Distribution network fault recovery method and system based on robust graph reinforcement learning

By using a robust graph reinforcement learning-based dual-agent zero-sum game model and the Lagrange multiplier method, the real-time performance and robustness of power grid fault recovery methods in large-scale or dynamic scenarios are addressed, enabling rapid and stable fault recovery of power systems.

CN122118698APending Publication Date: 2026-05-29WUHAN UNIV OF TECH
View PDF 0 Cites 0 Cited by

Patent Information

Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
WUHAN UNIV OF TECH
Filing Date
2026-02-03
Publication Date
2026-05-29

AI Technical Summary

Technical Problem

Existing power distribution network fault recovery methods suffer from poor real-time performance and insufficient robustness in large-scale or dynamic scenarios, making it difficult to meet the real-time recovery requirements of power systems.

Method used

A robust graph reinforcement learning-based approach is adopted, which constructs a constrained Markov decision process through a two-agent zero-sum game model and the Lagrange multiplier method. The RS-AG-PPO algorithm is then used to train and solve the distribution network fault recovery problem, simulating uncertainties in the power grid and optimizing the recovery strategy.

Benefits of technology

It improves the real-time performance and robustness of distribution network fault recovery, maintains stable performance in complex environments, and enhances the power system's anti-interference capability.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN122118698A_ABST
    Figure CN122118698A_ABST
Patent Text Reader

Abstract

The application provides a distribution network fault recovery method and system based on robust graph reinforcement learning, which comprises the following steps: obtaining power flow characteristic data and network topology structure data of a distribution network, and establishing a distribution network fault recovery problem based on the power flow characteristic data and the network topology structure data of the distribution network; modeling the distribution network fault recovery problem as a constraint Markov decision process; constructing a robust graph reinforcement learning mathematical model of double-agent zero-sum game based on a Maximin strategy, setting a target function and a constraint condition, converting the robust graph reinforcement learning mathematical model of double-agent zero-sum game into an unconstrained problem by using a Lagrange multiplier method, and determining a target function of the unconstrained problem; and training and solving the robust graph reinforcement learning model of double-agent zero-sum game by using an RS-AG-PPO algorithm to obtain the distribution network fault recovery method. The application can formulate an efficient recovery strategy through an intelligent algorithm, greatly shortens the recovery time of a fault problem, and improves the reliability and operation efficiency of a system.
Need to check novelty before this filing date? Find Prior Art