基于小波矩特征信息的沿面放电阶段识别方法

摘要

沿面放电是电力变压器绝缘局部放电的主要形式之一,开展沿面放电发展特征的研究,对甄别变压器潜伏性故障具有重要意义,为此,根据典型的沿面放电模型,利用恒压法试验对其发展特征进行研究分析.针对沿面放电不同时间的信号样本,提取不同时间阶段局部放电灰度图像的小波矩特征参量.通过对特征参量的无监督系统聚类分析,建立一种基于聚类-多级支持向量机 (support vector machine,SVM)的不同放电阶段识别机制,将整个放电过程划分为了初始阶段、发展阶段、稳定阶段、预击穿阶段,为局部放电发展特性研究提出了一种创新方法.%Surface discharge is one of main forms of partial discharge (PD) of power transformer insulation.It is of importance to study development features of surface discharge for identifying latent faults of the transformer.Therefore, according to typical surface discharge model, this paper studies and analyzes development features of surface discharge by using constant voltage method for experiment.For signal samples at different time of surface discharge, it extracts wavelet moment feature parameters of PD grey images in different stages.By means of unsupervised clustering analysis on feature parameters, it establishes an identification mechanism for different discharge stages based on clustering-multistage support vector machine (SVM) to divide the whole discharge process into four stages including initial stage, development stage, stabilization stage and pre-breakdown stage.This mechanism has presented a kind of innovative method for studying PD development features.