ConditioningAnalysisofParameterEstimationinaSynchronousGeneratorMiguelVklez-Reyes,SeniorMember,andLidaJauregui,StudentMemberCenterforPowerElectronicSystemsUniversityofPuertoRicoMayagiiezCampusP.O.Box9042,Mayaguez,PR00681-9042Tel.(787)265-3821,FAX(787)831-7564e-mail:mvelez@ece.uprm.edu,lida@ece.uprm.eduAbstract:ThispaperpresentstheconditioninganalysisoftheAparameterestimationproblemforasynchronousgenerator.wherey(a)istheN-vectorofmodelpredictionsfortheTheanalysisisperformedinasmallsignalmodelofthegenerator.Resultsdemonstratetheillconditioningoftheproblem.Methodologiestohandletheillconditioningarepresented.Theproposedmethodologiesrepresentdifferentwaystoincorporatepriorinformationaboutthesynchronousgeneratorparameterestimationproblemtoregularizedtheparameterestimates.I.INTRODUCTIONInpowersystemstudies,theusefulnessofthemodelofanV,(t)t40energyprocessingsystemcomponentdependsontheaccuracyofitsparameters.Inmanyinstances,thesemodelscansufferfromoverparameterizationortheavailabledataisParallelEstimationnotsufficientlyrichforaccurateparameteridentification.Ineithersituation,thenetresultisthattheresultingparameterAdaptationestimationproblemisillconditioned.ConditioninganalysisMechanismtreferstothemethodologyusedtostudythesensitivityoftheparameterestimatetoperturbationsintheproblemdataforFIG.1:OUTPUTERRORAPPROACHFORSYSTEMIDENTIFICATION.instance,noiseinthemeasuredvoltagesandcurrents.AnestimateisillconditionedifitisverysensitivetosuchandyistheN-vectorofactualmeasurements.Noticethatperturbations.predictionerrore(a)isanonlinearfunctionofa.TheThispaperpresentshowsensitivityanalysistechniquesareestimateofacanbefoundbyusedtostudytheconditioningofthelinearizedmodelofaAsynchronousgeneratorconnectedtoaninfinitebus.Herewea=argminV(a)showhow,usingrelativeconditionnumbers,the$conditioningofindividualmodelparameterestimatescanbequantified.ThisinformationcanbeusedtoassesstheneedwhereV(a)isameasureofmodelfittypicallyleastsquarestoincorporatepriorinfor