摘要
Inviewofthelimitationsinthepredictionofpollutionflashovervoltagebyleastsquaresregression,amethodtopredictpollutionflashovervoltagebyrobustregressionisproposed.Accordingtotestingvoltageandthedataofsaltdepositdensity(ρSDD)andnon-solubledepositdensity(ρNSDD),theregressioncoefficientissolvedbyacomplexweightingleastsquareiterationalgorithm.Initerativecalculations,theweightfunctionisadopted,inwhichtheweightcoefficientisthefunctionoftheresidualerroroflastiterationtoweakentheinfluenceofsingularvaluesontheregressioncoefficient.ThecharacteristicexponentdenotingρSDDinfluenceandcharacteristicexponentdenotingρNSDDinfluencearemappedbytheregressioncoefficient,andthusthepollutionflashovervoltageofinsulatorscanbepredicted.Throughthecomparisonoftestresults,robustregressionresultsandleastsquaresregressionresults,theeffectivenessoftheproposedrobustregression-basedforecastingmethodisverified.
出版日期
2011年05月15日(中国期刊网平台首次上网日期,不代表论文的发表时间)