Reinforcement Learning for Simultaneous SOC and Temperature Balancing of Li-ion Battery Cells in Battery Management Systems

The increasing adoption of electric vehicles (EVs) and battery energy storage systems (BESS) makes battery management systems (BMS) crucial for effectively managing battery cells, such as state of charge (SOC) and cell temperature. Significant imbalances in these parameters can degrade battery performance and reduce reliability. Although active balancing strategies offer greater control flexibility compared to passive methods, they typically face challenges in implementing complex control algorithms, particularly for multi-objective balancing tasks. To overcome these challenges, this paper proposes an enhanced active BMS employing reinforcement learning (RL) to simultaneously minimize deviations in SOC and cell temperature. Specifically, a proximal policy optimization (PPO)-based RL agent is trained within a simulation environment and subsequently deployed in a real-world experimental setup. Experimental results demonstrate that the proposed approach can quickly and effectively balance both SOC and cell temperature while efficiently managing the auxiliary power source (APS). This study validates the feasibility of intelligent multi-objective battery state balancing with RL.