A complete ensemble empirical mode decomposition with adaptive noise deep autoregressive recurrent neural network method for the whole life remaining useful life prediction of lithium-ion batteries.

Zhang, Chuyan; Wang, Shunli; Yu, Chunmei; Wang, Yangtao; Fernandez, Carlos

Abstract

The real-time prediction of the remaining useful life (RUL) of lithium-ion batteries provides an effective mean of preventing accidents. An improved adaptive noise-reduction deep learning method is applied to achieve adaptive noise-reduction decomposition of lithium-ion battery capacity using complete ensemble empirical mode decomposition with adaptive noise (CEEMDAN), and then the resulting intrinsic mode function (IMF) components are continued to be reconstructed, followed by input to a deep autoregressive recurrent neural network (DeepAR) to accurately predict the remaining useful life of lithium-ion batteries. To begin with, the lithium-ion battery data are screened and the correlation with capacity is analyzed by Pearson and Spearman to derive the indirect health factors. Then, the capacity data are decomposed by CEEMDAN to derive a relatively smooth IMF with the trend component for reconstruction, which is the output of the DeepAR model to predict the RUL of lithium-ion batteries with the indirect health factor as the output. The experimental results obtained after validation of the data set validate that the improved adaptive noise reduction DeepAR prediction model has superior prediction accuracy and greater stability, with all remaining life errors less than 5 times and all root mean square errors (RMSE) less than 2.5%.

Citation

ZHANG, C., WANG, S., YU, C., WANG, Y. and FERNANDEZ, C. 2023. A complete ensemble empirical mode decomposition with adaptive noise deep autoregressive recurrent neural network method for the whole life remaining useful life prediction of lithium-ion batteries. Ionics [online], 29(10), pages 4337-4349. Available from: https://doi.org/10.1007/s11581-023-05152-2

• Journal Article Type: Article
• Acceptance Date: Jul 28, 2023
• Online Publication Date: Aug 8, 2023
• Publication Date: Oct 31, 2023
• Deposit Date: Aug 24, 2023
• Publicly Available Date: Aug 9, 2024
• Journal: Ionics
• Print ISSN: 0947-7047
• Electronic ISSN: 1862-0760
• Publisher: Springer
• Peer Reviewed: Peer Reviewed
• Volume: 29
• Issue: 10
• Pages: 4337-4349
• DOI: https://doi.org/10.1007/s11581-023-05152-2
• Keywords: Lithium-ion batteries; Remaining useful life; Health factor; Complete ensemble empirical mode decomposition with adaptive noise; Deep autoregressive recurrent neural network
• Public URL: https://rgu-repository.worktribe.com/output/2048556