A case study to compare the methods of SOC estimation for lithium-ion batteries using second-order RC circuit model

Accurate modeling and state-of-charge (SOC) estimation of power batteries are key issues in battery management systems (BMSs). In this study, a 24 Ah lithium-ion power battery is taken as the research object. Based on HPPC test data and constant-current discharge data, a second-order RC equivalent circuit model is established, and the performance of three parameter identification methods (RLS, FFRLS, and VFFRLS) as well as two state estimation algorithms (EKF and UKF) for SOC estimation is systematically compared.The results show that all three algorithms can achieve online parameter identification, among which VFFRLS improves parameter convergence speed and identification accuracy through a dynamically adjusted forgetting factor. For SOC estimation, the EKF maintains terminal voltage errors within 80 mV, and the SOC estimation error converges after approximately 1700 s, with an RMSE of 8.93%. In contrast, the UKF exhibits better performance, with terminal voltage errors mainly within 20 mV (maximum below 50 mV) and SOC estimation errors within 2%. Overall, the VFFRLS–UKF combination demonstrates the best SOC estimation performance under dynamic operating conditions.The results provide a reference for the selection and optimization of online SOC estimation methods in practical BMS applications.