Deep Reinforcement Learning with Local Interpretability for Transparent Microgrid Resilience Energy Management

Abstract

Renewable energy integration into microgrids has become a key approach to addressing global energy issues such as climate change and resource scarcity. However, the variability of renewable sources and the rising occurrence of High Impact Low Probability (HILP) events require innovative strategies for reliable and resilient energy management. This study introduces a practical approach to managing microgrid resilience through Explainable Deep Reinforcement Learning (XDRL). It combines the Proximal Policy Optimization (PPO) algorithm for decision-making with the Local Interpretable Model-agnostic Explanations (LIME) method to improve the transparency of the actor network's decisions. A case study in Ongole, India, examines a microgrid with wind, solar, and battery components to validate the proposed approach. The microgrid is simulated under extreme weather conditions during the Layla cyclone. LIME is used to analyse scenarios, showing the impact of key factors such as renewable generation, state of charge, and load prioritization on decision-making. The results demonstrate a Resilience Index (RI) of 0.9736 and an estimated battery lifespan of 15.11 years. LIME analysis reveals the rationale behind the agent's actions in idle, charging, and discharging modes, with renewable generation identified as the most influential feature. This study shows the effectiveness of integrating advanced DRL algorithms with interpretable AI techniques to achieve reliable and transparent energy management in microgrids.