Enabling Accurate Battery Remaining Useful Life Prediction and Through-Life Degradation Mapping With Scarce End-of-Life Samples
IEEE Transactions on Industrial Informatics · 2026-05 · 22 (5) · 4492-4503
DOI: 10.1109/tii.2026.3658872 ↗Kurz gefasst
Fachartikel in der Zeitschrift IEEE Transactions on Industrial Informatics, 2026-05. Verfasst von Ruohan Guo und 6 weiteren Autor:innen. Die Arbeit wurde bisher 16-mal zitiert (Crossref) und nennt 30 Literaturangaben. Gefördert von: National Natural Science Foundation of China, PolyU startup fund.
Abstract (Originalsprache)
Accurate remaining useful life (RUL) prediction and through-life degradation mapping are essential for reliable prognostics and health management of lithium-ion batteries. However, their practical deployment is often constrained by the scarcity of high-quality labeled degradation data and the limitations of conventional single-point prediction paradigms. This article proposes a novel trajectory-aware few-shot learning framework to address these challenges. First, we develop an end-of-life (EOL) criterion-adaptive training-sample expansion strategy that constructs multiple virtual state of health (SOH) thresholds within a usable degradation range, thereby generating informative pseudosupervision and effectively leveraging unlabeled batteries that have not yet reached EOL. Second, we design a dual-input deep learning model that uses only a single-cycle constant-voltage charging profile as the primary exogenous input, together with a user-definable SOH threshold as a secondary configurable parameter, enabling flexible RUL estimation for arbitrary target thresholds. By fixing the charging-profile input and varying the threshold input, the model further outputs a sequence of ageing markers, which collectively characterize the through-life degradation trajectory of a target battery. Extensive experiments across three datasets with different chemistries and operating conditions demonstrate consistent accuracy gains over eight baselines, together with robustness under noise-corrupted supervision and improved generalization under domain shift.
Förderung
- National Natural Science Foundation of China
- PolyU startup fund
Zitieren
APA
Guo, R., Zhang, K., Ke, S., He, S., Tian, J., Shen, W., & Chung, C. Y. (2026). Enabling Accurate Battery Remaining Useful Life Prediction and Through-Life Degradation Mapping With Scarce End-of-Life Samples. IEEE Transactions on Industrial Informatics, 22(5), 4492-4503. https://doi.org/10.1109/tii.2026.3658872
Vancouver
Guo R, Zhang K, Ke S, He S, Tian J, Shen W, et al. Enabling Accurate Battery Remaining Useful Life Prediction and Through-Life Degradation Mapping With Scarce End-of-Life Samples. IEEE Transactions on Industrial Informatics. 2026;22(5):4492-4503. doi:10.1109/tii.2026.3658872
BibTeX
@article{Guo2026,
author = {Guo, Ruohan and Zhang, Kui and Ke, Song and He, Shangyang and Tian, Jinpeng and Shen, Weixiang and Chung, Chi Yung},
title = {Enabling Accurate Battery Remaining Useful Life Prediction and Through-Life Degradation Mapping With Scarce End-of-Life Samples},
journal = {IEEE Transactions on Industrial Informatics},
year = {2026},
volume = {22},
number = {5},
pages = {4492--4503},
doi = {10.1109/tii.2026.3658872}
}
EUVIMED