Abstract
Fast and reliable validation of new designs in complex physical systems such as batteries is critical to accelerating technological innovation. However, battery development remains bottlenecked by the high time and energy costs required to evaluate the lifetime of new designs1,2. Notably, existing lifetime forecasting approaches require datasets containing battery lifetime labels for target designs to improve accuracy and cannot make reliable predictions before prototyping, thus limiting rapid feedback3,4. Here we introduce Discovery Learning, a scientific machine learning approach that integrates active learning5, physics-guided learning6 and zero-shot learning7 into a human-like reasoning loop, drawing inspiration from educational psychology. Discovery Learning can learn from historical battery designs and reduce the need for prototyping, thereby predicting the lifetime of new designs from minimal experiments. To test Discovery Learning, we present industrial-grade battery data comprising 123 large-format lithium-ion pouch cells, including diverse material–design combinations and cycling protocols. Trained on public datasets of cell designs different from ours, Discovery Learning achieves 7.2% test error in predicting cycle life using physical features from the first 50 cycles of 51% of cell prototypes. Under conservative assumptions, this results in savings of 98% in time and 95% in energy compared with conventional practices. Discovery Learning represents a key advance in accurate and efficient battery lifetime prediction and, more broadly, helps realize the promise of machine learning to accelerate scientific discovery8.
This is a preview of subscription content, access via your institution
Access options
Access Nature and 54 other Nature Portfolio journals
Get Nature+, our best-value online-access subscription
$32.99 / 30 days
cancel any time
Subscribe to this journal
Receive 51 print issues and online access
$199.00 per year
only $3.90 per issue
Buy this article
- Purchase on SpringerLink
- Instant access to the full article PDF.
USD 39.95
Prices may be subject to local taxes which are calculated during checkout
Similar content being viewed by others
Data availability
The data required to reproduce the results of this study are available at Zenodo44 (https://doi.org/10.5281/zenodo.17654407) and Code Ocean (https://doi.org/10.24433/CO.5569470.v1). Commercially sensitive information not directly relevant to this study may be available from the corresponding authors on request. The public datasets used in this study are either accessible online or available on request from the original authors. The A123-M1A dataset and the LG-HG2 dataset are from Sandia National Laboratories19. The Sony-VTC5A dataset is from Technical University of Munich24. The LG-MJ1 dataset is from Flemish Institute for Technological Research21. The Sony-VTC6 dataset is from Carnegie Mellon University25. The Samsung-25R dataset is from Karlsruhe Institute of Technology23.
Code availability
The codes developed in this study are available at Code Ocean (https://doi.org/10.24433/CO.5569470.v1).
References
Figgener, J. et al. Multi-year field measurements of home storage systems and their use in capacity estimation. Nat. Energy 9, 1438–1447 (2024).
Degen, F., Winter, M., Bendig, D. & Tübke, J. Energy consumption of current and future production of lithium-ion and post lithium-ion battery cells. Nat. Energy 8, 1284–1295 (2023).
Severson, K. A. et al. Data-driven prediction of battery cycle life before capacity degradation. Nat. Energy 4, 383–391 (2019).
Attia, P. M. et al. Closed-loop optimization of fast-charging protocols for batteries with machine learning. Nature 578, 397–402 (2020).
Settles, B. Active Learning Literature Survey. University of Wisconsin–Madison https://minds.wisconsin.edu/handle/1793/60660 (2009).
Yu, R. & Wang, R. Learning dynamical systems from data: an introduction to physics-guided deep learning. Proc. Natl Acad. Sci. 121, e2311808121 (2024).
Xian, Y., Lampert, C. H., Schiele, B. & Akata, Z. Zero-shot learning—a comprehensive evaluation of the good, the bad and the ugly. IEEE Trans Pattern Anal. Mach. Intell. 41, 2251–2265 (2019).
Wang, H. et al. Scientific discovery in the age of artificial intelligence. Nature 620, 47–60 (2023).
Chen, S. et al. External Li supply reshapes Li deficiency and lifetime limit of batteries. Nature 638, 676–683 (2025).
