1、長(zhǎng)沙理工大學(xué)碩士學(xué)位論文電力設(shè)備絕緣油智能化診斷系統(tǒng)姓名：龔杰申請(qǐng)學(xué)位級(jí)別：碩士專業(yè)：高電壓與絕緣技術(shù)指導(dǎo)教師：周力行20090401ABSTRACTThetypicaloil—nlledelectricalequipmentsincludetransformers，reactorsandoil—nlledmutuaIinductoretcIt’swidlyappliedin，powersystemAsthehubofthepowert

2、ransmissionequipment，theoperationwilldirectlyaf佗ctthesafetyofpowersystemsandengenderedasignincanteconomicimpactanddamagewithafaultoccurredTheref0re，theresearchoninsulationconditionofthetransf0rmer，detectthepotentialtrans

3、formerfailureearlytoensuretheirsafeoperationandimproVethereliabilityofpowersupplyisalongstandingimportantissuetotheelectricitysectorBasedoninsulatingoilanalysis，theresearchofthetransfbrmerfaultdiagnosistechnologyhaVegrea

4、tsignificancetoimproVetheleVeloftransformeroperationandmaintenanceDGAisaneffectiVemethodintransformerfaultdiagnosisThispaperanalyzedtheformationofgasintransformeroil，thetraditionalfaultdiagnosismethodsandartificialintell

5、igencemethodsindomesticandforeignByreadinglotsofliteratures，usedtheLeVenberg—MarquardtalgorithmtoinlproVetheconvergencerateofneuralnetworks，whilereducingthepossibilityoflocalminimumpointsinnetworkSelectedtheRBFnetworkinf

6、aultdiagnosisaRercomparedawiderangeofneuralnetworkmethodsThiskindofnetworkhasacertaincredibilitycon6rmedbyalargenumberoftheactualdiagnosisInordertoexpresstheambiguityofthetransformerfault，usedthefuzzymathematicstopretrea

7、tmentthedataofcharacteristicsgasBasedonthefuzzymath，theneuralnetworkfaultdiagnosismethodofinsulatingoilwasbuiltandapplicatedininsulatingoilofintelligentfaultdiagnosissystemInthissystem，combinedwiththecharacteristicsofthe

8、gasproductionrate，inlproVingIECthreeratiomethodtoconductacomprehensiVediagnosisUsedVisualBasiclanguagedeVelopmenttoolstobuildaintelligentfaultdiagnosissysteminthispaperDirectlycalledthedesignedprogramofneuralnetworkfIunc

9、tionbytheMatlabproceduresAnalysisanddiagnosistheundiagnosticdataThesystemisuserfriendlyoperatingeasily，runfast，accuratefaultdiagnosisIthasbeenusedinconVentionalproductionofEHVAuthorityinHunanProVinceKeywords：transformer；