Deep Reinforcement Learning-Aided State-of-Health and State-of-Charge Management of Battery Energy Storage Systems

This paper proposes a deep reinforcement learning (DRL)-aided strategy for intelligently managing state-of-health (SoH) and state-of-charge (SoC) imbalances in a group of battery energy storage systems (BESSs) with varying SoH levels. The proposed method prioritizes SoH balancing during normal operation. However, when one or more BESSs approach full charge or discharge, it shifts priority to SoC balancing, as the aggregated system’s total power capability decreases unnecessarily if any BESS reaches its limits prematurely. A DRL agent is trained to determine when and how to switch between SoH and SoC balancing modes. Unlike state-of-the-art energy management methods for BESSs, whose performance heavily depends on predefined control parameters, the proposed strategy provides a more adaptive and robust solution.