基于密度均值聚类的配电网线损异常点检测方法  

Distribution network line loss anomaly detection method based on density mean clustering

作　　者：马超 张铁刚 冯世杰 王保强 苏泽铨 MA Chao;ZHANG Tiegang;FENG Shijie;WANG Baoqiang;SU Zequan(Information and Communication Branch of Hainan Power Grid Co.,Ltd.,Haikou 570203,China;China Southern Power Grid Hainan Digital Power Grid Research Institute Co.,Ltd.,Haikou 570100,China;Baisha Power Supply Bureau of Hainan Power Grid Co.,Ltd.,Baisha Hainan 572800,China)

机构地区：[1]海南电网有限责任公司信息通信分公司,海口570203 [2]南方电网海南数字电网研究院有限公司,海口570100 [3]海南电网有限责任公司白沙供电局,海南白沙572800

出　　处：《自动化与仪器仪表》2023年第6期152-155,160,共5页Automation & Instrumentation

基　　金：海南电网有限责任公司科技项目(20198GF53001)。

摘　　要：为突出异常节点处的线损特征,使配电网主机能够准确控制电量信号传输行为,提出基于密度均值聚类的配电网线损异常点检测方法。根据密度均值中心取值结果,定义目标聚类函数,再通过计量线损贡献度数值的方式,完成基于密度均值聚类的配电网线损特征体系构建。按照隐含层节点分布方式,确定缺失数据对线损异常点的影响,联合已知线损率指标,得到数据恢复特征的准确计算结果,完成基于密度均值聚类的配电网线损异常点检测方法设计。实验结果:在密度均值聚类算法作用下,异常节点处的线损特征指标与给定线损特征指标在数值方面具有明显差异性,配电网主机可以快速分辨出待检测异常点所处位置,从而准确控制电量信号传输行为。In order to highlight the characteristics of line loss at abnormal nodes and enable the host of distribution network to ac-curately control the transmission behavior of electric quantity signal,a method of detecting abnormal points of line loss in distribution network based on density mean clustering was proposed.According to the value result of density mean center,the target clustering function is defined,and then the construction of distribution network line loss feature system based on density mean clustering is com-pleted by measuring the value of line loss contribution.According to the node distribution mode of hidden layer,the influence of miss-ing data on abnormal points of line loss was determined,and the accurate calculation results of data recovery characteristics were ob-tained by combining the known line loss rate index.The detection method of abnormal points of distribution network line loss based on density mean clustering was designed.Experimental results:Under the effect of density mean clustering algorithm,the characteristic index of line loss at abnormal nodes has obvious difference with the given line loss characteristic index in numerical aspect,and the distribution network host can quickly distinguish the location of abnormal points to be detected,so as to accurately control the electricquantity signal transmission behavior.

关 键 词：密度均值聚类 配电网线损 异常点检测 线损贡献度 隐含层 恢复特征 

分 类 号：TP274[自动化与计算机技术—检测技术与自动化装置] TM714[自动化与计算机技术—控制科学与工程]