最小二乘支持向量机多分类法的变压器故障诊断

Fault Diagnosis of Transformer Using Multi-class Least Squares Support Vector Machine

为了提高变压器故障诊断正判率,提出了一种基于小样本的最小二乘支持向量机(LS-SVM)多分类电力变压器油中气体分析(DGA)法,即通过相关统计分析和数据的预处理,选择变压油中典型气体作为LS-SVM的输入,然后利用典型故障气体的体积分数在高维空间的分布特性诊断变压器故障类型。该法在小样本条件下可获得最优解,泛化能力很好,且没有传统支持向量机只能分两类的缺陷,很好地解决了变压器多种故障共存的实际情况。试验表明,该方法分类效果很好,可较好地解决变压器放电和过热共存时故障的难分辨问题,故障类型的正判率较高。

In order to improve the correct rate of dissolved gas analysis（DGA）, this paper investigates a method of the DGA of transformer based on multi-classification least squares support vector machine （LS-SVM）. The proposed approach is based on seeking the optimal solution by few training samples supporting, and it has important features such as good generalization, meanwhile, multi-classification LS-SVM makes up for deficiencies of traditional support vector machine, which can not distinguish among multi-fault but can distinguish the two types of failures, and it can resolve the actual problems that a great variety of failures coexist in power transformer. Based on correlation analysis and pretreatment, some key gases are selected as the inputs of LS-SVM, furthermore, the fault diagnosis is accomplished according to the concentration distribution of typical fault gases in higher dimensional space. This paper analyzes and compares the currently-used multi-classification methods. By discussing the experiment results, the proposed method of in this paper has very good classification results, and can figure out the problem that it is difficult to distinguish between failures when overheating and partial discharge coexist, meanwhile, its effectiveness and usefulness are proved.