基于主成分分析和概率神经网络的变压器局部放电模式识别  

Pattern recognition based on principal component analysis and probabilistic neural networks for partial discharge of power transformer

作　　者：苑津莎[1] 尚海昆[1] 

机构地区：[1]华北电力大学电气与电子工程学院,河北保定071003

出　　处：《电力自动化设备》2013年第6期27-31,共5页Electric Power Automation Equipment

基　　金：中央高校基本科研业务费专项资金资助项目(13X-S26)~~

摘　　要：提出利用主成分分析(PCA)的方法对变压器局部放电原始特征参数进行降维,并提取出新的主成分因子。结果表明,提取出的主成分因子可以很好地表征原始特征向量。通过概率神经网络(PNN)分类器分别对降维前和降维后的特征向量进行训练和识别。研究发现,提取出的新因子有效缓解了分类器负担,且PNN分类器的识别效果良好,优于传统BP神经网络分类器。For reducing the high dimension of original characteristic parameters in partial discharge pattern recognition of power transformer,PCA（Principal Component Analysis） is applied to extract new principal component factors,which represents the original characteristic parameters sufficiently.PNN（Probabilistic Neural Network） classifier is used to train and recognize the characteristic vectors before and after the dimension reduction respectively.It is found that,the extraction of new principal component factors mitigates effectively the load of PNN classifier and its effectiveness of pattern recognition is better than that of traditional BPNN classifier.

关 键 词：主成分分析 概率神经网络 变压器 局部放电 模式识别 

分 类 号：TM41[电气工程—电器]