同步发电机励磁电源设计软件部分文献翻译.docx
《同步发电机励磁电源设计软件部分文献翻译.docx》由会员分享,可在线阅读,更多相关《同步发电机励磁电源设计软件部分文献翻译.docx(15页珍藏版)》请在冰豆网上搜索。
同步发电机励磁电源设计软件部分文献翻译
南昌工程学院
2007级毕业(设计)论文文献翻译
机械与电气工程系(院)电气工程及其自动化专业
题目同步发电机励磁电源设计(软件部分)
学生姓名
班级07电气工程及其自动化
(1)班
学号
指导教师
日期2011年1月18日
南昌工程学院教务处订制
评语:
指导教师:
ExcitationControlofSelfExcitedInduction
GeneratorusingGeneticAlgorithmand
ArtificialNeuralNetwork
Abstract—Inductiongeneratorswhichmaybeoperatedingridorself-excitedmode,arefoundtobesuccessfulmachinesforwindenergyconversion.Outofthesetwoself-excitedmodeisgainingimportanceduetoitsabilitytoconvertthewindenergyintoelectricalenergyforlargevariationsinoperatingspeed.Howeverithasbeenfoundthatthesemachineexhibitsapoorvoltageregulation.Steady-stateanalysisofselfexcitedinductiongeneratorrevealsthatsuchgeneratorsarenotcapabletomaintaintheterminalvoltageandfrequencyintheabsenceofexpensivecontrollers.Inturnadditionofsuchcontrollersmayresultintoafallinpopularityofthismachineduetoitssimplicity.Anothersimplewaytocontroltheterminalvoltageisthroughexcitationcontrolusingseriescompensation.Inthispaperartificialintelligenttechniquesareusedtomodelthecontrolstrategyforproperreactivecompensationunderdifferentoperatingconditions.Geneticalgorithmalongwithartificialneuralnetworkhasbeenproposedtoestimatethevaluesofshuntandseriesexcitationcapacitancetomaintaintheterminalandloadvoltage.Simulatedresultsasfoundusingproposedcontroltechniqueareverifiedusingexperimentalresultsonatestmachine.Simulatedresultsarefoundtobeincloseagreementwithexperimentalresults.
Keywords—ArtificialNeuralNetwork,GeneticAlgorithm,SelfExcitedInductionGenerator,WindEnergyGeneration.
I.INTRODUCTION
Mostoftheelectricalpowergenerationacrosstheworldisduetotheuseoffossilfuel,whicharelimitedinreserve.Thesemayvanishfromearthifcontinuedtobeusedatthesamepace.Arapiddepletionoffossilfuelsaswellasafastgrowingpowerdemandhaspressurizedthescientiststothinkaboutnonconventionalsourcesofenergysuchaswindenergy,solarenergy,tidalenergy,etc.Useofsuchsourcesmaybehelpfultoretainourresourcesoffossilfuelforsomeadditionalyears.Scientistshaveobservedthatoutofallpotentialnonconventionalsourceswindenergyseemstobemoreattractiveandviable.Itisobservedthatwindscarryenormousamountofenergyandtheregionsinwhichstrongwindsprevailforasufficienttimeduringtheyearmayuseitforelectricalenergygeneration.Inadditiontothiswindenergygenerationprovidesacleanandpollutionfreeenvironmentanddoesnotleadtoglobalwarming.Furtherawindturbinegeneratormaybeaworthwhilepropositionforanisolatedremoteareaduetoabsenceofpowergrid.Therearemanyconsiderationsinthechoiceofgeneratorsforthewindturbineapplicationsandseveralviewsprevail.Howevermostoftheresearchersareinthefavourofinductiongeneratorsinself-excitedmodeduetoitsabilitytoconvertmechanicalpoweroverawiderangeofrotorspeeds.Inductiongeneratorsarealsopreferredduetoseveralotheradvantagessuchaslowcost,lessmaintenanceotheradvantagessuchaslowcost,lessmaintenanceandeasyoperation.Theself-excitedinductiongenerators(SEIG)arealsofoundsuitableforfewotherapplicationssuchastidalandminihydroelectricenergyconversion.Operationofinductiongeneratorinself-excitedmodeisusefulundervariablespeedoperationespeciallywhenwindspeedisfluctuatingwithinawiderange.ThereforeitbecomesthedutyofresearcherstoinvestigatethebehaviourofspecificproblemrelatedissuesofSEIG.TocomputethesteadystateperformanceofSEIG,researchersadopteddifferentmodels[1-12].Mainobservationwhichwaspointedoutincaseofself-excitedinductiongeneratorisitspoorvoltageregulation.Variousregulatingschemeswereproposedbyresearchpersons[13-25]toovercomethisissue.Howeverithasbeenrealizedthatsuchschemesmakesthesystemcomplicatedandexpensive.Reference[26]foundtheseriescompensationasimpleandcheapalternativetosuchschemes.Manyresearchscholars[27-31]studiedtheeffectsofseriescompensationusingdifferenttechniques.Ithasbeenobservedthatduringtheanalysisthedegreeofpolynomialequationinunknownfrequencyincreasesduetothepresenceofseriescapacitor.
InthispaperGAhasbeenproposedtoestimatethegeneratedfrequency,shuntandseriescapacitancefordifferentoperatingspeedsandwithdifferentloadconditions.AcontrolstrategyusingGAandANNcontrollerisproposedtocontroltheterminalandloadvoltageofSEIG.Simulatedresultshavebeenverifiedusingexperimentalobservationonatestmachine.
