基于LS_SVM的电力电子电路故障预测方法

15820118ELECTRICMACHINESANDCONTROLVol.15No.8Aug．2011LS-SVM121111．2100162．232001leastsquaressupportvectormachineLS-SVM。BuckLS-SVM。LS-SVM2%。TP206A1007－449X201108－0064－05ResearchonfaultpredictionmethodofpowerelectroniccircuitsbasedonleastsquaressupportvectormachineJIANGYuan-yuan12WANGYou-ren1CUIJiang1SUNFeng-yan11．CollegeofAutomationEngineeringNanjingUniversityofAeronauticsandAstronauticsNanjing210016China2．CollegeofElectricandInformationEngineeringAnhuiUniversityofScienceandTechnologyHuainan232001ChinaAbstractAimingattheissueoffaultpredictiontechniqueofpowerelectroniccircuitsamethodbasedoncharacteristicparameterdataandleastsquaressupportvectormachineLS-SVMforthepredictionofpowerelectroniccircuitswasproposed．TakingtheBuckconvertercircuitasanexamplethefaultpre-dictionofpowerelectroniccircuitswasachieved．Firstlytheoutputvoltagewasselectedasmonitoringsignalandthentheaveragevoltageandripplevoltagewereextractedascharacteristicparameters．LastlyLS-SVMalgorithmwasusedtopredictBuckconvertercircuit．TheexperimentalresultsshowthattheLS-SVMalgorithmisespeciallyaccurateinpredictingtheaveragevoltageandripplevoltagewiththerelativeerrorlessthan2%．Thenewmethodcantracethecharacteristicparameters’trendandcanbeeffectivelyappliedinfaultpredictionofpowerelectroniccircuits．KeywordspowerelectroniccircuitsfaultpredictioncharacteristicparameterdatadrivingleastsquaressupportvectormachineLS-SVM2010－07－24608710092009ZD52045CXLX11－0183NS20100631982—1963—、、1977—、1983—。0、1－2。3。、。4、5、63。、。、。、。7－9。leastsquaressupportvectormachineLS-SVMLS-SVM。11.1。。1Buck。1MOSFETIRF151f=50kHzD=0.22L=43μHD1MUR405C=220μFESRRL=1.25Ω。1BuckFig．1Buckconvertercircuit1BuckDC-DC。。Buck、。Buck。。2a。uo1UoNuoii。Uo=1N∑Ni=1uoi。1b。。－。、－Δu。1.2LS-SVMLS－SVMSVM1011supportvectormachineregressionSVR。LS-SVMxHilbert。SVMfx=wTx+bw∈Rnb∈R。2fx·568LS-SVMxwwTwbRnnR。supportvectorSV2minwbeJwe=12‖w‖2+γ∑ni=1e2iS．t．yi=wTxi+b+ei}。3eiγxix=x1…xi…xniyiy=y1…yi…yni。LagrangeLagrangeαiLwebα=Jwe－∑ni=1αiwTxi+b+ei－yi。4Karush-Kuhn-TuckerLS-SVMfx=∑ni=1αiKxix+b。5Kxy=xyMercer、、、Sigmoid。1.3LS-SVM。2①。、1h、5h、24h②③、④⑤LS-SVM⑥LS-SVM⑦。21BuckPspiceuoMatlab7.6LS-SVMlabBuck。2Fig．2Flowchartofthefaultprediction2.1Buck812CESR、L、MOSFETRds、Vthgm11。1Buck2。66151BuckTable1TheparametersofthecomponentsinBuck/hC/μFRESR/mΩL/μHRds/mΩVth/Vgm/μ1220512431.62.8320.5221852342.93.62.8520.4321553242.85.72.8720.3421254442.68.02.8920.2520855742.4112.9120.1620557042.2152.9320720058342192.9519.9819460841.7242.9719.7918962141.5293.019.51018363741.1343.0319.22Table2Thecharacteristicparametersdata/h/V/V15.1370.76825.1350.78235.1320.79345.1300.80955.1260.82565.1230.84275.1180.85885.1120.88995.1060.905105.1000.9282.2LS-SVM121～67712～7684。LS-SVMgam=1003。3LS-SVMBuck2%。UoΔu3、4。○+1～67～104。3、4LS-SVMBuck。3Table3Predictionresultsofcharacteristicparameters3Fig．3Predictioncurvesofaveragevoltage4Fig．4Predictedcurvesofripplevoltage2BuckBuck。0.5V10%1V。5.1378V768LS-SVM0.7V340.5V1V4h。3SVM。SVM。SVM。。、、。1gam2gamsig2RBF4。4Table4Predictioneffectofdifferentkernelfunction%%gam=100.06570.9419gam=1000.04890.8196gam=100sig2=100.05941.2189gam=1000sig2=500.02980.54344。。3LS-SVM。Buck。。1LALLPHANDEMBHATCetal．Prognosticshealthmonito-ringPHMforprior-damageassessmentinelectronicsequipmentunderthermo-mechanicalloadsC//ProceedingsofElectronicComponentsandTechnologyConferenceMay29－June12007NewYorkUSA．20071097－1111．2MAZhangshan．AnewlifesystemapproachtotheprognosticandhealthmanagementPHMwithsurvivalanalysisdynamichybridfaultmodelsevolutionarygametheoryandthree-layersurvivabil-ityanalysisC//ProceedingsofIEEEAerospaceConferenceMarch7－142009BigSkyMontanaUSA．20091－20．3ORSAGHRBROWNDROEMERM．Prognostichealthman-agementforavionicssystempowersuppliesC//ProceedingsofIEEEAerospaceConferenceMarch5－122005BigSkyMon-tanaUSA．20053585－3591．4．ARJ．200930191－95．LVKehongQIUJingLIUGuanjun．ResearchonlifeprognosismethodforelectronicsbasedondynamicdamageandoptimizationARmodelJ．ActaArmamentarii200930191－95．5．BPJ．200926152－54．ZOUXinyaoYAORuohe．LifepredictionofelectronicdevicesbasedonforecastsystemofbackpropagationneuralnetworkJ．MicroelectronicsandComputer200926152－54．6．LS-SVMJ．200621177－80．JIANGTianhanShUJiong．Multi-steppredictionofchaotictimeseriesusingtheleastsquaressupportvectormachineJ．ControlandDecision200621177－80．7AMARALACARDOSOA．UseofESRtopredictfailureofout-putfilteringcapacitorsinboostconvertersC//ProceedingsofIEEEInternationalSymposiumonIndustrialElectronicsMay4－72004NewYorkUSA．IEEE20041309－1314．8IMAMADIVANDHABETLERTetal．Real-timeconditionmonitoringoftheelectrolyticcapacitorsforpowerelectronicsappli-cationsC//ProceedingsofIEEEAppliedPowerElectronicsCon-ferenceandExpositionFebruary25－March12007NewYorkUSA．IEEE20071057－1061．9CHENYaow-MingCHOUHsu-WeiLEEKungyen．Onlinefail-urepredictionoftheelectrolyticcapacitorforLCfilterofswitching-modepowerconvertersJ．IEEETransactionsonIndustrialElec-tronics2008551400－406．10．J．2007271062－66．74861511LIUJunHUANGMengzhiWANGYang．Researchonvector-controlsystemofPMSMbasedoninternalmodelcontrolofcurrentloopC//SecondInternationalWorkshoponComputerScienceandEngineeringOctorber28－