基于K两端聚类的污秽绝缘子放电状态评估  

Discharge Condition Assessment of Polluted Insulator Based on K-two-end Clustering

作　　者：杜念[1] 黄建才[1] 朱永利[1] 

机构地区：[1]华北电力大学控制与计算机工程学院,河北保定071003

出　　处：《电瓷避雷器》2011年第4期23-28,共6页Insulators and Surge Arresters

基　　金：国家自然科学基金项目(编号:60974125)

摘　　要：沿面放电是绝缘子污秽程度及运行状态的重要体现,实时掌握绝缘子的污秽放电状态对污闪的预防具有重大意义。根据污秽绝缘子在放电过程中泄漏电流的波形变化以及放电区段的划分特点,提出了一种基于K两端聚类的动态划分方法。分析污秽绝缘子在运行中受潮、形成干燥带及局部电弧、局部电弧发展至极端闪络的全过程,并依据绝缘子运行中工作电压和泄漏电流的时频特征,提取了能够表征绝缘子放电状态的特征矢量。运用K两端聚类算法对三组同型号绝缘子的放电过程进行了区段划分。并用SVM(支持向量机)对聚类后的数据进行训练,得到SVM分类模型以用于绝缘子的运行状态评估。Discharge along the surface is the important reflection of pollution level and operating condition of insulator. Real-time monitoring pollution discharge condition plays an important role in the prevention of flashover accident. In accordance with leakage current waveforms variation during the polluted insulator discharge process and the division features of the discharge section, a dynamic division method based on K-two-end clustering is presented. The whole course of polluted insulator in service is analyzed, which including wetting, producing dry zone and local arc, and finally Local arc development to extreme flashover. And according to operation voltage and time-frequency characteristics of leakage current of insulator in service, the characteristic vector to symbolize discharge condition is obtained. K-two-end clustering is used for discharge processes division of three insulators with the same type number. And then support vector machine is used for the training of discharge data clustered, a classification model is determined, which could be used for the assessment on operation condition of insulator.

关 键 词：绝缘子 泄漏电流 聚类 状态评估 污秽放电区段 支持向量机 

分 类 号：TM216[一般工业技术—材料科学与工程]