,\Areviewofreliabilitypredictionmethods forelectronicdevices," MicroelectronicsReliability ,vol.42,no.8,pp.1155{1162,2002. [76]M.Economou,\Themeritsandlimitationsofreliabilitypredictions," Reliabilityand Maintainability,2004AnnualSymposium-RAMS ,pp.352{357,Jan2004. 125 [77]S.DistefanoandA.\Dynamicreliabilityblockdiagramsvsdynamicfault trees," ReliabilityandMaintainabilitySymposium,2007.RAMS'07.Annual ,pp.71{ 76,Jan2007. [78]S.Barbati,\Commonreliabilityanalysismethodsandprocedures,"Tech.Rep.1, ReliawindWPD,2008. [79]M.BoettcherandF.Fuchs,\Powerelectronicconvertersinwindenergysystems;con- siderationsofreliabilityandstrategiesforincreasingavailability,"in PowerElectronics andApplications(EPE2011),Proceedingsofthe2011-14thEuropeanConferenceon , pp.1{10,Aug2011. [80]M.Das,S.Panja,S.Chowdhury,S.Chowdhury,andA.Elombo,\Expert-basedfmeaof windturbinesystem," IndustrialEngineeringandEngineeringManagement(IEEM), 2011IEEEInternationalConferenceon ,pp.1582{1585,Dec2011. [81]L.Wu,J.Zhang,M.Zhang,R.Liu,X.Zhu,andW.Liu,\Reliabilityassessment ofcompositegenerationandtransmissionsystemcontainingwindfarms," Innovative SmartGridTechnologies-Asia(ISGTAsia),2012IEEE ,pp.1{5,May2012. [82]W.WangdeeandR.Billinton,\Reliabilityassessmentofbulkelectricsystemscontain- inglargewindfarms," InternationalJournalofElectricalPowerandEnergySystems , vol.29,no.10,pp.759{766,2007. [83]C.Zapata,L.Garces,andO.Gomez,\Reliabilityassessmentofenergylimitedsystems usingsequentialmontecarlosimulation," TransmissionDistributionConferenceand Exposition:LatinAmerica,2006.TDC'06.IEEE/PES ,pp.1{6,Aug2006. [84]F.Spinato,P.Tavner,G.vanBussel,andE.Koutoulakos,\Reliabilityofwindturbine subassemblies," RenewablePowerGeneration,IET ,vol.3,pp.387{401,December 2009. [85]P.J.TavnerandF.Spinato,\Reliabilityoftwindturbineconceptswithrele- vancetoapplication,"(Brussels),inEuropeanWindEnergyConference,2008. [86]S.Barbati,\Commonreliabilityanalysismethodsandprocedures,"tech.rep. [87]W.-J.GuandR.Liu,\Astudyofvolumeandweightvs.frequencyforhigh-frequency transformers,"in PowerElectronicsSpecialistsConference,1993.PESC'93Record., 24thAnnualIEEE ,pp.1123{1129,Jun1993. 126 [88]K.AlewineandW.Chen,\Areviewofelectricalwindingfailuresinwindturbine generators,"in ElectricalInsulationConference(EIC),2011 ,pp.392{397,June2011. [89]F.Blaabjerg,A.Isidori,andF.Rossi,\Impactofmodulationstrategiesonpower devicesloadingfor10mwmultilevelwindpowerconverter,"in PowerElectronicsfor DistributedGenerationSystems(PEDG),20123rdIEEEInternationalSymposiumon , pp.751{758,June2012. [90]S.Bala,J.Pan,D.Das,O.Apeldoorn,andS.Ebner,\Loweringfailureratesandim- provingserviceabilityinwindconversion-collectionsystems,"in PowerElec- tronicsandMachinesinWindApplications(PEMWA),2012IEEE ,pp.1{7,July 2012. [91]C.W.Zhang,T.Zhang,N.Chen,andT.Jin,\Reliabilitymodelingandanalysisfor anoveldesignofmodularconvertersystemofwindturbines," ReliabilityEngineering andSystemSafety ,vol.111,pp.86{94,2013. [92]S.Yang,D.Xiang,A.Bryant,P.Mawby,L.Ran,andP.Tavner,\Conditionmonitoring fordevicereliabilityinpowerelectronicconverters:Areview," PowerElectronics, IEEETransactionson ,vol.25,pp.2734{2752,Nov2010. [93]W.Liu,B.Tang,andY.Jiang,\Statusandproblemsofwindturbinestructuralhealth monitoringtechniquesinchina," RenewableEnergy ,vol.35,no.7,pp.1414{1418, 2010.SpecialSection: f IST g NationalConference2009. [94]R.Kumar,M.Manimozhi,andM.Enosh,\Asurveyoffaultdetectionandisolation inwindturbinedrives,"in Power,EnergyandControl(ICPEC),2013International Conferenceon ,pp.648{652,Feb2013. [95]P.Zhu,Y.Liu,R.Robert,andX.Hao,windconverterreliabilityevaluation," in PowerElectronicsandECCEAsia(ICPEECCE),2011IEEE8thInternational Conferenceon ,pp.966{971,May2011. [96]F.S.H.B.Tavner,P.andG.v.Bussel,\Reliabilityandavailabilityofwindturbine electricalandelectroniccomponents,"in EPEJournal,20(4) ,pp.45{50,May2010. [97]P.Tavner,\Deliverabled6.7-reportrecommendationsfromthereliawindconsortium forthestandardisationforthewindindustryofwindturbinereliabilitytaxonomy, terminologyanddatacollection,"tech.rep. [98]D.KovalandI.Floyd,H.L.,\Humanelementfactorsreliabilityandsafety," IndustryApplications,IEEETransactionson ,vol.34,pp.406{414,Mar1998. 127 [99]Z.Daneshi-Far,G.Capolino,andH.Henao,\Reviewoffailuresandconditionmoni- toringinwindturbinegenerators,"in ElectricalMachines(ICEM),2010XIXInter- nationalConferenceon ,pp.1{6,Sept2010. [100]M.M.Patriksson,A.-B.Stromberg,andL.Bertling,\Aredundancyopti- mizationmodelappliedtowindturbinepowerconverters,"in PowerTech (POWERTECH),2013IEEEGrenoble ,pp.1{6,June2013. [101]G.Mandic,A.Nasiri,E.Muljadi,andF.Oyague,\Activetorquecontrolforgear- boxloadreductioninavariable-speedwindturbine," IndustryApplications,IEEE Transactionson ,vol.48,pp.2424{2432,Nov2012. [102]T.HoandL.Ran,\Astudyontheelectricalsystemarrangementforwind turbinesandfactorsthevoltagelevelchoice,"in PowerElectronicsfor DistributedGenerationSystems(PEDG),20123rdIEEEInternationalSymposium on ,pp.442{449,June2012. [103]S.Lentz,\Reliabilitymodelingofregionalcabledobservatories,"in OCEANS2007- Europe ,pp.1{6,June2007. [104]B.Jones,\Electricalnoiseasareliabilityindicatorinelectronicdevicesandcompo- nents," Circuits,DevicesandSystems,IEEProceedings- ,vol.149,pp.13{22,Feb 2002. [105]F.Blaabjerg,R.Teodorescu,M.Liserre,andA.Timbus,\Overviewofcontroland gridsynchronizationfordistributedpowergenerationsystems," IndustrialElectronics, IEEETransactionson ,vol.53,pp.1398{1409,Oct2006. [106]C.Busca,R.Teodorescu,F.Blaabjerg,S.Munk-Nielsen,L.Helle,T.Abeyasekera,and P.Rodriguez,\Anoverviewofthereliabilitypredictionrelatedaspectsofhighpower f IGBTs g inwindpowerapplications," MicroelectronicsReliability ,vol.51,no.911, pp.1903{1907,2011. [107]H.Wang,K.Ma,andF.Blaabjerg,\Designforreliabilityofpowerelectronicsystems," in IECON2012-38thAnnualConferenceonIEEEIndustrialElectronicsSociety , pp.33{44,Oct2012. [108]A.BharathiSankar,\Performanceanalysisofpwminverterforwinddrivendou- blyfedinductiongenerator,"in AdvancesinEngineering,ScienceandManagement (ICAESM),2012InternationalConferenceon ,pp.244{248,March2012. 