State of health estimation of lithium-ion batteries based on multiple health factors and BO-Seq2Seq model

Abstract

State of health (SoH) is a key indicator for evaluating the degradation level of lithium-ion batteries (LIBs), and its accurate prediction is crucial for the reliability and safety of battery management systems. This paper proposes an improved sequence-to-sequence (Seq2Seq) model for SoH estimation based on the selection of relevant multiple types of health factors (HFs). In order to better describe the SoH of LIBs, Pearson's correlation coefficient is employed to select HFs that is highly correlated with the SoH. The hyperparameter optimization method based on Bayesian optimization (BO) is used to solve the challenge of hyperparameter selection in Seq2Seq models. This method effectively utilizes the strong global search ability of BO to obtain the global optimal solution. The proposed method is validated using a dataset provided by NASA. The results show that the BO-Seq2Seq model proposed in this work exhibits competitive accuracy in estimation performance, with an error of less than 1 %. Compared with methods that only consider a single HF, the method that considers multiple HFs in this work significantly improves the accuracy of estimation. The estimation performance indicators of the proposed BO-Seq2Seq model, such as the root mean square errors are all within 1 %, the mean absolute errors are all within 0.5 % and the mean square errors are all within 0.004 %, are significantly lower than other methods.