State of charge estimation for lithium-ion batteries using an adaptive cubature Kalman filter based on improved generalized minimum error entropy criterion

Abstract

Accurate and robust estimation of the State of Charge (SOC) in complex environments is vital to achieving high battery performance, extended lifespan, enhanced safety, and improved user experience. Conventional estimation methods often neglect the impact of temperature on battery during the modeling process. To address this, this paper presents a battery modeling method applicable across a comprehensive temperature range that encompasses all potential operating temperatures. A second-order equivalent circuit model (ECM) is developed for the battery, with parameters defined within a specified temperature range. Furthermore, an adaptive cubature Kalman filter based on an improved minimum error entropy criterion (IMEF-ACKF) is introduced to address the significant accuracy degradation of traditional methods under non-Gaussian noise and substantial outlier interference. The generalized minimum error entropy criterion is combined with the generalized maximum correntropy criterion to replace the traditional minimum mean-square error (MMSE) criterion, addressing the influence of non-Gaussian noise. Then, exponential transformation of system residual is used to mitigate the impact of large outliers, and adaptive filter is incorporated to improve the stability of the calculation process. Predictions at various cycle tests and temperatures show that RMSE values below 0.3% and MAX values below 0.6% could be achieved by the proposed method in environments without additional noises. Even under non-Gaussian and impulsive noise conditions, the RMSE value of the optimized method remains below 0.9%. The results indicate that this method consistently maintains excellent estimation accuracy and robustness across all application scenarios.