State of charge estimation for lithium-ion batteries based on physics-embedded neural network

Abstract

Expansion of lithium-ion batteries (LIBs) impacts performance and safety. Therefore, accurately estimating the state of swelling displacement (SoD) and state of charge (SoC) is crucial for battery health management. However, SoC estimation methods often ignore the impact of expansion on battery performance, leading to estimation errors. To address this issue, this paper proposes a convolutional neural network (CNN)-long short-term memory (LSTM) estimation framework embedded with physical information. First, at the physical level, the relationship between displacement and charge state is analyzed using an electrochemical-mechanical coupling model, which provides certain prior physical knowledge for subsequent estimation. At the mathematical level, Pearson correlation analysis is used to quantify the correlation between displacement and SoC. Next, a CNN-LSTM framework is employed to estimate the displacement and use it as key physical information for SoC estimation. Finally, the proposed method is validated using test data under various operating conditions. The results show that the accuracy of SoC estimation is significantly improved with including displacement, with the mean absolute error (MAE) reduced by about 16.07 % compared to when displacement is not included. The proposed method depicts good prediction accuracy and computational efficiency under different charge-discharge rates, validating the effectiveness of displacement as key physical information.