Deep Reinforcement Learning–Based Cascading Failure Mitigation in New Power Systems with Wind Power Penetration
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
Cascading failure in power systems constitute the primary cause of large-scale blackouts, frequently inflicting substantial socio-economic losses. With the integration of high proportions of renewable energy sources, the inherent randomness and volatility of generation have markedly heightened operational uncertainties within power grids. This renders traditional rule-based or model-based emergency control strategies ill-suited to adapting to diverse fault scenarios. Addressing this challenge, this paper proposes a chain fault interruption method grounded in deep reinforcement learning. First, a cascading failure simulation model incorporating primary frequency regulation, islanding handling, and probabilistic line tripping mechanisms was established. Weibull distributions were employed to model the randomness of wind power output, simulating the conditions of modern power systems. Subsequently, the Proximal Policy Optimisation (PPO) algorithm was used to train the agent, enabling it to generate optimal line-disconnection strategies in real-time during the dynamic evolution of wind power fluctuations and fault propagation. Case studies on an enhanced IEEE 39-node system demonstrate that the proposed method effectively blocks fault propagation under varying wind power penetration levels and initial fault scenarios. Compared to traditional inactive and random strategies, this approach significantly improves system retention rates, exhibiting robust performance and strong generalisation capabilities.
Fields of Study & Subjects
Power System Optimization and Stability
Power System Reliability and Maintenance
Electric Power System Optimization
Publication details
- Journal / source
- International Conference on Electrical Automation and Artificial Intelligence (ICEAAI)
- Pages
- 107–112
- Publication date
- 16 January 2026
- Language
- EN