State of charge accurate estimation of lithium-ion batteries based on augmenting observation dimension estimator over wide temperature range

This paper proposes an observation-based incremental dimension state estimator utilizing electrochemical mechanism information feedback. The state of charge (SOC) of lithium ternary battery is estimated by combining it with an adaptive cubature Kalman filter (ACKF) algorithm under wide temperature ranges and various dynamic operating conditions. The proposed algorithm is compared with the ACKF algorithm, which implements online parameter identification using the Forgetting Factor Recursive Least Squares (FFRLS) method. The results demonstrate that the proposed method outperforms the FFRLS-ACKF, with root mean square error (RMSE) and mean absolute error (MAE) of 0.83 % and 0.71 %, respectively, compared to 1.14 % and 1.25 % for the FFRLS-ACKF. However, a significant degradation in SOC estimation accuracy is observed when the temperature drops to around −10 °C. To address this issue, this paper also presents a joint ACKF and deep learning-based algorithm model, which achieves precise SOC estimation, with MAE and RMSE not exceeding 0.24 % and 0.32 %, respectively. In conclusion, the proposed algorithm offers notable advantages in SOC estimation and provides valuable guidance for practical engineering applications.