A deep reinforcement learning approach for state of charge and state of health estimation in lithium-ion batteries

Abstract

Accurate estimation of state variables such as State of Charge (SOC) and State of Health (SOH) is pivotal in the management of lithium-ion batteries. Existing methods, including the unscented Kalman filter (UKF), often require manual tuning of parameters and may not adapt well to the non-linear and non-stationary characteristics of batteries. This paper introduces a novel approach to optimize the parameters of an adaptive unscented Kalman filter (AUKF) using deep reinforcement learning (DRL). The DRL agent learns to adjust the parameters of the AUKF to maximize the estimation accuracy through interaction with the battery environment. This approach is capable of adapting to different battery types and operating conditions, eliminating the need for manual parameter tuning. Our results indicate that the DRL-optimized AUKF outperforms traditional UKF methods in terms of SOC and SOH estimation accuracy, demonstrating the potential of this approach for improving battery management systems.