类均值核主元法在GIS局部放电模式识别中的应用研究  

GIS partial discharge pattern recognition research based on class kernel mean principal component analysis

作　　者：何樱[1] 华征[2] 侯智剑 王召盟 

机构地区：[1]中国电力科学研究院,北京100192 [2]华北电力大学河北省输变电设备安全防御重点实验室,河北保定071003

出　　处：《电测与仪表》2016年第2期84-89,共6页Electrical Measurement & Instrumentation

基　　金：国家高技术研究发展计划项目(863计划)(2011AA05A121)

摘　　要：GIS局部放电的模式识别对于评估其运行状态、确定检修策略具有重要意义。论文设计了4种典型的GIS局部放电模型,并通过实验建立了相应的局部放电超高频信号图谱数据库,然后根据信号特点提取了原始特征量,由于原始特征量维数较高,不利于模式识别,因此论文引入类均值核主元分析法,首先求出各类映射数据的类均值矢量,然后根据建立的类均值核矩阵建立类均值核主元算法。研究结果表明,该方法得到的特征量涵盖原始样本中的全部信息,并且维数低于绝缘缺陷种类数,能够实现信息的无损降维。GIS partial discharge pattern strategy determination. The author has d recognition is an important part of its state evaluation, and its maintenance esigned four kinds of typical partial discharge models in laboratory, then es- tablished corresponding partial discharge UHF signal mapping database through the experimental method, and also ex- tracted the original feature parameters, because the original characteristic dimension is high, which is bad for pattern recognition, based on this, this paper uses a species mean kernel principal component analysis method, it mapped the partial discharge original data samples to high-dimensional feature space. Firstly, it calculates all kinds of class mean vector data, and then builds the class average kernel matrix, finally, the class kernel mean principal component analysis algorithm is established. Research results show that characteristic of this method contains all the information of the original data, and the dimension is less than GIS insulation defect category numbers, and it can realize data dimension reduction without information loss, which improves the pattern recognition rate.

关 键 词：气体绝缘组合电器 局部放电 信息无损降维 特征提取 模式识别 

分 类 号：TM835.4[电气工程—高电压与绝缘技术]