State of Charge Estimation of Lithium-Ion Battery Pack in Electric Vehicles Based on Inconsistency Compensation Modeling

This paper proposes a novel approach to address the state of charge (SOC) estimation challenges of lithiumion battery packs in electric vehicles characterized by inherent inconsistencies. The proposed method aims to achieve a delicate equilibrium between estimation accuracy and model complexity. Firstly, an inconsistency compensation model is proposed, utilizing nonzero mean virtual measurement noise to offset external discrepancies induced by battery pack inconsistency. Subsequently, a variational Bayesian unscented Kalman filter (VBUKF) is derived for simultaneous estimation of SOC and the mean and variance of virtual measurement noise. Concurrently, within the same time scale, a variational Bayesian Kalman filter (VBKF) is employed for internal resistance (R) estimation. Finally, the proposed method is validated through experiments on two battery packs with varying degrees of inconsistency under pulsed-discharge and urban dynamometer driving schedule conditions at various temperatures. Experimental results demonstrate that the proposed SOC+R compensation model yields significantly more accurate SOC estimates than the big cell model and the compensation model without R estimation. In addition, it exhibits comparable accuracy to the representative cell model but without the burden of selecting representative cells, thus effectively balancing complexity and accuracy. Moreover, the proposed model is proved to be more robust through simulations in the case of random errors and abrupt disturbances