基于离群数据挖掘的低压窃电行为辨识方法研究  

Research on identification method of low⁃voltage electricity stealing behavior based on outlier data mining

作　　者：唐伟宁 刘颖 于旭 董冠良 TANG Weining;LIU Ying;YU Xu;DONG Guanliang(State Grid Jilin Electric Power Research Institute,Changchun 130021,China;School of Management Science and Information Engineering,Jilin University of Finance and Economics,Changchun 130117,China)

机构地区：[1]国网吉林省电力有限公司电力科学研究院,吉林长春130021 [2]吉林财经大学管理科学与信息工程学院,吉林长春130117

出　　处：《电子设计工程》2021年第23期56-59,64,共5页Electronic Design Engineering

基　　金：吉林省科技发展计划项目(20180101337JC);吉林省教育厅科研规划项目(JJKH20200139KJ)。

摘　　要：传统的低压窃电行为辨识方法的特征参数优化能力较弱,导致其辨识能力较差。为解决此问题,基于离群数据挖掘设计了一种新的低压窃电行为辨识方法。通过密度聚类方法分析不同方向的用电模式,在分析离群距离的基础上,挖掘不同的离群数据点,并设定评价矩阵,从而提取离群阈值。基于此分析窃电行为的可能性,进而完成对低压窃电行为的有效辨识。实验结果表明,该方法能够有效提高特征参数的优化能力,增强了对低压窃电行为的辨识效果。The characteristic parameter optimization ability of traditional identification methods for low voltage larceny behavior is weak,which leads to its poor identification ability.To solve this problem,a new identification method of low voltage power theft behavior is designed based on outlier data mining.The power consumption patterns in different directions are analyzed by density clustering method.On the basis of analyzing the outlier distance,different outlier data points are mined and evaluation matrix is set up to extract the outlier threshold.Based on this,the possibility of stealing electric power is analyzed,and then the effective identification of low-voltage stealing electric power is completed.Based on this,the possibility of stealing electric power is analyzed,and the effective identification of low-voltage stealing electric power is completed.The experimental results show that the proposed method can effectively improve the optimization ability of characteristic parameters and enhance the identification effect of low-voltage power theft.

关 键 词：离群数据挖掘 低压窃电 窃电行为辨识 密度聚类 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]