基于SOM神经网络的变电站设备红外热像诊断研究

王佳林; 崔昊杨; 许永鹏; 孙运涛; 张同乔; 盛戈皞

###

DOI:10.3969/j.issn.1006-4729.2016.01.017

Journal of ShangHai University of Electric Power :2016,32(1):78-82

View/Add Comment Archive Advanced search HTML

←Previous | Next

Download full text

基于SOM神经网络的变电站设备红外热像诊断研究

王佳林¹, 崔昊杨¹, 许永鹏¹, 孙运涛², 张同乔², 盛戈皞³

(1.上海电力学院电子与信息工程学院;2.国网山东省电力公司济南供电公司;3.上海交通大学电气工程系)

Infrared Image Diagnosis Method of Transformer Substation Equipment Based on SOM Neural Network

WANG Jialin¹, CUI Haoyang¹, XU Yongpeng¹, SUN Yuntao², ZHANG Tongqiao², SHENG Gehao³

(1.School of Electronics and Information Engineering, Shanghai University of Electric Power, Shanghai 200090, China;2.Jinan Power Supply Company, Shandong Province Electric Power Company State Grid, Jinan 250000, China;3.Department of Electrical Engineering, Shanghai Jiao Tong University, Shanghai 200240, China)

Abstract

Image-text

References

Similar literature

本文已被：浏览 1538次下载 925次
Received:May 24, 2015

中文摘要: 提出了基于自组织神经网络(SOM)判别变电站设备热故障类型的红外图像诊断方法.采用了最大类间差法(OTSU)对电力设备红外热像进行了分割处理,从中提取出包括设备红外热像的温度特征值、Zernike不变矩等12个参数,以此作为设备状态识别的信息输入量,将设备的状态分类信息作为输出向量.通过训练56组红外热像数据,确定了SOM神经网络识别模型中的参数值.试验结果表明:该方法可用于变电站设备状态诊断,相对于传统的神经网络方法的诊断结果,该方法对设备运行状态评估的准确率高达85.7%,如将诊断模型产生的可疑状态列入故障状态,则故障的诊断率可达到95%以上.

中文关键词: 红外热像 SOM神经网络故障诊断 OTSU法

Abstract:A method of diagnosing substation equipment's state based on infrared image diagnosis method of SOM neural network is proposed by using the OTSU method.The equipment's infrared thermography is obtained through thermal infrared image,which could extract temperature characteristic value,Zernike invariant moment of infrared thermography. These values can be regarded as properties of distinguishing equipment state information. Then through treating classified information of equipment as the output vector and training 56 groups of the infrared image data,SOM neural network identification model is gained,which can be utilized on diagnosis of substation equipment. The experiment results show this method is highly accurate and its accuracy rate of diagnosis of running state is 85.7%,and if suspicious state of diagnosis model is treated as fault state,the fault rate of diagnosis can be above 95%.

keywords: infrared thermography SOM neural network Fault diagnosis OTSU method

文章编号：20160117 中图分类号： 文献标志码：

基金项目:

Reference text:

Author Name	Affiliation	E-mail
WANG Jialin	School of Electronics and Information Engineering, Shanghai University of Electric Power, Shanghai 200090, China
CUI Haoyang	School of Electronics and Information Engineering, Shanghai University of Electric Power, Shanghai 200090, China	cuihy@shiep.edu.cn
XU Yongpeng	School of Electronics and Information Engineering, Shanghai University of Electric Power, Shanghai 200090, China
SUN Yuntao	Jinan Power Supply Company, Shandong Province Electric Power Company State Grid, Jinan 250000, China
ZHANG Tongqiao	Jinan Power Supply Company, Shandong Province Electric Power Company State Grid, Jinan 250000, China
SHENG Gehao	Department of Electrical Engineering, Shanghai Jiao Tong University, Shanghai 200240, China