Electricity stealing behavior detection method based on improved multi-agent artificial immune network

By improving the multi-agent artificial immune network and utilizing DTW distance and electrical physical constraints, the problems of timing mismatch and sample bias in electricity theft detection were solved, achieving high-precision electricity theft identification with low false alarms and improving detection efficiency.

CN122241173APending Publication Date: 2026-06-19HUIAN COUNTY POWER SUPPLY CO OF STATE GRID FUJIAN ELECTRIC POWER CO LTD +1
View PDF 0 Cites 0 Cited by

Patent Information

Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
HUIAN COUNTY POWER SUPPLY CO OF STATE GRID FUJIAN ELECTRIC POWER CO LTD
Filing Date
2026-02-28
Publication Date
2026-06-19

AI Technical Summary

Technical Problem

Existing technologies for detecting electricity theft suffer from problems such as timing mismatch, amplified sample bias, and physical infeasibility, making it difficult to effectively identify high-dimensional sparse and time-abrupt electricity theft.

Method used

An improved multi-agent artificial immune network is adopted, which calculates affinity through dynamic time warping (DTW) distance, embeds an imbalance learning strategy and power physics constraints, and forms a detection method with time-series perception, imbalance perception and physical perception to ensure the model accurately identifies electricity theft patterns.

Benefits of technology

It significantly improved the detection rate of electricity theft, controlled the false alarm rate, enhanced the interpretability and reliability of the model, and met the power grid inspection standards.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN122241173A_ABST
    Figure CN122241173A_ABST
Patent Text Reader

Abstract

This invention proposes a method for detecting electricity theft based on an improved multi-agent artificial immune network (AIN). It employs an AIN-based algorithm to map the electricity theft detection task into an antigen-antibody identification problem. The algorithm targets the data characteristics of electricity theft, using the electricity consumption curve of the user to be tested as the antigen and known user templates as the antibody. It integrates electricity consumption time-series features through a neighborhood cloning selection mechanism to capture non-rigid time-series distortions caused by electricity theft; embeds an imbalanced learning strategy in the antibody collaboration mechanism to alleviate model bias caused by the sparsity of negative samples; and combines competitive operations with power physics constraints to ensure that the generated memory cells conform to the physical laws of the power grid. This invention effectively solves the problems of time-series mismatch, amplified sample bias, and physical infeasibility in existing technologies by reconstructing the algorithm for the three core aspects of electricity theft scenarios.
Need to check novelty before this filing date? Find Prior Art