基于LS-SVM和数学形态谱的局部放电模式识别  

The pattern recognition of partial discharge based on LS-SVM and pattern spectrum

作　　者：薛雷[1] 宋阳[1] 郭丹丹[1] 沈重[1] 

机构地区：[1]东北电力大学,吉林吉林132012

出　　处：《黑龙江电力》2011年第1期16-19,22,共5页Heilongjiang Electric Power

摘　　要：阐述了1种基于最小二乘支持向量机的局部放电模式识别的新方法。该方法采用多尺度数学形态学开运算提取局部放电信号的数学形态谱,通过数学形态谱可以看出每种类型局部放电具有不同的形态特征,把提取的形态特征作为最小二乘支持向量机的输入,然后应用最小二乘支持向量机的多分类算法来识别不同的放电类型。识别结果表明,该方法有效、可行,具有较高的识别率。To accurately recognize types of partial discharge,this paper proposes a new recognition method based on Least Squares Support Vector Machines.Multi-scale mathematical morphology is adopted to extract mathematical patter spectrum of partial discharge signal which presents different morphological features of every type of partial discharge.Different discharge type can be recognized by inputting extracted morphological feature as Least Squares Support Vector Machines and applying its multi-classification algorithm.Recognition result proves the feasibility and high recognition rate of this method.

关 键 词：局部放电 最小二乘支持向量机 数学形态学 形态谱 模式识别 

分 类 号：TM855[电气工程—高电压与绝缘技术]