基于小波变换和HOG特征的变压器局部放电类型识别方法  

Recognition Method of Partial Discharge Type of Transformer Based on Wavelet Transform and Hog Feature

作　　者：李平 田秋松 霍明 陈熙伦 林雨 李佳伟 LI Ping;TIAN Qiusong;HUO Ming;CHEN Xilun;LIN Yu;LI Jiawei(Construction Branch,State Grid Inner Mongolia East Power Co.,Ltd.,Hohhot 010010,Nei Monggol,China)

机构地区：[1]国网内蒙古东部电力有限公司建设分公司,内蒙古呼和浩特010010

出　　处：《电气传动》2021年第24期52-56,62,共6页Electric Drive

摘　　要：不同类型的电力变压器局部放电对变压器绝缘造成的破坏程度不同,正确识别变压器局放类型对于评价变压器的绝缘状况至关重要。提出一种基于小波变换和梯度直方图(HOG)特征的变压器局放模式识别方法,首先根据变压器绝缘缺陷结构特点,设计制作了3种典型的局放缺陷模型,在实验室搭建测试平台并采用脉冲电流法获取变压器局放数据;其次对局放信号进行小波时频变换,获取局放信号的时频谱图并对该时频谱图进行灰度化和归一化处理;最后利用HOG算法提取局放时频谱图上的特征参量并送入分类器,实现变压器不同类型局部放电的模式识别。识别结果表明,该方法的平均识别准确率高达98%,能够有效识别变压器放电类型。The damage degree of transformer insulation caused by partial discharge(PD)will vary from one type to another type,so correctly identifying PD type is very important for evaluating the insulation status of the transformer.A PD pattern recognition method based on wavelet transform and histogram of oriented gradient(HOG)features was proposed.Firstly,three typical PD models were designed according to the structure characteristics of insulation defect in transformer and the PD experiment platform of transformer was built to collect PD signal by pulse current method.On this basis,the time-frequency(T-F)spectrum of PD signal was obtained by wavelet transform and then was grayed and normalized.Finally,the characteristic parameters on the T-F spectrum of PD were extracted by HOG algorithm and putted into classifier to realize PD pattern recognition.The experimental results show that the method has good performance for PD pattern recognition of transformer,and the average recognition accuracy is up to 98%.

关 键 词：电力变压器 局部放电 小波变换 梯度方向直方图 模式识别 

分 类 号：TM76[电气工程—电力系统及自动化]