A Resilience Enhancement Method for Multi-Source Coordinated Restoration Based on Graph Reinforcement Learning

View at publisher Summarise with VIRAT
Authors: Zhang Xue, Xiaolu Li, Baokun Yuan, Hongzhong Li, Jinsong Liu
Cited by: 0
References: 4

Summary

No summary yet. VIRAT can read this paper and write a plain-language summary of its aim, methods, key findings and limitations.

Summarise with VIRAT Free to try with your Science Catalogs account.

Abstract

Extreme weather events have posed growing threats to the resilience of urban distribution networks, highlighting the urgent need for fast and reliable restoration strategies. This paper proposes a resilience enhancement method for multi-source coordinated restoration based on graph reinforcement learning (GRL). The restoration problem is formulated as a Markov decision process (MDP), and a learning framework that combines Graph Attention Networks (GAT) with Double Deep Q-Networks (DDQN) is used to train effective restoration policies. To further improve feasibility and accelerate convergence, expert knowledge pre-training mechanisms are incorporated. Simulation results on the improved IEEE 33-bus system demonstrate that the proposed method yields high-quality restoration plans with low constraint violations under extreme disaster scenarios, providing an effective and practical solution for resilient distribution system restoration.

Fields of Study & Subjects

Optimal Power Flow Distribution Power System Optimization and Stability Infrastructure Resilience and Vulnerability Analysis

Publication details

Journal / source
IEEE PES International Meeting (PES IM)
Pages
1–5
Publication date
18 January 2026
Language
EN