128 [109]Y.SongandB.Wang,\Surveyonreliabilityofpowerelectronicsystems," Power Electronics,IEEETransactionson ,vol.28,pp.591{604,Jan2013. [110]F.BlaabjergandK.Ma,\Futureonpowerelectronicsforwindturbinesystems," EmergingandSelectedTopicsinPowerElectronics,IEEEJournalof ,vol.1,pp.139{ 152,Sept2013. [111]H.Lu,C.Bailey,andC.Yin,\Designforreliabilityofpowerelectronicsmodules," MicroelectronicsReliability ,vol.49,no.911,pp.1250{1255,2009.20thEuropean SymposiumontheReliabilityofElectronDevices,FailurePhysicsandAnalysis20th EuropeanSymposiumontheReliabilityofElectronDevices,FailurePhysicsandAnal- ysis. [112]F.KanellosandN.Hatziargyriou,\Optimalcontrolofvariablespeedwindturbines inislandedmodeofoperation," EnergyConversion,IEEETransactionson ,vol.25, pp.1142{1151,Dec2010. [113]J.Carrasco,L.Franquelo,J.Bialasiewicz,E.Galvan,R.Guisado,M.Prats,J.Leon, andN.Moreno-Alfonso,\Power-electronicsystemsforthegridintegrationofrenewable energysources:Asurvey," IndustrialElectronics,IEEETransactionson ,vol.53, pp.1002{1016,June2006. [114]\Ananalyticalliteraturereviewofstand-alonewindenergyconversionsystemsfrom generatorviewpoint," RenewableandSustainableEnergyReviews ,vol.28,no.0, pp.597{615,2013. [115]E.Bueno,A.Hernandez,F.Rodriguez,C.Giron,R.Mateos,andS.Cobreces,\Adsp andfpgabasedindustrialcontrolwithhighspeedcommunicationinterfacesforgrid convertersappliedtodistributedpowergenerationsystems," IndustrialElectronics, IEEETransactionson ,vol.56,pp.654{669,March2009. [116]L.Yaolin,P.Xuange,andL.Peipei,\Basedoncanbuswindgeneratingsetonline monitoringandfaultdiagnosissystem,"in SustainableEnergyTechnologies,2008. ICSET2008.IEEEInternationalConferenceon ,pp.195{197,Nov2008. [117]I.Abdelsalam,G.Adam,D.Holliday,andB.Williams,\Assessmentofawinden- ergyconversionsystembasedonasix-phasepermanentmagnetsynchronousgenerator withatwelve-pulsepwmcurrentsourceconverter,"in ECCEAsiaDownunder(ECCE Asia),2013IEEE ,pp.849{854,June2013. 129 [118]F.Blaabjerg,Z.Chen,andS.Kjaer,\Powerelectronicsastinterfaceindis- persedpowergenerationsystems," PowerElectronics,IEEETransactionson ,vol.19, pp.1184{1194,Sept2004. [119]S.Zhang,K.-J.Tseng,D.Vilathgamuwa,T.D.Nguyen,andX.-Y.Wang,\Design ofarobustgridinterfacesystemforpmsg-basedwindturbinegenerators," Industrial Electronics,IEEETransactionson ,vol.58,pp.316{328,Jan2011. [120]A.DobakhshariandM.Fotuhi-Firuzabad,\Areliabilitymodeloflargewindfarmsfor powersystemadequacystudies," EnergyConversion,IEEETransactionson ,vol.24, pp.792{801,Sept2009. [121]G.VanBusselandM.Zaaijer,\Reliability,availabilityandmaintenanceaspectsof large-scalewindfarms,aconceptsstudy,"in ProceedingsofMAREC ,vol.2001, 2001. [122]R.AllanandP.C.Avella,\Reliabilityandeconomicassessmentofgeneratingsystems containingwindenergysources," IEEProceedingsC(Generation,Transmissionand Distribution) ,vol.132,no.1,pp.8{13,1985. [123]R.BillintonandG.Bai,\Generatingcapacityadequacyassociatedwithwindenergy," EnergyConversion,IEEETransactionson ,vol.19,pp.641{646,Sept2004. [124]P.GiorsettoandK.F.Utsurogi,\Developmentofanewprocedureforreliabilitymod- elingofwindturbinegenerators," PowerApparatusandSystems,IEEETransactions on ,vol.PAS-102,pp.134{143,Jan1983. [125]FP7-ENERGY-2007-1-RTD,\DeliverableD1.1.eportonpreviouspublicationsdealing withwindturbinereliability,,"tech.rep.,2007. [126]K.Xiangyu,F.Dazhong,andT.Chung,\Economyandreliabilityevaluationofgener- atingincludingwindenergysystems,"in