基于支持向量机的电网设备绝缘材料缺陷太赫兹成像方法  

Terahertz Imaging Method of Insulation Material Defects in Power Grid Equipment Based on Support Vector Machine Algorithm

作　　者：王磊 梅红伟 刘建军 陈大兵 江怀远 王黎明 WANG Lei;MEI Hongwei;LIU Jianjun;CHEN Dabing;JIANG Huaiyuan;WANG Liming(Tsinghua Shenzhen International Graduate School,Tsinghua University,Guangdong Shenzhen 518055,China;State Grid Jiangsu Electric Power Co.,Ltd.,Electric Power Research Institute,Nanjing 211103,China)

机构地区：[1]清华大学深圳国际研究生院,广东深圳518055 [2]国网江苏省电力有限公司电力科学研究院,南京211103

出　　处：《高压电器》2025年第3期160-167,共8页High Voltage Apparatus

基　　金：国家电网公司科技项目(5500-202118134A-0-0-00)。

摘　　要：电网设备绝缘材料在生产、运输、施工、现场运行等过程中容易产生缺陷,这严重影响了电网的安全运行。为实现对绝缘材料缺陷的无损检测,文中提出了基于支持向量机(SVM)的太赫兹二维成像方法。制备了聚乙烯(PE)、聚氯乙烯(PVC)和硅橡胶3种带缺陷的典型电力绝缘材料平板样品进行测试,得到缺陷识别准确率的不低于92.3%。在成像方面,与传统的单一特征参数成像方法相比,SVM成像方法对缺陷区域与非缺陷区域具有更好的区分度,边缘过渡区域明显变窄,成像质量较高。Insulation materials of power grid equipment are prone to defects in the process of production,transportation,construction and operation,which seriously affects safe operation of power grid.In order to realize the nondestructive detection of insulation material defects,a terahertz 2D imaging method based on support vector machine(SVM)is proposed in this paper.Such Three kinds of insulation materials samples as PE,PVC and silicone rubber are prepared for testing,and the accuracy of defect identification is not less than 92.3%.In terms of imaging,the SVM imaging method,compared with the traditional feature parameter imaging method,has a better distinction between defect and non-defect region,the edge transition region is obviously narrower and the imaging quality is higher.

关 键 词：绝缘材料 太赫兹成像 支持向量机 缺陷识别 

分 类 号：TM21[一般工业技术—材料科学与工程] O441.4[电气工程—电工理论与新技术] TP391.41[理学—电磁学]