Enhanced few-shot state-of-health estimation for lithium-ion batteries via Masked Autoencoder

Accurately estimating the state-of-health (SOH) of lithium-ion batteries (LIBs) is crucial for optimizing performance, ensuring operational safety, and enabling predictive maintenance in battery management systems. With the widespread adoption of LIBs, a large amount of field data has been generated...

Celý popis

Uloženo v:
Podrobná bibliografie
Vydáno v:Energy (Oxford) Ročník 335; s. 138263
Hlavní autoři: Shen, Yifan, Guo, Dongxu, Wang, Yu, Chen, Jianguo, Liu, Xuyang, Han, Xuebing, Zheng, Yuejiu, Ouyang, Minggao
Médium: Journal Article
Jazyk:angličtina
Vydáno: Elsevier Ltd 30.10.2025
Témata:
Electric vehicle data
Lithium-ion battery
Masked Autoencoder
Semi-supervised learning
SOH estimation
SOH estimation
Lithium-ion battery
Electric vehicle data
Semi-supervised learning
Masked Autoencoder
ISSN:0360-5442
On-line přístup:Získat plný text
Tagy: Přidat tag
Žádné tagy, Buďte první, kdo vytvoří štítek k tomuto záznamu!
Popis
Shrnutí:Accurately estimating the state-of-health (SOH) of lithium-ion batteries (LIBs) is crucial for optimizing performance, ensuring operational safety, and enabling predictive maintenance in battery management systems. With the widespread adoption of LIBs, a large amount of field data has been generated, yet current data-driven SOH estimation methods often fail to fully utilize it due to the lack of labeled data. To address this, we propose a method based on semi-supervised learning to exploit large-scale unlabeled data for accurate SOH estimation. A generative unsupervised model, the Masked Autoencoder (MAE), is pre-trained on unlabeled field charging data to automatically extract latent representations related to SOH. The model is then fine-tuned with a small amount of labeled data. Experimental results show that using only 20 % of the labeled data usually required for supervised learning, the method achieves an RMSE of 2.14 %. The latent representation extraction capability of the MAE is validated via incremental capacity (IC) analysis, which explains the 14 % improvement in estimation accuracy (RMSE of 1.84 %) when using data from a specific voltage range (3.8–3.9 V). Furthermore, experiments demonstrate that even with only 21.33 min of charging data—consisting of only charge quantity and voltage signals—the model can still achieve a competitive RMSE of 1.94 %. This work introduces a novel approach for SOH estimation using large-scale, unlabeled field data and provides valuable insights for battery management in the era of artificial intelligence. •SOH labeling reduced by 80 % via MAE-based semi-supervised learning.•MAE-extracted SOH features match IC curves in mid-voltage zone.•1.94 % SOH error using charge quantity and voltage from a 21-min charge segment.•Semi-supervised MAE yields 1.7 % error on 2-year electric vehicle data.
ISSN:0360-5442
DOI:10.1016/j.energy.2025.138263