Lithium-Ion Battery Management System with Reinforcement Learning for Balancing State of Charge and Cell Temperature

As an indispensable interface, a battery management system (BMS) is used to ensure the reliability of Lithium-Ion battery cells by monitoring and balancing the states of the battery cells, such as the state of charge (SOC). Since many battery cells are used in the form of packs, cell temperature imbalance may occur. Current approaches do not solve the multi-objective active balancing problem satisfyingly considering SOC and temperature. This paper presents an optimal control method using reinforcement learning (RL). The effectiveness of BMS based on Proximal Policy Optimization (PPO) agents obtained from hyperparameter optimization is validated in simulation narrowing the values to be balanced at least 28%, in some cases up to 72%. The RL agents let the active BMS select the optimal cell and regulates current for the balance of SOC and temperature between battery cells.