Lam, V. N. et al. A decade of insights: delving into calendar aging trends and implications. Joule 9, 101796 (2025).
Zhu, Y., Gu, X., Liu, K., Zhao, W. & Shang, Y. Rapid test and assessment of lithium-ion battery cycle life based on transfer learning. IEEE Trans. Transp. Electrification 10, 9133–9143 (2024).
Edge, J. S. et al. Lithium ion battery degradation: what you need to know. Phys. Chem. Chem. Phys. 23, 8200–8221 (2021).
Zhang, H. et al. Battery lifetime prediction across diverse ageing conditions with inter-cell deep learning. Nat. Mach. Intell. 7, 270–277 (2025).
Guo, N. et al. Semi-supervised learning for explainable few-shot battery lifetime prediction. Joule 8, 1820–1836 (2024).
Aykol, M., Herring, P. & Anapolsky, A. Machine learning for continuous innovation in battery technologies. Nat. Rev. Mater. 5, 725–727 (2020).
Ward, L. et al. Principles of the Battery Data Genome. Joule 6, 2253–2271 (2022).
Merchant, A. et al. Scaling deep learning for materials discovery. Nature 624, 80–85 (2023).
Bruner, J. S. The act of discovery. Harvard Educ. Rev. 31, 21–32 (1961).
Preger, Y. et al. Degradation of commercial lithium-ion cells as a function of chemistry and cycling conditions. J. Electrochem. Soc. 167, 120532 (2020).
Lain, M. J., Brandon, J. & Kendrick, E. Design strategies for high power vs. high energy lithium ion cells. Batteries 5, 64 (2019).
Trad, K. Lifecycle ageing tests on commercial 18650 Li ion cell @ 25 °C and 45 °C. 4TU. ResearchData https://doi.org/10.4121/13739296.v1 (2021).
Heenan, T. M. M. et al. An advanced microstructural and electrochemical datasheet on 18650 Li-ion batteries with nickel-rich NMC811 cathodes and graphite-silicon anodes. J. Electrochem. Soc. 167, 140530 (2020).
Zhu, J. et al. Data-driven capacity estimation of commercial lithium-ion batteries from voltage relaxation. Nat. Commun. 13, 2261 (2022).
Wildfeuer, L. et al. Experimental degradation study of a commercial lithium-ion battery. J. Power Sources 560, 232498 (2023).
Bills, A. et al. A battery dataset for electric vertical takeoff and landing aircraft. Sci. Data 10, 344 (2023).
Yang, X.-G., Liu, T. & Wang, C.-Y. Thermally modulated lithium iron phosphate batteries for mass-market electric vehicles. Nat. Energy 6, 176–185 (2021).
Kim, J.-H. et al. Kosmotropic aqueous processing solution for green lithium battery cathode manufacturing. Nat. Commun 16, 1686 (2025).
Ko, S. et al. Rapid safety screening realized by accelerating rate calorimetry with lab-scale small batteries. Nat. Energy 10, 707–714 (2025).
Wang, C.-Y. et al. Fast charging of energy-dense lithium-ion batteries. Nature 611, 485–490 (2022).
Zhang, J., Che, Y., Teodorescu, R., Song, Z. & Hu, X. Energy storage management in electric vehicles. Nat. Rev. Clean Technol. 1, 161–175 (2025).
Szymanski, N. J. et al. An autonomous laboratory for the accelerated synthesis of novel materials. Nature 624, 86–91 (2023).
Angello, N. H. et al. Closed-loop transfer enables artificial intelligence to yield chemical knowledge. Nature 633, 351–358 (2024).
Cranmer, K., Brehmer, J. & Louppe, G. The frontier of simulation-based inference. Proc. Natl Acad. Sci. 117, 30055–30062 (2020).
Brehmer, J. Simulation-based inference in particle physics. Nat. Rev. Phys. 3, 305–305 (2021).
