A Lithium-ion Battery SOH Estimation Method Based on WT-EMD-Informer Model

Given the numerous advantages of lithium-ion battery and their widespread application across various fields, precisely assessing their state of health (SOH) holds paramount importance in safeguarding the stable operation of devices. To enhance the quality of lithium battery datasets and boost the precision of SOH estimation, this paper introduces an approach for estimating the SOH of lithium-ion battery by leveraging the WT-EMD-Informer model. Firstly, this method extracts four health factors (HFs) from the public dataset based on energy features approach and builds a HFs dataset (HFs-dataset). Secondly, the HFs-dataset is denoised using wavelet transform (WT) to obtain the processed dataset (Processed-Dataset). Thirdly, the Processed-Dataset undergoes signal decomposition and reconstruction using empirical mode decomposition (EMD) combined with deep learning model to obtain the reconstructed dataset (Reconstructed-Dataset); The Pearson correlation analysis is employed to determine whether the Reconstructed-Dataset can form the required dataset (Dataset) for this paper. Finally, the Dataset is divided into a training set and a testing set, with 70 percent of the data allocated for training and 30 percent for testing. This division is used to train and test the Informer model for estimating the SOH of the lithium battery. The experimental outcomes demonstrate that our proposed method outperforms existing approaches, achieving an average RMSE of under 0.6%, an average MAE of less than 0.5%, and an average MAPE of under 1%.