电力系统暂时过电压多级支持向量机分层识别

Temporary overvoltage layered pattern identification based on multistage support vector machine in power system

提出了一种电力系统暂时过电压多级支持向量机(M-SVM)分层识别的方法。根据暂时过电压分类,建立暂时过电压分层识别系统,并采用'二分树'法构建多级支持向量机分类器。在变电站实测过电压数据的基础上,提取了三相及零序电压的时域统计特征和小波时频特征,同时对特征量进行逐级选择,将这些特征量作为M-SVM的输入,实现暂时过电压类型辨识。现场数据测试表明,采用的M-SVM分层识别方法具有训练样本少、训练时间短、识别率高的优点,可较好地应用于电力系统暂时过电压类型识别。

This paper proposes an approach of temporary overvoltage layered recognition based on multistage support vector machine in power system. Firstly, on the basis of temporary overvoltage classification, a temporary overvoltage multi-level recognition system is built, and the classifiers are constructed based on binary trees of multistage support vector machine. Then, according to field overvoltage data in a substation, time-domain statistical features and wavelet time-frequency features of three-phase and zero sequence voltage are extracted and selected step by step, and finally these features are used as input vectors of M-SVM to classify temporary overvoltage. The field data indicate that the proposed method has advantages of less training samples, shorter training time and high recognition rate, and can be well applied to power system temporary overvoltage identification.