第46卷第1期2023年2月电子器件ChineseJournalofElectronDevicesVol.46No.1Feb．2023项目来源:云南电网有限责任公司科技项目(059300KK52190006)收稿日期:2021－04－27修改日期:2021－12－21MechanicalFaultDetectionMethodinHighVoltageCircuitBreakersbyUsingHybridFeaturesandIntegratedExtremeLearningMachine*TIANYuan*，HUANGZuyuan，ZHANGHang，SUWenwei，GENGZhenwei，GAOYudou(InformationcenterofYunnanPowerGridCo．，Ltd．，KunmingYunnan650011，China)Abstract:Amechanicalfaultdetectionmethodforhighvoltagecircuitbreakerbasedonhybridfeatureextractionandintegratedextremelearningmachine(IELM)isproposed．Firstly，thevibrationsignalisdecomposedbyusingfullyintegratedadaptivenoiseempiricalmodedecomposition(CEEMDAN)toobtaintheintrinsicmodefunction(IMF)．Then，thetime-frequencymatrixisobtainedthroughsub-bandre-constructionofeachorderIMFcomponentcombinedwithHilberttransformandband-passfilter．Thetime-frequencymatrixistransformedintoenergymatrix，thefrequencybandisnormalizedbyusingnormalcumulativedistributionfunction(NCDF)，andthetime-frequencyen-tropyandsingularentropyareextractedtoformthemechanicalfaultfeaturevector．Inaddition，afaultclassificationsystemisestablished．TheadvantageofCEEMDANschemecombinedwithband-passfilteringisthatitcaneliminatemodalaliasing，reducetheadditionofaux-iliarynoiseandimprovethedecompositionefficiency．Inaddition，thenormalizedsingularentropyofNCDFhasmorestableperformance．IELMcomposedofmultipleweakclassescansolvetheshortcomingsoftraditionalextremelearningmachines．Experimentalresultsbasedonmeasureddatashowthatmechanicalfaultscanbeeffectivelydetectedbyusingthismethodthroughsmallsamples．Keywords:high-voltagecircuitbreakers;vibrationsignals;empiricalmodedecomposition;time-frequencyentropy;singularentropy;ex-tremelearningmachineEEACC:1210;7220doi:10．3969/j．issn．1005－9490．2023．01．034基于混合特征和集成极限学习机的高压断路器机械故障检测*田园*，黄祖源，张航，苏文伟，耿贞伟，高宇豆(云南电网有限责任公司信息中心，云南昆明650011)摘要:提出了一种基于混合特征提取和集成极限学...
