基于BP神经网络的复合绝缘子憎水性等级的判定  

Determination to Hydrophobic Level of Composite Insulators Based on BP Neural Network

作　　者：汪佛池[1] 闫康[1] 张重远[1] 石鹏[2] 律方成[1] 

机构地区：[1]河北省输变电设备安全防御重点实验室(华北电力大学),保定071003 [2]保定供电公司,保定071000

出　　处：《高压电器》2013年第12期19-25,共7页High Voltage Apparatus

基　　金：中央高校基本科研业务费专项资金资助项目(13MS71)~~

摘　　要：憎水性检测对于确保复合绝缘子安全可靠运行具有重要意义。笔者提出把图像处理技术和BP神经网络引入到绝缘子憎水性检测中。首先,运用对比度受限自适应直方图均衡和数学形态学滤波对憎水性图像进行增强。然后,利用自适应阈值对图像进行分割,并提取图像中与憎水性相关的4个特征量。最后,选择BP神经网络判定绝缘子憎水性等级。分别采用BP标准算法和4种改进算法对网络进行训练,并对测试样本进行了憎水性等级判定。基于4个特征量的BP网络在一定程度上能够准确地判定绝缘子的憎水性等级。各种算法的判定结果表明L-M算法是比较合理的判定绝缘子憎水性等级的BP神经网络算法。Hydrophobic measurement is of great significance to ensure safe and reliable operation of composite insulators. In this paper, the image processing technology and BP neural network are introduced to the hydrophobic measurement of insulators. Firstly, the contrast limited adaptive histogram equalization and mathematical morphology filter are used to enhance the hydrophobic image. Then the image was segmented by adaptive threshold, and four features associated with hydrophobic are extracted in the image. Finally, the BP neural network is selected to determine hydrophobic level of insulators. The standard and four improved algorithms of BP are used to train network, and then the trained network is used to determine the hydrophobic level of the test sample. It is concluded that BP network based on four features could determine the hydrophobic level of insulators to a certain extent. Moreover, it is suggested the L-M algorithm is the most reasonable algorithm for determining the hydrophobic level of insulators according to determination results of different algorithms.

关 键 词：憎水性检测 BP神经网络 对比度受限自适应直方图 数学形态学滤波 自适应阈值 ￡埘算法 

分 类 号：TM216[一般工业技术—材料科学与工程]