Towards efficient state of charge estimation of lithium-ion batteries using canonical correlation analysis

State of charge (SOC) estimation plays an important role for lithium-ion batteries indicating the remaining charge during a cycle. The deep networks adopt the complicated network structure with a large number of parameters, which are sophisticated and lack generality. This paper presents a novel and facile data-driven method based on canonical correlation analysis (CCA) for battery SOC estimation. Firstly, CCA is demonstrated in a regression form and given with an optimizing algorithm for battery SOC estimation. Then the offline training results are followed by the Kalman filter (KF) for online error correction. Finally, a robust canonical correlation analysis (RCCA) is proposed for noise corruption on the input data. Simulation results on different dynamic profiles show the effectiveness of RCCA compared with CCA with improved accuracy by 40% for input noise, and the final results of RCCA with KF achieve root mean squared error (RMSE) of 0.71%. The proposed method achieves superior results in accuracy under input noise and is also computationally efficient with less training time compared with other methods. (C) 2022 Elsevier Ltd. All rights reserved.