基于高维随机矩阵的异常用电行为识别方法  

Recognition Method for Abnormal Electricity Consumption Behavior Based on High Dimensional Random Matrix

作　　者：王鹏 刘长江 刘攸坚 韦景康 邱凌 吴远超 李自怀 WANG Peng;LIU Changjiang;LIU Youjian;WEI Jingkang;QIU Ling;WU Yuanchao;LI Zihuai(Foshan Power Supply Bureau of Guangdong Power Grid Co.,Ltd.,Foshan,Guangdong 528000,China;Wuhan Xindian Electrical Co.,Ltd.,Wuhan,Hubei 430073,China)

机构地区：[1]广东电网有限责任公司佛山供电局,广东佛山528000 [2]武汉新电电气股份有限公司,湖北武汉430073

出　　处：《广东电力》2019年第6期78-85,共8页Guangdong Electric Power

基　　金：中国南方电网有限责任公司科技项目(GDKJQQ20161009)

摘　　要：基于高维随机矩阵理论,提出一种用户异常用电行为识别方法。该方法利用电网用户侧大数据构建高维随机矩阵,得出其矩阵特征值的谱分布和谱密度函数,通过M-P定律和单环定律判断样本数据有无异常;并通过平均谱半径确定用户用电异常时间区段。通过对某电网公司某台区近一年半的日用电量和有功功率数据进行分析,验证该方法的准确性和有效性。结果表明该方法能解决传统异常用电行为识别方法耗费人力大、时效性差及识别不精准等问题。Based on the high dimensional random matrix theory, this paper proposes a kind of recognition method for users abnormal electricity consumption behavior. By using big data of power grid user side, this method constructs a high dimensional random matrix and obtains spectral distribution and spectral density functions of matrix eigenvalues. According to M-P s law and single loop law, the method is used to decide whether sample data is abnormal or not. Meanwhile, it determines abnormal time section of electricity consumption according to mean spectral radius. On the basis of analyzing data of daily electricity consumption and reactive power in recent a year and half of a certain area of a power grid, correctness and effectiveness of this method is verified. The results indicate the method can solve problems of traditional recognition methods for abnormal electricity consumption behavior such as large manpower, poor timeliness, inaccurate recognition, and so on.

关 键 词：异常用电 窃电行为 非技术性损失 行为识别 高维随机矩阵 谱密度函数 大数据 

分 类 号：TM711.2[电气工程—电力系统及自动化]