RIME-VMD-LSSVM在气体绝缘电器局放故障识别的应用  

Application of RIME-VMD-LSSVM in Partial Discharge Fault Identification of Gas-insulated Switchgear

作　　者：张超 张运 张士勇 高鹏 刘虹 ZHANG Chao;ZHANG Yun;ZHANG Shiyong;GAO Peng;LIU Hong(Yancheng Power Supply Branch,State Grid Jiangsu Electric Power Co.,Ltd.,Yancheng 224000,China)

机构地区：[1]国网江苏省电力有限公司盐城供电分公司,江苏盐城224000

出　　处：《电工技术》2024年第19期178-183,共6页Electric Engineering

基　　金：国网江苏省电力有限公司科技项目(编号J2023002)

摘　　要：气体绝缘组合电器中存在多种绝缘故障,准确识别GIS的故障类型对保障电力安全具有重要意义。为此,提出一种基于霜冰优化算法(Rime optimization algorithm,RIME)优化变分模态分解(VMD)与最小二乘支持向量机(LSSVM)的GIS局部放电分类识别方法。首先引入RIME以最小包络熵作为目标函数对VMD中K和α两参数进行优化。然后对IMFs进行选取,并采用峭度、裕度、波形提取特征。最后将提取的特征向量输入RIME-LSSVM进行识别诊断。经过对4种局放特高频信号进行处理分析,表明相比于传统算法,该方法的RIME-VMD-LSSVM诊断效果更好,能有效识别不同的绝缘缺陷故障,识别正确率相较于其他传统算法最高可提升约16%,对GIS等高压电力设备故障识别有进步意义。Gas insulated switchgear(GIS)may encounter various types of insulation faults,of which the accurate identification is of great significance to power safety.This work studied a GIS partial discharge(PD)classification and identification method based on the rime optimization algorithm(RIME),which optimizes variational mode decomposition(VMD)and least squares support vector machine(LSSVM).First the RIME was introduced with minimum envelope entropy as the objective function for K and in VMDαOptimize two parameters.Then IMFs was selected and features were extracted using kurtosis,margin,and waveform.Finally the extracted feature vectors were input into RIME-LSSVM for identification and diagnosis.The proposed method was indicated,by testing treatment of four types of PD ultra-high frequency signals,to have better diagnostic performance with accuracy improvement up to 16%and achieve effective identification of different fault types compared to the selected conventional algorithms,and thereby to be potentially significant for fault identification of high-voltage power equipment such as GIS.

关 键 词：GIS 局部放电 VMD 霜冰优化算法 最小二乘支持向量机 

分 类 号：TM561[电气工程—电器]