A Hybrid-Driven Method of Knowledge Guidance and Unsupervised Learning for Power Users Anomaly Discrimination

doi:10.3969/j.issn.1008-0198.2025.04.011

Abstract

Abstract: In view of the problems of high manual annotation cost and limited data feature representation ability of traditional rule-based detection methods, a hybrid model for anomaly detection of power users that integrates knowledge guidance and unsupervised learning is proposed. First, a multi-dimensional load characteristic index system containing load rate, peak-to-valley difference rate and other features is constructed based on power consumption behavior analysis, and principal component analysis is used to realize feature extraction and visualization of high-dimensional data. Then, the abnormal deviation degree of users' power consumption modes and abnormal probability of spatial distribution are quantified by reconstruction error and local outlier factor algorithm respectively. Secondly, the abnormality degree and suspected probability of users' power consumption behavior are ranked, so that most abnormal users can be found by detecting only a few users with high abnormality degree. Finally, the proposed method is verified by small and large data sets respectively in the case analysis. The results show that the AUC score of this method can reach 0.8251, which has significant superiority. In addition, this method can not only effectively identify abnormal users, but also reduce the false alarm rate and improve the accuracy of detection, providing an efficient and reliable power consumption anomaly analysis tool for power companies.

Key words: anti theft techniques, anomaly discrimination, unsupervised learning, principal component analysis, local matrix reconstruction

CLC Number:

TM714

LYU Weijia, LI Xiaohui, TENG Yongxing, CHEN Juan, MEN Yingjun, DU Tianshuo. A Hybrid-Driven Method of Knowledge Guidance and Unsupervised Learning for Power Users Anomaly Discrimination[J]. Hunan Electric Power, 2025, 45(4): 74-82.

References

[1] 张利生. 电力网电能损耗管理及降损技术:2版[M]. 北京:中国电力出版社,2008.
[2] 林振智,崔雪原,金伟超,等. 用户侧窃电检测关键技术[J]. 电力系统自动化,2022,46(5):188-199.
[3] PUNMIYA R,CHOE S.Energy theft detection using gradient boosting theft detector with feature engineering-based prepro-cessing[J]. IEEE Transactions on Smart Grid,2019,10(2):2326-2329.
[4] RAZAVI R,GHARIPOUR A,FLEURY M,et al.A practical feature-engineering framework for electricity theft detection in smart grids[J]. Applied Energy,2019,238:481-494.
[5] 游文霞,申坤,杨楠,等. 基于Bagging异质集成学习的窃电检测[J]. 电力系统自动化,2021,45(2):105-113.
[6] MONEDERO I,BISCARRI F,LEÓN C,et al. Detection of frauds and other non-technical losses in a power utility using Pearson coefficient,Bayesian networks and decision trees[J]. International Journal of Electrical Power & Energy Systems,2012,34(1):90-98.
[7] YAP K S,TIONG S K,NAGI J,et al.Comparison of supervised learning techniques for non-technical loss detection in power utili-ty[J].International Review on Computers and Software,2012,7(2):626-636.
[8] NAGI J,YAP K S,TIONG S K,et al.Nontechnical loss detection for metered customers in power utility using support vector machines[J]. IEEE Transactions on Power Delivery,2010,25(2):1162-1171.
[9] JOKAR P,ARIANPOO N,LEUNG V C M. Electricity theft detection in AMI using customers′ consumption patterns[J]. IEEE Transactions on Smart Grid,2016,7(1):216-226.
[10] 刘增明. 供电企业防窃电方法和对策的研究[D]. 北京:华北电力大学,2013.
[11] 张承智,肖先勇,郑子萱. 基于实值深度置信网络的用户侧窃电行为检测[J]. 电网技术,2019,43(3):1083-1091.
[12] 杨挺,徐嘉成,叶芷杉,等. 基于最优变换关联因子和优化平移分裂法的窃电检测[J]. 电网技术,2023,47(9):3913-3924.
[13] NAGI J,YAP K S,TIONG S K,et al.Improving SVM-based nontechnical loss detection in power utility using the fuzzy inference system[J].IEEE Transactions on Power Delivery,2011,26(2):1284-1285.
[14] ZANETTI M,JAMHOUR E,PELLENZ M,et al.A tunable fraud detection system for advanced metering infrastructure using short-lived patterns[J]. IEEE Transactions on Smart grid,2019,10(1):830-840.
[15] 胡天宇,郭庆来,孙宏斌. 基于堆叠去相关自编码器和支持向量机的窃电检测[J]. 电力系统自动化,2019,43(1):119-125.
[16] ANGELOS E W S,SAAVEDRA O R,CORTÉS O A C,et al. Detection and identification of abnormalities in customer consumptions in power distribution systems[J]. IEEE Trans actions Power Delivery,2011,26(4):2436-2442.
[17] PASSOS L A JR,OBA RAMOS C C,RODRIGUES D,et al. Unsupervised non-technical losses identification through optimum-path forest[J]. Electric Power Systems Research,2016,140:413-423.
[18] ZHENG K D,WANG Y,CHEN Q X,et al.Electricity theft detecting based on density-clustering method[C]//2017 IEEE Innovative Smart Grid Technologies-Asia(ZISGT-Asia). Auckland,Newzealand,IEEE,2017:1-6.
[19] NIZAR A H,DONG Z Y,WANG Y.Power utility nontechnical loss analysis with extreme learning machine method[J]. IEEE Transactions on Power Systems,2008,23(3):946-955.
[20] 王天真,刘远,汤天浩,等. 基于相对主元分析的动态数据窗口故障检测方法[J]. 电工技术学报,2013,28(1):142-148.
[21] LEE Y J,YEH Y R,WANG Y C F. Anomaly detection via online oversampling principal component analysis[J]. IEEE Transactions on Knowledge and Data Engineering,2012,25(7):1460-1470.
[22] 王建元,张少锋. 基于线性判别分析和密度峰值聚类的异常用电模式检测[J]. 电力系统自动化,2022,46(5):87-98.
[23] 刘敏,方义治,孙廷玺,等. 基于邻域保持嵌入-主成分分析的高压电缆状态数据异常检测及分析[J]. 科学技术与工程,2019,19(27):192-199.
[24] SU H J,WU Z Y,DU Q,et al.Hyperspectral anomaly detection using collaborative representation with outlier removal[J]. IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing,2018,11(12):5029-5038.