A lithium-ion battery state of charge estimation method based on the fusion of data-driven and Kalman filter-based method

State of charge (SOC) estimation is crucial for battery management systems (BMS), which relies on accurate battery mathematical models. The conventional equivalent circuit model, however, does not describe the actual electrochemical nonlinear dynamic response of a lithium-ion battery. In this work, a novel a nonlinear equivalent circuit model (NLECM) is established. It is based on an odd random phase multisine signal for parameter estimation. The signal allows parametrization over a bandwidth broader than that of a conventional HPPC signal. Based on the established NLECM model, a window-varying adaptive extended Kalman filter (WVAEKF) with data-driven algorithm is first applied for SOC estimation. The designed WVAEKF can identify variations in the error innovation sequence distribution and modify the window’s length. Learning from a large number of battery operating data, the data-driven algorithm extracts useful features for the estimation of SOC error and improves the accuracy of the SOC estimation. The experimental results show that the error of SOC estimation by WVAEKF with data-driven algorithm is limited to 1% under dynamic stress test (DST) conditions and 0.5C constant current discharge. Compared with artificial neural network and traditional AEKF, the RMSE of the proposed algorithm is reduced by 93 % and 96 % respectively, which shows that the algorithm has higher accuracy under DST conditions .