ElectricUtilityDeregulationandRestructur- ingandPowerTechnologies,2008.DRPT2008.ThirdInternationalConferenceon , pp.2653{2657,IEEE,2008. [127]A.Ehsani,M.Fotuhi,A.Abbaspour,andA.Ranjbar,\Ananalyticalmethodforthe reliabilityevaluationofwindenergysystems,"in TENCON20052005IEEERegion 10 ,pp.1{7,Nov2005. [128]D.A.FiruzeabadM.F.,\Relibility-basedselectionofwindturbinesforlargescalewind farm," WorldAcademyofScience,EngineeringandTechnology ,2009. 130 [129]M.Zhao,Z.Chen,andF.Blaabjerg,\Generationratioavailabilityassessmentof electricalsystemsforwindfarms," EnergyConversion,IEEETransactions on ,vol.22,pp.755{763,Sept2007. [130]W.Abdel-Latif,A.Alaboudy,H.Mostafa,andM.Fekry,\Evaluationandmitigation ofvoltagekercausedbyconstantspeedwindturbines,"in InnovativeSmartGrid Technologies-MiddleEast(ISGTMiddleEast),2011IEEEPESConferenceon ,pp.1{ 6,Dec2011. [131]P.ClarksonandP.Wright,\Sensitivityanalysisofkermeterimplementationsto waveformsfortestingtotherequirementsofiec61000-4-15," Science,Measurement Technology,IET ,vol.4,pp.125{135,May2010. [132]N.More,\NewEnergyNews,year=2013,url= http://nenmore.blogspot.com/2011/03/uk-study-clears-wind-again.html,urldate =2011-3-19." [133]S.Tentzerakis,N.Paraskevopoulou,S.Papathanassiou,andP.Papadopoulos,\Mea- surementofwindfarmharmonicemissions,"in PowerElectronicsSpecialistsConfer- ence,2008.PESC2008.IEEE ,pp.1769{1775,June2008. [134]L.Sainz,J.Mesas,R.Teodorescu,andP.Rodriguez,\Deterministicandstochastic studyofwindfarmharmoniccurrents," EnergyConversion,IEEETransactionson , vol.25,pp.1071{1080,Dec2010. [135]I.STANDARD,\Iec61400-21windturbinegeneratorsystems,part21:Measurement andassessmentofpowerqualitycharacteristicsofgridconnectedwindturbines,"tech. rep. [136]S.LumbrerasandA.Ramos,windfarmelectricaldesign:areview.," Wind Energy ,vol.16,no.3,pp.459{473,2013. [137]S.Lundberg, ationstudyoflargewindparks,Goteborg,Sweden .PhDthesis, ChalmersUniversityofTechnology,DepartmentofElectricPowerEngineering,2003. [138]W.KuoandM.J.Zuo, Optimalreliabilitymodeling:principlesandapplications .John Wiley&Sons,2003. [139]A.Rocha,B.Filho,I.Pires,andH.DePaula,\Fault-tolerantmedium-voltagepower convertersforhigh-capacitybeltconveyorsystems,"in IndustrialElectronicsSociety, IECON2014-40thAnnualConferenceoftheIEEE ,pp.4471{4477,Oct2014. 131 [140]K.Ma,A.Bahman,S.Beczkowski,andF.Blaabjerg,\Completelossandthermal modelofpowersemiconductorsincludingdeviceratinginformation," PowerElectron- ics,IEEETransactionson ,vol.30,pp.2556{2569,May2015. [141]Z.Zhou,M.Khanniche,P.Igic,S.Kong,M.Towers,andP.Mawby,\Afastpower losscalculationmethodforlongrealtimethermalsimulationofigbtmodulesfora three-phaseinvertersystem,"in PowerElectronicsandApplications,2005European Conferenceon ,pp.9pp.{P.10,2005. [142]J.KolarandF.Zach,\Lossesinpwminvertersusingigbts," ElectricPowerApplica- tions,IEEProceedings- ,vol.142,pp.285{288,Jul1995. [143]Y.Yu,T.