基于多元模糊支持向量数据描述的高压电缆缺陷识别  

Defect Recognition of High Voltage Cable Based on Multi-Fuzzy Support Vector Data Description

作　　者：刘敏 方义治 孙廷玺 何伟 罗思琴 兰雪珂 周念成[2] LIU Min;FANG Yi-zhi;SUN Ting-xi;HE Wei;LUO Si-qin;LAN Xue-ke;ZHOU Nian-cheng(Zhuhai Power Supply Bureau of Guangdong Power Grid Co.Ltd.,Zhuhai Guangdong 519000;State Key Laboratory of Power Transmission Equipment&System Security and New Technology,Chongqing University,Shapingba Chongqing 400044)

机构地区：[1]广东电网有限责任公司珠海供电局,广东珠海519000 [2]重庆大学输配电装备及系统安全与新技术国家重点实验室,重庆沙坪坝400044

出　　处：《电子科技大学学报》2020年第2期240-247,共8页Journal of University of Electronic Science and Technology of China

基　　金：国家自然科学基金(51577018);广东电网有限责任公司科技项目(GDKJXM20162022)。

摘　　要：为了实现高压电缆缺陷状态准确识别,该文提出基于多元模糊支持向量描述的高压电缆缺陷识别方法。针对原始支持向量描述(SVDD)对异常数据敏感的问题,首先利用竞争凝聚算法对高压电缆训练样本进行模糊聚类和筛选,并利用筛选后样本的隶属度对SVDD加权处理,进而获得高压电缆每种状态的训练样本的最小超球体,实现对高压电缆多种缺陷状态的有效识别。最后,利用广东珠海高压电缆的实际数据对所提模型的可行性和准确性进行验证。In order to realize the accurate defect recognition of high voltage,this paper proposes defect recognition of high voltage cable based on the multi-fuzzy support vector data description.The original support vector description(SVDD)is sensitive to abnormal data,competitive aggregation algorithm is used to perform fuzzy clustering and screening for training samples,and SVDD is weighted by the selected samples.Then the minimum hypersphere of training sample of each state of the high voltage cable is obtained,and effective identification of various defect states of the high voltage cable is also realized.Finally,the feasibility and accuracy of the proposed model are verified by the actual data of high voltage cable in Zhuhai,Guangdong province.

关 键 词：竞争凝聚算法 缺陷识别 超球体 高压电缆 多元模糊支持向量描述 

分 类 号：TM76[电气工程—电力系统及自动化]