资源描述：
《基于PLS分析的高压绝缘子污秽等级判定.pdf》由会员上传分享，免费在线阅读，更多相关内容在学术论文-天天文库。
1、ComputerEngineeringandApplications计算机工程与应用2014，50（15）223基于PLS分析的高压绝缘子污秽等级判定11222宋宛净，姚建刚，张彦，匡少林，孙谦11222SONGWanjing,YAOJiangang,ZHANGYan,KUANGShaolin,SUNQian1.湖南大学电气与信息工程学院，长沙4100822.湖南湖大华龙电气与信息技术有限公司，长沙4100821.CollegeofElectricalandInformationEngineering,HunanUniversity,
2、Changsha410082,China2.HunanUniversityHualongElectric&InformationTechnologyCo.,Ltd,Changsha410082,ChinaSONGWanjing,YAOJiangang,ZHANGYan,etal.Determinationofpollutionclassforhigh-voltageinsulatorsbasedonPartialLeastSquaresregressionanalysis.ComputerEngineeringandApplicati
3、ons,2014,50（15）：223-227.Abstract：Inviewofthecharacteristicsofmulti-correlationofinfraredthermographyofpollutedinsulators,amethodbasedonpartialleastsquaresregressionanalysisisproposedtodeterminethepollutionclassofhighvoltageinsulators.Onthepremiseofpreservingtheoriginald
4、atatothemaximumextent,thepartialleastsquaresregressionequationbetweenthecharacteristicparametersofhigh-voltageinsulatorcontaminationandcontaminationgradesisbuilt,andthroughanalyzingtheimportanceofindicatorsofvariableprojectionoftheregressionmodelequation,theeffectdegree
5、ofvariouscharac-teristicparametersoncontaminationgradesisobtained.Themethodsolvesthemulti-correlationproblemoftheindepen-dentvariableseffectively,andquantifiestherelationshipbetweenthecharacteristicparametersandcontaminationgrades.Thetestresultshowsthat,judgingthehigh-v
6、oltageinsulatorcontaminationgradesbyapplyingpartialleastsquaresregres-sionanalysisisscientificandreliable,withhighaccuracyandstrongpracticability.Keywords：insulatorspollutionclass;characteristicparameters;PartialLeastSquares（PLS）regression;modelequation;variableimportan
7、ceinprojection摘要：针对污秽绝缘子红外热像特征数据具有多重相关性的特点，提出基于PLS(PartialLeastSquares,PLS)回归分析的高压绝缘子污秽等级判定方法。在最大限度保留原有数据信息的前提下，建立起高压绝缘子污秽特征量与污秽等级之间的PLS回归模型方程，通过对回归模型方程进行变量投影重要性指标分析，可以得到各个特征量对污秽等级判定结果的影响程度。此方法有效解决了自变量之间的多重相关性问题，量化了污秽特征量与污秽等级之间的关系。测试结果表明，将PLS回归分析应用于高压绝缘子污秽等级的判定，科学可靠，准确率
8、高，具有较强的实用性。关键词：绝缘子污秽等级；特征量；偏最小二乘（PLS）；模型方程；变量投影重要性指标文献标志码：A中图分类号：TM83doi：10.3778/j.issn.1002-8331.1208-02591引言
