基于RBF神经网络的无刷直流电机无位置传感器控制

RBFBasedonRBFNeuralNetworkPositionSensorlessControlforBrushlessDCMotors(300072)XiaChangliangWenDeFanJuanYangXiaojun(TianjinUniversity300072China),(RBF)RBF,k2(RLS),,,,::TM92115AbstractInthispaper,theprincipleofpositionsensorlesscontrolforbrushlessDCMotors(BLDCM)isanalyzed,andanewcontrolmethodforBLDCMisproposed,whichisbasedonradialbasisfunction(RBF)neuralnetwork.Thenetworkgetstheinitialcentersthroughak2meansalgo2rithm,andgetstheweightsthroughrecursiveleastsquares(RLS)inoff-linetraining.Inon-linetraining,thenetworkupdatesthenetworkparametersusingagradientdescendingerroralgorithm.Theoutputsofthenetwork,rectifiedbythelogicalmodule,areregardedasthecontrolsignalsaswellastheteacherfornetworktraining.Bymappingtheterminalvoltagesandphasecurrentstocom2municationsignals,thenetworkcanreplacethetraditionalpositionsensors.Thetheoryinthispaperisverifiedbytheexperimentalresults.Keywords:BrushlessDCmotor,radialbasisfunction,neuralnetwork,positionsensor(013800811),1968,,,,1977,,1,,,()()[1],,,,[2];,,[1];,,,,()17320026,[3],(BP),RBFRBF,,,BP,,,[4]RBF,k2[5]RBF,[6],,,,,TITMS320LF2407,,RBF2Y,vavbvc=R000R000Riaibic+(L-M)ddtiaibic+ddtm()m(-2/3)m(-4/3)(1)va,vb,vcia,ib,icL,MRmm,m,,(1),m,,,,,(1)mx(n)=mx(n-1)+Tvx(n-1)-(T+L-M)ix(n-1)+(L-M)ix(n)x=abc(2),Sx(n)Sx(n)=f[vx(n-1),ix(n-1),ix(n)](3)Sx(n){-1,0,1},x,SaSbScSa+Sb+Sc=0(4)(4),SaSbSc,,,(4),Sa(n)=fa[va(n-1),vb(n-1),ia(n-1),ib(n-1),ia(n),ib(n)]Sb(n)=fb[va(n-1),vb(n-1),ia(n-1),ib(n-1),ia(n),ib(n)]Sc(n)=-[Sa(n)+Sb(n)](5)3311RBF,RBF,(),,,,,,[6]172173RBF1RBFFig11ThetopologyofRBFNetwork(5),X=(va(n-1),vb(n-1),ia(n-1),ib(n-1),ia(n),ib(n)),Sa(n),Sb(n);RBFt(X)=Gt(X)=e(-Xi-Xt2/t)(6)XiXttttSo(Xi)=6kt=0wtot(Xi)o=a,b(7)kwto312,,k2,:(1)k:Z01Z02Z0k;(2)nX-ZnjX-Zni,XRniRniZni;(3)Zn+1j=1Nj6XRnjXj=1,2k(8)NjRj(4)Zn+1j=Znj,,(2),dm,Nj,[6]t=dm/2Nj(9),(RLS),313,,22RBFFig12RBFnetworkon-linelearningfiguration,,,[7]E(n)=126nj=n-Me2jej=dj(Xj)-S3j(Xj)(10)nMdjS3j:(1)9E(n)9wto(n)=6ni=n-Mei(n)t(Xi)wto(n+1)=wto(n)-19E(n)9wto(n)(11)(2)9E(n)9Xt(n)=2wto(n)6ni=n-Mei(n)t(Xi)t(n)[Xi-Xt(n)]Xt(n+1)=Xt(n)-29E(n)9Xt(n)(12)8220026(3)9E(n)9t(n)=-wto(n)6ni=n-Mei(n)t(Xi)QjiQit=[Xi-Xt(n)][Xi-Xt(n)]Tt(n+1)=t(n)-39E(n)9t(n)(13)1,2,3Sf(S(n))=-1S(n)-0170-0125S(n)01251S(n)017f(S(n-1))(14):(1)Sa(n),Sb(n)0;(2)(1),;(3)(5)Sc(n)RBF,4TI(DSP)TMS320LF2407,PC,DSP33Fig13Configurationofsystemcontrol:UN=36V;Tl=0106Nm;nN=2500r/min;R=115;L-M=3147mH(113%);Ke=0.048V/(rad/s);J=410-4kgm2120r/min2700r/min120r/min,,4SaHa4,,,,,5RBFA4SaHaFig14ThelogicalsignalSaandHallpositionsignalHaatratedspeed5AFig15ThecurrentwaveformofphaseAatratedspeed5,,RBF,RBF(76)92173RBF,30,22,,2Tab12Theresultofrecognition(%)96313585124,,,,,,,,,1BartnikasR1Detectionofpartialdischarge(corona)inelec2tricalapparatus1IEEETrans1EI,1990,25(1)1111202WifriedH1Partialdischarge-part5:Partialdischargede2tectioninrotatingmachines1IEEEElectricalMagazine,1992,8(3)21323,,,11,2001,20(2)20244,,,11:,20005,,,11(),2001,35(10)109811002001-09-10(29),,,,1ErtugrulN,AcarnleyP.Anewalgorithmforsensorlessoperationofpermanentmagnentmotors[J].IEEETrans.onIndustryApplications,1994,30(1)1261332IizukaK,UzuhashiH,KanoM,etal.Microcomputercontrolforsensorlessbrushlessmotor[J].IEEETrans.onIndustryApplication,1985,IA-215956013El-SharkawiMA,El-SamahyAA,El-SayedML.HighperformancedriveofDCbrushlessmotorsusingneuralnetwork[J].IEEETrans.onEnergyconversion.1994,9(2)3173224MulgrewB.Applyingradialbasisfuctions.IEEESignalProcessingMagazine[J],1996,13(2)50655,1[M]1:,19916,1[M]1:,19997,1[M]1:,19882002-02-04(66)1JenqYC,CrosbyPhil.Sinewaveparameterestimational2gorithmwithapplicationtowaveformdigitizereffectivebitsmeasurement.IEEETransactiononInstrumentationandMeasurement,1988,37(4)529532211:,199731()1:,19902001-09-296720026