基于人工免疫聚类算法的配电网故障状态相似性分析方法  

Fault State Similarity Analysis Method for Distribution Network Based on Artificial Immune Clustering Algorithm

作　　者：董文娜 王增平 赵乔 李钰洋 辛忠良 陈玉蛟 DONG Wenna;WANG Zengping;ZHAO Qiao;LI Yuyang;XIN Zhongliang;CHEN Yujiao(State Key Laboratory of Alternate Electrical Power System with Renewable Energy Sources(North China Electric Power University),Beijing 102206,China;Zhengzhou Power Supply Company,State Grid Henan Electric Power Company,Zhengzhou 450006,China)

机构地区：[1]新能源电力系统国家重点实验室(华北电力大学),北京102206 [2]国网河南省电力有限公司郑州供电分公司,郑州450006

出　　处：《电力系统及其自动化学报》2021年第6期60-66,共7页Proceedings of the CSU-EPSA

基　　金：国家电网公司总部科技资助项目(521710180008)。

摘　　要：为实现事故预案匹配,加快供电恢复策略生成速度,提出一种基于人工免疫聚类算法的配电网故障状态相似性分析方法。首先,将故障恢复策略相似作为相似故障状态的评价标准,构建配电网故障相似特征指标集,包括负荷失电量、联络线备用容量以及失电负荷-源点最短电气距离。在此基础上,鉴于聚类分析能够揭示无标记训练样本数据的内在关联特征,提出一种基于人工免疫聚类算法的故障相似状态匹配方法。最后,应用免疫系统信息处理机制对算法进行改进,通过引入免疫浓度机制、变异机制,弥补k-means聚类算法对初始中心点位置敏感的不足。算例分析表明,所提出的配电网故障状态相似性分析方法能够表征配电网故障样本间的内在关联性,且改进算法较传统算法聚类效果较优,为后续供电恢复策略的匹配生成提供了理论依据。To realize accident plan matching and accelerate the generation speed of power supply recovery strategy,a method of fault state similarity analysis for distribution network based on the artificial immune clustering algorithm is proposed.First,the similarity of fault recovery strategy is taken as the evaluation standard for similar fault states,and a set of indicators for the similar fault characteristics of distribution network is constructed,including the power loss of load,reserve capacity of tie line,and the shortest electrical distance between the power loss load and the source point.On this basis,considering that cluster analysis can reveal the intrinsic correlation characteristics of unlabeled training sample data,a fault similarity state matching method based on the artificial immune clustering algorithm is proposed.Finally,the immune system information processing mechanism is applied to improve the algorithm.By introducing the immune concentration mechanism and mutation mechanism,the k-means clustering algorithm is compensated for its lower sensitivity to the position of the initial center point.The analysis of an example shows that the proposed method of fault state similarity analysis for distribution network can characterize the intrinsic correlation between fault samples of distribution network.In addition,the clustering effect of the improved algorithm is better than that of the traditional method,providing a theoretical basis for the matching and generation of the subsequent power supply recovery strategy.

关 键 词：配电网故障状态相似性 供电恢复 K-MEANS聚类 人工免疫聚类 轮廓值系数 

分 类 号：TM727.2[电气工程—电力系统及自动化]