Resilience Optimization and Dynamic Stability Defense in Active Distribution Networks Under Extreme Disasters: A Graph Learning and Cooperative Control Approach

Chutao Zheng¹Wei Li^2*Xinsen Yang¹Diwei Lin²Hao Bai¹Guowei Guo²

• ¹Guangdong Power Grid Corporation Foshan Power Supply Bureau, foshan, China
• ²Electric Power Research Institute of China Southern Power Grid, guangzhou, China

The escalating integration of high-penetration renewable energy sources introduces severe dynamic stability challenges-such as low inertia and fast transients-to modern power systems, particularly in the context of Active Distribution Networks (ADNs). These vulnerabilities are critically amplified when the system is subjected to extreme natural disasters, potentially leading to widespread instability and destructive cascading failures. To address this pressing need for robust operation, this paper pro-poses a novel framework for Resilience Optimization and Dynamic Stability Defense in ADNs, utilizing a Graph Learning and Cooperative Control Approach. The core of the methodology is a topology-aware stability assessment integrated with multi-objective, risk-informed decision-making. We first employ a GraphSAGE-based predictor to establish a high-fidelity, nonlinear mapping between the system's uncertain operating states and its dynamic stability margin. This data-driven approach over-comes the computational limitations of conventional physical models in real-time stability prediction and precursor identification following an extreme event. Furthermore, a lightweight cooperative control mechanism is designed to maximize operational resilience under catastrophic conditions. This mechanism coordinates the immediate dynamic stability defense with long-term system recovery by optimizing generation dispatch, power flow, and stability margins against cascading failure propagation. By coupling offline strategic planning with online real-time optimization, the framework ensures adaptive and rapid system response. Case studies confirm that the proposed framework drastically enhances the system's ability to resist instability and mitigates cascading failure sequences under extreme stress. It provides an efficient, accurate, and real-time solution for robust stability management and resilience optimization in high-renewable active distribution networks.