Performance Evaluation of Kalman-based Algorithms for SoC Estimation in Electric Vehicle Batteries

Accurate estimation of the State of Charge (SoC) is vital for the reliable and safe operation of lithium-ion batteries used in Electric Vehicles (EVs). The Battery Management System (BMS) relies on precise prediction of SoC to ensure optimal battery usage, prevent overcharging or deep discharging, and extend battery life in long term. This paper attempts to sincerely investigate and compare the performance of two widely used filtering techniques— Standard Kalman Filter (KF) and Extended Kalman Filter (EKF)—for SoC estimation using battery data from the real-world. The KF is often implemented under the assumption of linear battery dynamics. On the other hand, the EKF incorporates the nonlinear relationship between Open Circuit Voltage (OCV) and SoC through a fitted curve extracted from experimental data. A publicly available open source dataset from high-rate discharge tests of a lithium-cobaltoxide pouch cell is used for validation purpose. Based on cumulative discharge capacity, the true SoC is computed and is then compared against SoC estimates from both filters. The results are indicative of the fact that the EKF, with its nonlinear measurement model, more closely follows the true SoC trajectory, especially under dynamic current conditions. The improved accuracy of EKF over the standard KF is confirmed by evaluation metrics such as RMSE and MAE. These findings emphasize the importance of model selection and also highlight the advantages of nonlinear filtering for SoC estimation in practical battery systems. The findings also contribute to the ongoing efforts in enhancing the accuracy of battery state estimation which in turn supports safer, reliable, and more efficient operation of electric vehicles.