電力變壓器局部放電超高頻在線監(jiān)測方法研究.pdf

1、重慶大學碩士學位論文電力變壓器局部放電超高頻在線監(jiān)測方法研究姓名：寧佳欣申請學位級別：碩士專業(yè)：電氣工程指導(dǎo)教師：李劍20070526重慶大學碩士學位論英文摘要AbstractInsulationdeterioration，whichismainlycausedbypartialdischarge口D)occurringinsidepowertransformers，isoneoftheprimereasonstocausetransf

2、oITnerfaultsInordertopreventaccidentfaultsandinsurestableperformanceofpowersystem，itiSvaluabletojudgeaccuratelytheconditionoftransformersthroughon—linemonitoringPDactivitiesoftransformersUltra—highfrequency(UHF)monitorin

3、gapproachisfocusedonrecentyearsbecauseofitseffectivenesstoavoidlowfrequencynoisesHowevertherealestillunsolvedproblemstoobstacletheonsiteapplicationofUHFon—linemonitoringsystemforPDintransformersInthispaperUHFonlinemonito

4、ringapproachforPDintransformersisstudiedonthebasisofconcludingandanalyzingtheresearchsituationofonlinemonitoringforPDactivities111ispaperconcentratesonthreeaspects：theoptimizeddesignofUHFsensor(UHFantenna)，interferencesu

5、ppressionofUHFsi驢a1andtherecognitionofUHFsignalallofwhichareshownbelow(1)ThebasicprinciplesofHilbertfractalantennaisintroducedandtheoptimizationanddesignapproachesofHilbertfractalantennaarepresentedforUHFonlinemonitoting

6、forPDintransformersbasedonfractalandantennamagnetictheoriesneperformanceofHilbertfractalantennaiSdiscussedandtheinfluenceofgeometryparameterstotheperformanceoftheantennaiSstudiedbasedonthemagneticsimulationsothvareAnsoft

7、Designer；a3rdHilbertfractalantennaisdesignedtakingthestructureoftransfoITllersintoaccount(2)AnimprovedwaveletdenoisingmethodispresentedtosuppressthewhitenoisemixedwithintheUHFsignalgeneratedbyPDactivitiesTheoptimalbasicw

8、aveletiscalculatedineachscalethroughanalyzingtheinfluenceofbasicwavelettosi鄂lalenergydecomposedineachscalenlemethodiSappliedtodenoiseUmsignalsgeneratedbyfourtypesofclassicartificialinsulationdefectsandthedenoisingresults

9、derivedfromdifferentthresh01dsandthresholdmethodsarealSOcomparedf31n圮differenceboxcountingmethodforfractaldimensionisstudiedtoextractfractalfeaturesfromwavelettoe伍cientsofUHFsignalgeneratedbyPDAccordingtotheextractedfeat

10、ure，radialbasisfunctionartificialneuronnetworkandprobabilityartificialneuronnetworkaspatternclassifierrespectivelyiSusedtecoghizefourtypesofUHFsignalsderivedfromartificialdefectsandtherecognitionresultsoftwoANNsarecompar