Education, Science, Technology, Innovation and Life
Open Access
Sign In
  1. Home
  2. Journals
  3. Automation and Machine Learning
  4. Vol 7, Issue 2, 2026

Deep Reinforcement Learning-Based Approach for Abnormal Line Loss Diagnosis and Electricity Theft Detection in Distribution Transformer Areas

Download as PDF

DOI: 10.23977/autml.2026.070208 | Downloads: 0 | Views: 104

Author(s)

Keqiang Chen 1, Wanqing Zhong 1

Affiliation(s)

1 State Grid Gansu Electric Power Company Lanzhou Power Supply Company, Lanzhou, Gansu, China

Corresponding Author

Keqiang Chen

ABSTRACT

When abnormal line loss shows up in a low-voltage transformer area, it can often be traced to non-technical losses like electricity theft. Traditional approaches to tracking down abnormal line loss have leaned heavily on manual inspections and rule-based expert systems, but they eat up a lot of time and money, and their efficiency drops sharply once the number of users grows. The transformer area is modeled as a partially observable environment in which an agent has to decide which user to check next, drawing on total line loss information and the electricity consumption records gathered from smart meters. A dueling deep Q-network (DQN) with action masking is used to learn a better inspection policy, and the goal is to recover more of the lost revenue while keeping the inspection cost down. Experiments were run on a synthetic dataset containing 100 users, and the proposed method ended up needing far fewer inspections than a greedy baseline—a 43.2% reduction in the average number of inspections was achieved, while a theft detection rate of 97.8% was still reached. The method also performed better than traditional rule-based methods and supervised anomaly detection models. These results show that the learned inspection strategy is able to make use of the connection between abnormal line loss and user electricity behavior, so theft can be located more accurately and at a lower cost.

KEYWORDS

Deep reinforcement learning, abnormal line loss diagnosis, electricity theft detection, distribution station area, smart meter, sequential decision making

CITE THIS PAPER

Keqiang Chen, Wanqing Zhong. Deep Reinforcement Learning-Based Approach for Abnormal Line Loss Diagnosis and Electricity Theft Detection in Distribution Transformer Areas. Automation and Machine Learning (2026). Vol. 7, No. 2, 68-75. DOI: http://dx.doi.org/10.23977/autml.2026.070208.

REFERENCES

[1] A. H. Nizar, Z. Y. Dong, and Y. Wang, "Power utility nontechnical loss analysis with extreme learning machine method," IEEE Trans. Power Syst., vol. 23, no. 3, pp. 946–955, 2008.
[2] P. Jokar, N. Arianpoo, and V. C. M. Leung, "Electricity theft detection in AMI using customers' consumption patterns," IEEE Trans. Smart Grid, vol. 7, no. 1, pp. 216–226, 2016.
[3] Z. Zheng, Y. Yang, X. Niu, H.-N. Dai, and Y. Zhou, "Wide and deep convolutional neural networks for electricity-theft detection to secure smart grids," IEEE Trans. Ind. Informat., vol. 14, no. 4, pp. 1606–1615, 2018.
[4] V. Mnih et al., "Human-level control through deep reinforcement learning," Nature, vol. 518, pp. 529–533, 2015.
[5] H. Van Hasselt, A. Guez, and D. Silver, “Deep reinforcement learning with double Q-learning," in Proc. AAAI, 2016, pp. 2094–2100.
[6] Z. Wang, T. Schaul, M. Hessel, H. Van Hasselt, M. Lanctot, and N. De Freitas, “Dueling network architectures for deep reinforcement learning," in Proc. ICML, 2016, pp. 1995–2003.
[7] T. Schaul, J. Quan, I. Antonoglou, and D. Silver, "Prioritized experience replay," in Proc. ICLR, 2016.
[8] S. Huang, Y. Wu, Y. Chi, and Y. Li, "Line loss rate prediction method based on deep learning with long short-term memory," Power Syst. Technol., vol. 43, no. 4, pp. 1125–1132, 2019.
[9] J. B. E. A. R. S. Sutton and A. G. Barto, Reinforcement Learning: An Introduction, 2nd ed. Cambridge, MA: MIT Press, 2018.
[10] R. S. Niculescu, T. M. Mitchell, and R. B. Rao, "Bayesian network learning with a deep exploration strategy for sequential diagnosis," in Proc. AAAI, 2007, pp. 605–610.

Subscription
E-Mail Alert
Most popular papers
Call for Special Issue
Publication Ethics & OA Statement
Plagiarism Policy
Contact Us
Downloads: 5602
Visits: 307411

Sponsors, Associates, and Links

All published work is licensed under a Creative Commons Attribution 4.0 International License.

Copyright © 2016 - 2031 Clausius Scientific Press Inc. All Rights Reserved.