II.ARTIFICIALINTELLIGENTTECHNIQUES
A.ArtificialNeuralNetworks
Artificialneuralnetworks(ANN)[32-34],alsocalledparalleldistributedprocessingsystemsandconnectionist,aregenerallyusedforfunctionapproximation,nonlinearmodeling,systemidentification,patternassociation,patternclassificationetc.
Fig.1givestheANNarchitecture.
Thisnetworkhasanaturaltendencyforstoringexperimentalknowledgeandmakingitavailableforuse.Thisarchitecturewillnotcompletelybeconstrainedbytheproblemtobesolved.Thenumberofinputandoutputneuronsdependonthegivenproblembutthenumberofhiddenlayersandassociatedneuronswilldependonthedesigner.Inthispaper,thetwo-layerbackpropagationfeedforwardneuralnetworkhasbeenused.Atotalof5neuronsinhiddenlayer,twoneuronininputlayerandtwoneuroninoutputlayerhavebeenused.Thenetworkissetwith‘logsig’activationfunctionatthemiddlelayerand‘purelin’activationfunctionattheoutputlayer.Levenberg-Marquadrtmethodhasbeenadoptedforsupervisedlearning.Experimentaldatafortestmachine-1(seeAppendixI)andcomputeddatafromGAhavebeenusedfortrainingpurpose.
B.GeneticAlgorithm
Overthepastfewyears,manyresearchershavebeenpayingattentiontoreal-codedevolutionaryalgorithms,particularlyforsolvingreal-worldoptimizationproblems.GeneticAlgorithm(GA)[35]isoneofthem.Sincetheperformancevariablesevaluation(,CashandCse)inSEIGmaytakeanyrealnumber,therefore,inthispaperareal-codedgeneticalgorithmhasbeenusedtoinvestigatetheperformance.Inareal-codedGA,variablesarecodedinrealnumbersitself[36].GAoperatorsaredirectlyappliedontherealnumbers(,CshandCse)ThreemainoperatorsresponsiblefortheworkingoftheGAsarereproduction,crossover,andmutation.Reproductionoperatorallowshighlyproductivestringstoliveandreproduce,wheretheproductivityofanindividualisdefinedasastring’snon-negativeobjectivefunctionvalue.Therearemanywaystoachieveeffectivereproduction.Here,tournamentselectionisusedinsteadofroulette-wheelselection,whichisgenerallyused.Highervaluesoftournamentsizegivehigherselectionpressure,makingtheconvergencefaster.Thesecondoperator,crossover,exchangesgeneticinformation.Thestudyrevealsthatanumberofcrossoveroperatorssuchasblendcrossover(BLX),simulatedbinarycrossover(SBX),unimodalnormaldistributioncrossover(UNDX),simplexcrossover(SPX)arecommonlyused.Adetailedstudyofsuchoperatorscanbefoundin[37-40].Hereparentcentricrecombinationoperator(PCX)hasbeenused,asthisoperatorassignsmoreprobabilityforanoffspringtoremainclosertotheparentsthanawayfromparents.Thesearchpowerofthiscrossoverisbetterthanthesimplecrossoveri.e.localorbroadersearchcanbedone.Differentmutationoperatorsareusedbasedonthecodingofthevariable.Sincecontinuousvariablesarecodeddirectly,thealgorithmisflexibleinnature.AsPCXandthereal-codedmutationoperatorshavebeenusedtohaveasearchpowersimilartotheircounterparts,theoverallalgorithmperformsbetterthanthebinary-codedGAs.ModifiedGA[35]processesfastconvergenceincomparisontoconventionalGAasreportedin[41].
III.MODELINGFORSTEADYSTATEANALYSIS
Thesteady-stateoperationoftheself-excitedgeneratorwithseriesandshuntcapacitorsmaybeanalyzedbyusingtheequivalentcircuitrepresentationasshowninFig.2.
Fig.2.Perphaseequivalentcircuitrepresentationfortwo-capacitorself-excitedinductiongenerator.
A.SelectionofCshatNoLoad
GAhasbeenusedtofindtheshuntcapacitanceatdifferentspeedstomaintaintheratedvoltageacrossthestatorterminalsatnoload.Thefitnessfunction(FF)usedhereis,
Fig.2.Perphaseequivalentcircuitrepresentationfortwo-capacitorself-excitedinductiongenerator.
Inthiscircuitmodelallparametersareassumedtobeindependentofsaturationexceptformagnetizingreactance.Thecorelosseshavebeenignored.ThenetworkabovecanbefurthertransformedintoFig.3.
Where,
Forgivenvalueofoperatingspeedandshuntcapacitanceascalculatedabove,GAmaybeusedtofindtheoptimumvaluesofgeneratedfrequency,aandseriescapacitanceCseforagivenloadimpedanceandcorrespondingtooperatingspeed.
Thiswillleadtotheestimationof‘a’and‘Cse’throughGAforconstantvoltageforanyoperatingspeed‘b’andloadresistance‘R’.
Fig.4.Variationofseriescapacitancewithspeedatdifferentloadsforconstantloadvoltage.
Fig4showsthevariationofcomputedvaluesofCsewithoperatingspeedfordifferentvaluesofloadresistance.Itisobservedthatwithanincreaseinspeed,initiallythevalueofseriescapacitanceCsedecreasesandafterattainingaminimumvalueitstartsincreasing.Thesecharacteristicsarealsousefultodeterminetheoperatingspeedforanyloadresul