基于多维行为分析的窃电高风险客户精准定位方法

Accurate Positioning Method of High-risk Customers of Electricity Theft Based on Multi-dimensional Behavior Analysis

关键词:  多维行为分析  窃电高风险客户  特征提取  数据去噪  精准定位

Key words:  multi-dimensional behavior analysis  high-risk customers of electricity theft  feature extraction  data noise  accurate positioning

摘要:

窃电行为对国家电力系统及供电公司造成了极大的损失，故反窃电技术是电力行业的重要研究方向之一。传统的窃电用户定位方法存在定位不准确、查处效率低等问题，为了解决上述问题，提出基于多维行为分析的窃电高风险客户精准定位方法。首先通过相关矩阵R及特征值谱熵正则化完成用户数据去噪，其次在UFS-MI模型内提取用户数据特征，分析用户用电的多维行为，最后根据逻辑回归算法完成窃电高风险客户的精确定位。实验结果表明，所提方法的窃电高风险客户定位精准度较高，误判率较低，整体定位效果较好。

Abstract:

Electricity theft has caused great losses to the national power system and power supply companies, so anti-electricity theft technology is one of the important research directions in the power industry. The traditional positioning method of electricity theft users has the problems of inaccurate positioning and low efficiency of investigation and punishment. In order to solve the above problems, an accurate positioning method for high-risk customers of electricity theft based on multi-dimensional behavior analysis is proposed. Firstly, the user data denoising is completed by the correlation matrix R and the eigenvalue spectral entropy regularization. Secondly, the data characteristics of electricity users are extracted in the UFS-MI model, and the multi-dimensional behavior of electricity users' consumption is analyzed. Finally, according to the logistic regression algorithm, the precise positioning of high-risk customers for electricity theft is completed. The experimental results show that the proposed method has high positioning accuracy for high-risk customers of electricity theft, low misjudgment rate and good overall positioning effect.