A State‐of‐Health Estimation Method for Lithium Batteries Based on Incremental Energy Analysis and Bayesian Transformer

Abstract

Lithium‐ion batteries (LIBs) have wide‐ranging applications in areas such as electric vehicles and mobile devices. Accurate estimation of the state of health (SOH) of batteries is an important aspect of battery state estimation. Battery capacity cannot be precisely measured due to negative factors such as aging effects. To address this issue, this paper proposes a LIB’s SOH estimation method based on incremental energy analysis (IEA) and transformer. First, data collected during the constant‐current (CC) charging phase of the battery are used to create and analyze the IEA curve. Then, the peaks and areas of the curve are proposed as health characteristics of the LIB. Cosine similarity analysis (CSA) is employed to determine the correlation between each health characteristic and SOH, as well as the correlations between different health characteristics. Finally, an accurate estimation model of battery health was developed using the Bayesian‐transformer model by plotting the relationship between health characteristics and battery health. To validate the reliability of the model, comparisons with regression evaluation metrics of other models such as support vector regression (SVR) and recurrent neural network (RNN) are conducted under different charging rates. The conclusion is drawn that this model exhibits an R2 greater than 98%, MAE less than 0.22%, RMSE less than 0.26%, and MAPE less than 0.0026%. Its accuracy is significantly improved compared to the same type of methods, and it can be used for high accuracy estimation of SOH in LIBs. The multistep prediction of a single step adopted by the model can effectively overcome the capacity regeneration problem in the field of SOH estimation, which will inspire future experts and scholars to improve the accuracy of SOH estimation.