摘要
Thechaoticcharacteristicsoftimeseriesoffivepartialdischarge(PD)patternsinoil-paperinsulationarestudied.TheresultsverifyobviouschaoticcharacteristicofthetimeseriesofdischargesignalsandthefactthatPDisachaoticprocess.Thesetimeserieshavedistinctivefeatures,andthechaoticattractorsobtainedfromtimeseriesdifferedgreatlyfromeachotherbyshapesinthephasespace,sotheycouldbeusedtoqualitativelyidentifythePDpatterns.Thephasespaceparametersareselected,thenthechaoticcharacteristicquantitiescanbeextracted.ThesequantitiescouldquantificationallycharacterizethePDpatterns.TheeffectsonpatternrecognitionofPRPDandCAPDarecomparedbyusingtheneuralnetworkofradialbasisfunction.Theresultsshowthatbothofthetworecognitionmethodsworkwellandhavetheirrespectiveadvantages.Then,boththestatisticaloperatorsunderPRPDmodeandthechaoticcharacteristicquantitiesunderCAPDmodeareselectedcomprehensivelyastheinputvectorsofneuralnetwork,andthePDpatternrecognitionaccuracyistherebygreatlyimproved.
出版日期
2011年06月16日(中国期刊网平台首次上网日期,不代表论文的发表时间)