State-of-Health Estimation of Lithium-Ion Batteries Based on EIS and CNN-Transformer Network

Estimating the state of health (SOH) and remaining useful life (RUL) of lithium-ion batteries is a challenging task. In this paper, we propose a novel convolution neural network (CNN)-Transformer model for the estimation of SOH, using only electrochemical impedance spectroscopy (EIS) data. The publicly accessible datasets are employed for both the training and testing phases. In the data preprocessing stage, highly correlated EIS features are selected using the Pearson correlation coefficient (PCC). The proposed model is implemented in MATLAB for training and prediction on a low-performance platform devoid of Graphical Processing Unit (GPU) support, thereby demonstrating its broad applicability. The simulation outcomes indicate that the mean absolute error (MAE) and root mean square error (RMSE) are maintained within 3%, and the coefficient of determination (R2) reaches 98.72<sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">%</sup>, outperforming standalone Transformer and LSTM models. The proposed model is therefore both accurate and practical, offering the possibility of applying deep learning models into battery management systems (BMS).