State of charge estimation of lithium-ion battery based on double deep Q network and extended Kalman filter

Abstract

The state of charge (SOC), as an important parameter of the lithium-ion battery management system (BMS), is an important factor affecting the current battery detection and safety issues. Due to the non-linear characteristics of the lithium-ion battery charging and discharging process, it is difficult to model the BMS and estimate the SOC accurately. This paper takes lithium-ion battery as the research object, designs the Markov decision process (MDP) model of lithium-ion battery energy storage prediction, establishes the second-order RC equivalent circuit model (ECM) and extended Kalman filter (EKF), and proposes a method based on double deep Q network (double DQN) to optimize the EKF parameters SOC estimation method to solve the problem of dimensionality disaster and value function overestimation generated in deep reinforcement learning (DRL). Through the effective verification of the lithium-ion battery SOC estimation algorithm, the simulation results show that compared with the traditional deep Q network (DQN) algorithm, double DQN has better convergence, adaptive ability and better estimation accuracy. It can provide new ideas for the accurate estimation of SOC.