Enhanced multi-scale signal decomposition transformer neural network for state of health estimation of lithium-ion batteries.

Li, Yang; Shi, Haotian; Huang, Qi; Li, Ke; Liu, Chunmei; Nie, Shiliang; Jia, Xianyi; Fernandez, Carlos

Abstract

The accurate estimation of battery state of health (SOH) is important in the fields of electric vehicles, energy storage devices, and renewable energy. To address the accuracy challenges of SOH estimation caused by the phenomenon of small-scale capacity regeneration during battery aging, this paper proposes the efficient SOH estimation framework based on the combination of multi-scale signal decomposition and improved Transformer model. The improved sparrow search algorithm (ISSA) through coupling multiple strategies is used in the framework to automatically find the best hyperparameters for the encoder, decoder and optimizer structures in the Transformer model. Another important core of the proposed framework is the use the complete ensemble empirical mode decomposition with adaptive noise (CEEMDAN) technology, which effectively separates the characteristics of local capacity regeneration and overall capacity decline to enhance the predictability of input data. Finally, the feasibility and validity of the proposed framework are validated on public lithium-ion battery datasets. The results demonstrate the root mean square error (RMSE) consistently below 0.47 % across all test cases. Remarkably, the estimation error remains within 0.02 % even when only 50 % of the training data is utilized.

Citation

LI, Y., SHI, H., HUANG, Q., LI, K., LIU, C., NIE, S., JIA, X. and FERNANDEZ, C. 2025. Enhanced multi-scale signal decomposition transformer neural network for state of health estimation of lithium-ion batteries. Journal of energy storage [online], 134(B), article number 118191. Available from: https://doi.org/10.1016/j.est.2025.118191

• Journal Article Type: Article
• Acceptance Date: Aug 21, 2025
• Online Publication Date: Aug 28, 2025
• Publication Date: Oct 30, 2025
• Deposit Date: Aug 29, 2025
• Publicly Available Date: Aug 29, 2026
• Journal: Journal of energy storage
• Print ISSN: 2352-152X
• Electronic ISSN: 2352-1538
• Publisher: Elsevier
• Peer Reviewed: Peer Reviewed
• Volume: 134
• Issue: B
• Article Number: 118191
• DOI: https://doi.org/10.1016/j.est.2025.118191
• Keywords: Lithium-ion battery; State of health; Signal mode decomposition; Improved sparrow search algorithm; Transform
• Public URL: https://rgu-repository.worktribe.com/output/2989549