Hierarchical deep Q-network-based optimization of resilient grids under multi-dimensional uncertainties from extreme weather.

The increasing frequency of extreme weather events, coupled with the rising adoption of renewable energy, has exposed critical vulnerabilities in modern power systems. Addressing these challenges requires a holistic approach to enhance resilience, operational flexibility, and cybersecurity. This paper introduces an innovative optimization framework leveraging Deep Q-Networks (DQN) to integrate renewable energy sources, energy storage systems, and advanced resilience metrics into a cohesive decision-making model. By dynamically addressing uncertainties arising from weather-induced disruptions and demand variability, the framework provides a robust solution for real-time system adaptation. Through extensive case studies based on a realistic 118-node distribution network, the proposed model achieves significant performance improvements, including up to 35% reduction in load shedding, 30% enhancement in resilience scores, and €15 million in additional cost savings compared to conventional methods. The results underscore the framework's ability to balance economic efficiency and system reliability while ensuring scalability across diverse operational scenarios. This research paves the way for next-generation power systems capable of withstanding disruptions, setting a new benchmark for sustainable and resilient energy management.