Enhanced SOC and SOH estimation for Li-ion batteries based on combining adaptive central difference Kalman filter and discrete-time sliding mode observer

Abstract

A combined adaptive central difference Kalman filter and discrete-time sliding mode observer (ACDKF-DSMO) algorithm is proposed to improve the robustness and accuracy of estimating both state-of-charge (SOC) and state-of-health (SOH) for Li-ion batteries. The pulse discharge method is utilized to identify the model parameters using the dual polarization (DP) model. Since the discrete-time sliding mode observer (DSMO) demonstrates good robustness and adaptability in effectively addressing uncertainties and nonlinearities, this advantage is combined with the accuracy of the central difference Kalman filter (CDKF). Specifically, the state equation of DSMO is incorporated into the state update phase of CDKF, thereby improving SOC estimation through synergy with Sage-Husa adaptive filter. Aiming at the problem that battery capacity will decline during actual operation, the extended Kalman filter (EKF) is employed to estimate the capacity over a long-time scale while the ACDKF-DSMO algorithm is employed to estimate the SOC over a short-time scale, since the capacity changes slowly but the SOC changes quickly. The single SOC estimation experiment based on the ACDKF-DSMO was conducted under the non-noise condition, Gaussian white noise and colored noise respectively. The mean absolute error (MAE) were 0.556%, 0.87%, and 0.82%, and the root mean square error (RMSE) were 0.557%, 0.88%, and 0.83% respectively, demonstrating good estimation accuracy and robustness. The multi-time scale joint estimation of SOC and SOH was validated under mixed cycle conditions, and the MAE and RMSE of SOC estimation were 0.47% and 0.56%, respectively. Compared with the single SOC estimation algorithm, the results were reduced by 1.16% and 1.39% respectively, indicating that the joint estimation method can significantly enhance the accuracy of SOC estimation.