Enhanced Particle Swarm Assisted Adaptive Hybrid Filtering with Hysteresis Characteristic Correction for State-of-Charge and State-of-Energy Joint Prediction of Lithium-Ion Batteries

Abstract Accurate estimation of state-of-charge (SOC) and state-of-energy (SOE) is crucial for ensuring the safety and range capability of battery systems. This paper presents an enhanced particle swarm-assisted adaptive extended Kalman particle filtering framework incorporating hysteresis correction for joint SOC and SOE estimation. Firstly, a second-order equivalent circuit model that accounts for battery hysteresis is established. To enhance the accuracy and dynamic performance of parameter identification, a recursive extended least squares algorithm equipped with an adaptive forgetting factor is devised. Secondly, to alleviate particle impoverishment and diversity loss inherent in conventional particle filters, a hybrid strategy integrating extended Kalman filtering with particle swarm optimization is introduced, thereby bolstering the stability and precision of state estimation. Moreover, an adaptive mechanism for noise covariance is incorporated to strengthen the algorithm's robustness and convergence under complex operating conditions. Experimental results demonstrate that during parameter identification, the mean-absolute-error (MAE) and root-mean-square-error (RMSE) of the proposed method are confined to the ranges of 6.45 mV to 10.61 mV. For SOC and SOE estimation, the MAE and RMSE are maintained between 0.19% and 0.51%. Furthermore, the proposed method maintains stable convergence even with inaccurate initial values, confirming its good accuracy and robustness.