Deep Reinforcement Learning-Based Resilient Resource Allocation for Smart Grid Network Slicing

With the growing number of users and increasing business volume, network resource allocation and fault recovery in the smart grid have become critical challenges. However, existing approaches do not address both business requirements and fault recovery in dynamically changing network environments, impacting user experience and business continuity. We propose a deep reinforcement learning-based resilient resource allocation algorithm DRL-CRO for smart grid network slicing. DRL-CRO incorporates a business transfer module to ensure service recovery after node failures, enhancing system reliability and user experience. Simulation results demonstrate that, compared to the baseline algorithm, DRL-CRO reduces operational costs, improves the revenue-cost ratio, boosts the overall reliability of the smart grid, and shows strong robustness under fluctuating network loads.