利用红外热像检测高压绝缘子污秽度的影响因素研究  

Influence Factors Research of Insulator Contamination Grades Detection Based on Infrared Image

作　　者：何洪英[1] 杨迎建[2] 姚建刚[1] 蒋正龙[3] 钱艳萍[1] 

机构地区：[1]湖南大学电气与信息工程学院,长沙410082 [2]国网电力科学研究院,武汉430074 [3]湖南省电力试验研究院,长沙410007

出　　处：《高电压技术》2010年第7期1730-1736,共7页High Voltage Engineering

基　　金：湖南省科技计划(2006GK3043);国网武汉高压研究院项目(2007)~~

摘　　要：针对不同环境温度、环境湿度、实验电压下绝缘子污秽程度的识别,提出了一种基于红外热像特征和BP神经网络的智能识别方法。以污秽绝缘子表面温度分布的方差值及其相对于环境平均温度的最大温升和平均温升作为污秽特征,结合环境温度、湿度、实验电压共6个量作为BP神经网络输入,通过神经网络强大的非线性映射能力实现污秽等级的自动识别。实验分析表明污秽特征的选取剔除了环境温度的影响,在不同电压等级与湿度条件下,不同污秽等级特征数据虽有部分重合,但神经网络引入电压等级与湿度作为多维输入,因此同样可以准确区分不同污秽等级。实验结果证明该方法能有效实现不同环境温度、环境湿度、实验电压下的绝缘子污秽等级识别。Aiming at the contamination grades recognition of the insulators under different circumstance temperature,different circumstance humidity and different loading voltage,a method integrating infrared image features of contaminated insulators and neural network was presented. The temperature variance of the contaminated insulator surface,the highest,the mean temperature raise of the insulator surface comparing to the mean temperature of the circumstance were taken as the contamination features extracted from the infrared image,and these three features integrating circumstance temperature,circumstance humidity and loading voltage were all used as six inputs of the BP neural network. Then,the automatic recognition of the contamination grades was achieved by the great nonlinear mapping ability of the neural network. Analysis indicates that the selection of the contamination features eliminates the influences of the circumstance temperature by subtracting the mean circumstance temperature. Though contamination feature data under different circumstance humidity and different loading voltage may partally overlap,contamination grades recognition are accurately achieved by using loading voltage and circumstance humidity as inputs of neural network. Experimental results prove that the method is feasible and efficient for the insulator contamination grades recognition under different circumstance temperature,different circumstance humidity and different loading voltage.

关 键 词：绝缘子污秽等级识别 环境温度 环境湿度 实验电压 BP神经网络 污秽绝缘子红外热像特征 

分 类 号：TM216[一般工业技术—材料科学与工程]