基于粗糙集降维理论的GIS超高频局放包络模式识别方法  

Pattern Recognition Method of UHF PD Envelope Signal in GIS Based on Rough Set Reduction

作　　者：李清[1] 段大鹏 邱武斌[1] 蔡开穗 盛戈皞[3] 黄兴泉[1] 江秀臣[3] 

机构地区：[1]河南电力试验研究院,郑州450000 [2]北京电力科学研究院,北京100000 [3]上海交通大学电子信息与电气工程学院,上海200240

出　　处：《高压电器》2012年第3期6-11,共6页High Voltage Apparatus

基　　金：上海市科委资助项目(10dz1203000)~~

摘　　要：判断局部放电类型与局放包络信号之间的关系是实现评估气体绝缘金属封闭开关设备(GIS)的绝缘状态以及制定合理的维修策略的重要前提。笔者利用粗糙集降维理论,先对表征UHF PD包络信号的37个特征向量所组成的特征矩阵进行了离散化与差别矩阵约简,然后利用简约后的特征向量结合BP神经网络分类器,对4种不同类型的UHF PD包络信号进行了模式识别,结果表明此方法具有较高的识别率。Determining the relationship between partial discharge（PD） type and PD envelope signal is an important premise to assess the insulation state of gas-insulated metal-enclosed switchgear（GIS） and to make rational maintenance strategy.The matrix of 37 feature vectors,with which a UHF PD envelope signal is characterized,is discretized,and its difference matrix is reduced with the dimension reduction rough set theory.Then the reduced feature vectors combining with the BP neural network classifier are applied to pattern recognition of four different types of UHF PD envelope signals.The results show a high recognition rate of the proposed method.

关 键 词：UHF PD包络信号 局放类型 模式识别 粗糙集降维 BP神经网络 

分 类 号：TM855[电气工程—高电压与绝缘技术] TM595