第47卷第4期电网技术Vol.47No.42023年4月PowerSystemTechnologyApr.2023文章编号：1000-3673（2023）04-1470-08中图分类号：TM721文献标志码：A学科代码：470·40基于改进灰狼算法与最小二乘支持向量机耦合的电力变压器故障诊断方法李云淏1，咸日常1，张海强2，赵飞龙1，李嘉洋1，王玮1，李增悦1（1．山东理工大学电气与电子工程学院，山东省淄博市255049；2．国网山东省电力公司淄博供电公司，山东省淄博市255000）FaultDiagnosisforPowerTransformersBasedonImprovedGreyWolfAlgorithmCoupledWithLeastSquaresSupportVectorMachineLIYunhao1,XIANRichang1,ZHANGHaiqiang2,ZHAOFeilong1,LIJiayang1,WANGWei1,LIZengyue1(1.CollegeofElectricalandElectronicEngineering,ShandongUniversityofTechnology,Zibo255049,ShandongProvince,China;2.ZiboPowerSupplyCompany,StateGridShandongElectricPowerCompany,Zibo255000,ShandongProvince,China)ABSTRACT:Inordertoachievetheaccuratefaultclassification,thearticleproposesafaultdiagnosismethodforpowertransformersbasedontheimprovedgreywolfalgorithmcoupledwiththeleastsquaressupportvectormachinebyusingfivetypicalgasesdissolvedintheoilasthecharacteristicquantityforthefaultdiagnosis.ThismethodseekstheoptimalpenaltycoefficientCandthekernelfunctionparametergintheleastsquaressupportvectormachinethroughtheimprovedgreywolfalgorithmtoimprovethefaultdiagnosisaccuracy.Firstly,theimprovementpointsoftheleastsquaressupportvectormachineandthegreywolfalgorithmareelucidatedandcoupled.Theyareputintothe413setsofdissolvedgasinoildetectiondataofthepowertransformersaresubstitutedtodiagnosethefaulttypesandcomparedwiththeotherdiagnosticmethods;secondly,theinfluencelawofthepenaltycoefficientCandthekernelfunctionparametergontheaccuracyofthefaulttypeidentificationofthepowertransformersisinvestigated;finally,withthehelpofthecouplingofthetrainedimprovedgreywolfalgorithmandtheleastsquaressupportvectormachine,theeffectivenessofthefaultdiagnosismethodisverifiedthroughtheanalysisoftwotransformerfaultexamplesunderdifferentvoltagelevels.Theresultsofthestudyshowthat,comparedwiththeleastsquaressupportvectormachineandthetraditi...
