State-of-charge estimation for lithium-ion battery during constant current charging process based on model parameters updated periodically

Abstract

With online established battery model, model-based estimation method can track battery state-of-charge (SOC) precisely under dynamic conditions. Nevertheless, both recursive least square-based and filter-based methods cannot distinguish whether the voltage difference comes from SOC difference or internal resistance difference during constant current (CC) conditions, further leading to erroneously identified model parameters and inaccurate SOC estimation. To address this issue, a novel SOC estimation method during CC charging process by fusion of global optimization algorithm and Kalman filter family algorithm is developed in this paper. Firstly, some key parameters that are helpful for initialization and lower/upper bounds setting for global optimization method are extracted from electric vehicles’ driving process. Secondly, considering the shortcomings in traditional global optimization methods, including possible premature convergence, slow search speed in the late stage and relatively large computational cost, an improved particle swarm optimization is designed to periodically update model parameters during CC charging process. With obtained model parameters, SOC is further tracked via extended Kalman filter (EKF). The verification results based on experimental data demonstrates that the developed method can significantly weaken the strong cross-interference between model parameters and SOC, further achieving much more accurate SOC estimation than existing dual/joint EKF during CC charging process.