State of charge estimation for lithium-ion batteries based on gate recurrent unit and unscented Kalman filtering

Abstract

Accurate and robust state of charge (SOC) estimation for lithium-ion batteries is crucial for battery management systems. In this study, we proposed an SOC estimation approach for lithium-ion batteries that integrates the gate recurrent unit (GRU) with the unscented Kalman filtering (UKF) algorithm. This integration aims to enhance the robustness of SOC estimation under complex working conditions and varying temperatures. The GRU neural network is employed to establish an offline training model, while the fusion of the UKF online estimation is utilized to obtain smooth SOC estimation results for lithium-ion batteries. This approach realized a closed-loop SOC estimation strategy. The 18,650 and 26,650 LiFePO₄ batteries were selected for experiments conducted under different charging and discharging conditions at operating temperatures of 10℃, 25℃, and 40 °C. The experiment verified the high accuracy and robustness of the proposed GRU and UKF fusion approach, with both the root mean square error (RMSE) and the mean absolute error (MAE) maintained within 1%.