Deep Reinforcement Learning for Resilient Predictive Control in Self-Healing Smart Grids

The transition to self-healing smart grids demands advanced control strategies to ensure resilience against faults, disturbances, and cyber-physical threats. This paper proposes a deep reinforcement learning (DRL)-based resilient predictive control (DRL-RPC) framework, enabling autonomous fault recovery, voltage stabilization, and adaptability to dynamic grid conditions. Using a Proximal Policy Optimization (PPO) algorithm, the DRL agent is trained on a large-scale dataset exceeding ۱۵,۰۰۰ samples, capturing realistic grid dynamics. Extensive simulations demonstrate that our approach outperforms traditional methods, achieving faster recovery, improved voltage stability, and reduced load shedding. The framework’s scalability and robustness position it as a transformative solution for next-generation smart grids.

کلیدواژه ها:

Deep Reinforcement Learning ، Smart Grids ، Predictive Control ، Resilience ، Self-Healing