Speed_sensorless_direct_torque_control_of_IMs_with

SpeedsensorlessdirecttorquecontrolofIMswithrotorresistanceestimationMuratBaruta,SetaBogosyanb,*,MetinGokasanaaDepartmentofElectricalandElectronicEngineering,FacultyofEngineering,IstanbulTechnicalUniversity,Maslak,34390Istanbul,TurkeybDepartmentofElectricalandComputerEngineering,UniversityofAlaska,P.O.Box750145,Fairbanks,AK997750145,USAReceived17January2004;accepted29April2004Availableonline2July2004AbstractDirecttorquecontrol(DTC)ofinductionmotors(IMs)requiresanaccurateknowledgeontheampli-tudeandangularpositionofthecontrolledfluxinadditiontotheinformationrelatedtoangularvelocityforvelocitycontrolapplications.However,unknownloadtorqueanduncertaintiesrelatedtostator/rotorresistancesduetooperatingconditionsconstitutemajorchallengesfortheperformanceofsuchsystems.Thedeterminationofstatorresistancecanbeperformedbymeasurements,butmethodsmustbedevelopedforestimationandidentificationofrotorresistanceandloadtorque.Inthisstudy,anEKFbasedsolutionissoughtfordeterminationoftherotorresistanceandloadtorqueaswellastheabovementionedstatesrequiredforDTC.TheEKFalgorithmusedinconjunctionwiththespeedsensorlessDTCistestedunderelevenscenarioscomprisedofvariouschangesmadeinthevelocityreferencebesidetheloadtorqueandrotorresistancevaluesassignedinthemodel.Withnoaprioriinformationintheestimatedstatesandparameters,ithasbeendemonstratedthattheEKFestimationandsensorlessDTCperformquitewellinspiteoftheuncertaintiesandvariationsimposedonthesystem.
2004ElsevierLtd.Allrightsreserved.Keywords:Inductionmotor;ExtendedKalmanfilter;Sensorlessdirecttorquecontrol;LoadtorqueandrotorresistanceestimationEnergyConversionandManagement46(2005)335–349*Correspondingauthor.Tel.:[phone];fax:[phone].E-mailaddress:[e-mail](S.Bogosyan).0196-8904/$-seefrontmatter
2004ElsevierLtd.Allrightsreserved.doi:10.1016/j.enconman.2004.04.0021.IntroductionHighefficiencycontrolandestimationtechniquesrelatedtoinductionmotors(IMs)havebeenfindingmoreandmoreapplicationwithBlaschke’swell-knownfieldorientedcontrol(FOC)method,establishedin1971.TherehasbeenanextensiveamountofworktoimprovethedynamicresponseandreducethecomplexityofFOCmethods.Onesuchtechniqueisthedirecttorquecontrol(DTC)methoddevelopedbyTakahashiin1984[1],whichhasbeengettingincreasedattentionduetotheimproveddynamicperformanceandsimplifiedcontrolstrategythatitofferswithrespecttotheFOCmethods.TheDTCmethodinvolvesthedirectchoiceoftheappropriate/optimumswitchingmodes,inordertokeepthefluxandtorqueerrorswithinapredeterminedbandlimit(inahysteresisband)[2].Theerrorsaredefinedasthedifferencebetweenthereferenceandthemeasured/estimatedvaluesoffluxandtorque.UnlikeFOCmethods,DTCtechniquesrequireutilizationofhysteresisbandcomparatorsinsteadoffluxandtorquecontrollers.Toreplacethecoordinatetransfor-mationsandpulsewidthmodulation(PWM)signalgeneratorsofFOC,DTCuseslook-uptablestoselecttheswitchingprocedurebasedontheinverterstates.However,bothmethodsrequireaccurateknowledgeoftheamplitudeandangularpositionofthecontrolledflux(withrespecttothestationarystatoraxis)inadditiontotheangularvelocityforvelocitycontrolapplications.Asiswellknown,speedsensorsliketachometersorincrementalencodersincreasethesizeandcostofsystemsunnecessarily.SimilarproblemsarisewiththeadditionofsearchcoilsorHalleffectsensorstothemotorformeasurementoftheflux,hinderingfunctionalityintermsofimplementation.Thus,toimproveoverallsystemperformance,stateestimatorsorobserversareusuallymorepreferablethanphysicalmeasurements.However,thefifthorderandnonlinearstructureoftheIMmodel[3],inadditiontothesensitivityofthesystemparameterstotem-perature[4]andfrequency[5],makesthedesignofobserversforIMsachallenge.Inthisregard,forhighperformancesensorlessvectorcontrolofIMs,itisessentialtoknowthetemperatureandfrequencydependentvariationofthestatorandrotorresistancesinadditiontotheloadtorque.Inastudyaddressingthisissue[6],ithasbeenstatedthatsimultaneousestimationofthestatorandrotorresistancesgivesrisetoinstability.Moreover,whilethevalueofthestatorresistancecouldbeobtainedbymeasuringthestatortemperature,therearephysicaldifficultiesindeterminingtherotorresistanceinasquirrelcageIM.Thus,estimatingtherotorresistanceandtheloadtorqueappearstobeareasonableapproach.InDTC,thefluxisconventionallyobtainedfromthestatorvoltagemodel,usingthemeasuredstatorvoltagesandcurrents.Thismethod,utilizingopenlooppureintegrationsuffersfromthewellknownproblemsofintegrationeffectsindigitalsystems,especiallyinthelowspeedoperationrange[7],evenwiththecorrectknowledgeofthestatorresistance.Moreover,itwillrequiretherotorangularvelocityforvelocitycontrolapplications.AmongtherecentstudiesconductingsimultaneousfluxandvelocityestimationforDTC,Ref.[8]studiedthesensitivitytoparametervariationswithanartificialneuralnetworkapproach,andarobustperformanceto50%variationsinthestatorresistancehasbeenobtainedinRef.[9]withaslidingmodeapproach,whiletheadaptivefluxobserverinRef.[10],theextendedLuenbergerobserverinRef.[11]andthenon-linearobserverinRef.[12]demonstraterobustnesstostepshapedloadtorquevariations.Amongstudi