-Y.Lee,andV.Chiriac,\Compactthermalresistor-capacitor-networkap- proachtopredictingtransientjunctiontemperaturesofapowermodule," Components,PackagingandManufacturingTechnology,IEEETransactionson ,vol.2, pp.1172{1181,July2012. [144]D.Murdock,J.Torres,J.Connors,andR.Lorenz,\Activethermalcontrolofpower electronicmodules," IndustryApplications,IEEETransactionson ,vol.42,pp.552{ 558,March2006. [145]S.Clemente,\Transientthermalresponseofpowersemiconductorstoshortpower pulses," PowerElectronics,IEEETransactionson ,vol.8,pp.337{341,Oct1993. [146]L.Yang,G.Yang,Z.Xu,Z.Dong,K.Wong,andX.Ma,\Optimalcontrollerdesign ofadoubly-fedinductiongeneratorwindturbinesystemforsmallsignalstability enhancement," Generation,TransmissionDistribution,IET ,vol.4,pp.579{597,May 2010. [147]D.W.NovotnyandT.A.Lipo,\Vectorcontrolanddynamicsofacdrives,"in Oxford , ClarendonPress,1996. [148]A.Ammous,K.Ammous,M.Ayedi,Y.Ounajjar,andF.Sellami,\Anadvancedpwm- switchmodelincludingsemiconductordevicenonlinearities," PowerElectronics,IEEE Transactionson ,vol.18,pp.1230{1237,Sept2003. [149]K.Ma,M.Liserre,F.Blaabjerg,andT.Kerekes,\Thermalloadingandlifetimeesti- mationforpowerdeviceconsideringmissioninwindpowerconverter," Power Electronics,IEEETransactionson ,vol.30,pp.590{602,Feb2015. 132 [150]K.Ma,M.Liserre,F.Blaabjerg,andT.Kerekes,\Thermalloadingandlifetimeesti- mationforpowerdeviceconsideringmissioninwindpowerconverter," Power Electronics,IEEETransactionson ,vol.PP,no.99,pp.1{1,2014. [151]H.Huang,A.Bryant,andP.Mawby,\Electro-thermalmodellingofthreephasein- verter,"in PowerElectronicsandApplications(EPE2011),Proceedingsofthe2011- 14thEuropeanConferenceon ,pp.1{7,Aug2011. [152]C.Lalanne,\Mechanicalvibrationandshockanalysis,fatiguedamage,"JohnWiley andSons,2014. [153]L.Wei,J.McGuire,andR.Lukaszewski,\Analysisofpwmfrequencycontroltoim- provethelifetimeofpwminverter," IndustryApplications,IEEETransactionson , vol.47,pp.922{929,March2011. [154]A.Hangleiter,A.Grabmaier,andG.Fuchs,\Anomalousdampinginmqwlasersdueto slowinter-welltransport," ElectronDevices,IEEETransactionson ,vol.40,pp.2106{, Nov1993. [155]A.Isidoril,F.Rossi,F.Blaabjerg,andK.Ma,\Thermalloadingandreliabilityof 10-mwmultilevelwindpowerconverteraterentwindroughnessclasses," Industry Applications,IEEETransactionson ,vol.50,pp.484{494,Jan2014. [156]K.Ma,M.Liserre,F.Blaabjerg,andT.Kerekes,\Thermalloadingandlifetimeesti- mationforpowerdeviceconsideringmissioninwindpowerconverter," Power Electronics,IEEETransactionson ,vol.30,pp.590{602,Feb2015. [157]H.-T.K.Yung-LiLee,MarkE.Barkey [158]I.Kovacevic,U.Drofenik,andJ.Kolar,\Newphysicalmodelforlifetimeestima- tionofpowermodules,"in PowerElectronicsConference(IPEC),2010International , pp.2106{2114,June2010. [159]A.Nies lony,\Determinationoffragmentsofmultiaxialserviceloadingstrongly encingthefatigueofmachinecomponents," MechanicalSystemsandSignalProcessing , vol.23,no.8,pp.2712{2721,2009. [160]N.Kaminski,\LoadCycleCapabilityofHiPaks,ApplicationNotes5SYA2043-01," tech.rep.,ABBSwitzerlandLtd,Semiconductors,Sep2004. [161]H.HuangandP.Mawby,\Alifetimeestimationtechniqueforvoltagesourceinverters," PowerElectronics,IEEETransactionson ,vol.28,pp.4113{4119,Aug2013. 133 [162]M.MusallamandC.Johnson,\Antimplementationofthewcounting algorithmforlifeconsumptionestimation," Reliability,IEEETransactionson ,vol.61, pp.978{986,Dec2012. 134