Evaluation of electrochemical models based battery state-of-charge estimation approaches for electric vehicles

Real-time and accurate state-of-charge (SoC) estimation of lithium-ion batteries is a critical issue for efficient monitoring, control and utilization of advanced battery management systems (BMS) in electric vehicles (EVs). The electrochemical mechanism model can accurately describe the spatially distributed behavior of the internal states of the battery, but the model is complex and computationally huge, which is difficult to simulation in vehicle BMS. To solve these problems, it is necessary to simplify the battery mechanism model and study the model-based SoC estimation approaches. In this paper, two order-reduced models including an average-electrode model (AEM) and a single particle model (SPM) are first proposed. Additionally, the reduced-models combined with algorithms, including an extended Kalman filter (EKF), a sliding-mode observer (SMO) with a uniform reaching law (URL) and an SMO with an exponential reaching law (ERL), are employed to design battery SoC observers. To achieve an optimal trade-off between the tracking accuracy and convergence ability, the performances of these approaches are compared under an Urban Dynamometer Driving Schedule (UDDS) test. The comparison results indicate that the SPM-EKF approach can obtain a reliable battery voltage response and a more accurate SoC estimation than other approaches.