Durkan, C., Bekasov, A., Murray, I. & Papamakarios, G. Neural spline flows. In Proc. Advances in Neural Information Processing Systems 32 (Curran Associates, 2019).
Tejero-Cantero, A. et al. sbi: a toolkit for simulation-based inference. J. Open Source Softw. 5, 2505 (2020).
Sulzer, V., Marquis, S. G., Timms, R., Robinson, M. & Chapman, S. J. Python battery mathematical modelling (PyBaMM). J. Open Res. Softw. 9, 14 (2021).
Lake, B. M. & Baroni, M. Human-like systematic generalization through a meta-learning neural network. Nature 623, 115–121 (2023).
Greenberg, D., Nonnenmacher, M. & Macke, J. Automatic posterior transformation for likelihood-free inference. In Proc. 36th International Conference on Machine Learning 2404–2414 (PMLR, 2019).
Zou, H. & Hastie, T. Regularization and variable selection via the elastic net. J. R. Stat. Soc. B Stat. Methodol. 67, 301–320 (2005).
Awad, M. & Khanna, R. in Efficient Learning Machines: Theories, Concepts, and Applications for Engineers and System Designers (eds Awad, M. & Khanna, R.) 67–80 (Apress, 2015).
Pedregosa, F. et al. scikit-learn: machine learning in Python. J. Mach. Learn. Res. 12, 2825–2830 (2011).
Schulz, E., Speekenbrink, M. & Krause, A. A tutorial on Gaussian process regression: modelling, exploring, and exploiting functions. J. Math. Psychol. 85, 1–16 (2018).
Zhang, J. et al. Discovery Learning predicts battery cycle life from minimal experiments. Zenodo https://doi.org/10.5281/zenodo.17654407 (2025).
Acknowledgements
This study was supported by Farasis Energy USA. The large-format lithium-ion pouch cells and all relevant data were provided by Farasis Energy USA. We thank Q. Hu for early-stage data processing. We thank B. Li and W. Xu for early-stage processing and analysis of public datasets. We also thank all referenced articles and their authors for providing the motivation, justification and inspiration for this study.
Author information
Authors and Affiliations
Contributions
J.Z. and Z.S. conceived the study and led the research programme. J.Z. developed the Discovery Learning concepts, including the physics-guided learning approach, the zero-shot learning approach and the active-learning approach. W.J., Y.R. and Q.J. provided experimental data and performed data management. J.Z., Y.R. and Q.J. interpreted the experimental data. J.Z. and Y.Z. developed the algorithms for physics-guided learning and zero-shot learning. J.Z. and B.Y. developed the algorithm for active learning. J.Z. and H.B. performed experimental cost analysis. Y.Z. and J.Z. performed code management. J.Z. reviewed previous literature and wrote the paper. J.Z., Y.Z. and B.Y. prepared the visualization items. J.Z., Y.Z., B.Y. and H.B. prepared the supplementary material. All authors edited and reviewed the manuscript. W.J. and Z.S. supervised the work.
Corresponding authors
Ethics declarations
Competing interests
Y.R., Q.J. and W.J. are employees of Farasis Energy USA, which is a commercial entity. This work was supported in part by funding from Farasis Energy USA provided to the University of Michigan and the National University of Singapore. The other authors declare no competing interests.
Peer review
Peer review information
Nature thanks Kandler Smith and the other, anonymous, reviewer(s) for their contribution to the peer review of this work. Peer reviewer reports are available.
Additional information
Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
Supplementary information
Rights and permissions
Springer Nature or its licensor (e.g. a society or other partner) holds exclusive rights to this article under a publishing agreement with the author(s) or other rightsholder(s); author self-archiving of the accepted manuscript version of this article is solely governed by the terms of such publishing agreement and applicable law.
About this article
Cite this article
Zhang, J., Zhang, Y., Yi, B. et al. Discovery Learning predicts battery cycle life from minimal experiments. Nature 650, 110–115 (2026). https://doi.org/10.1038/s41586-025-09951-7
Received:
Accepted:
Published:
Version of record:
Issue date:
DOI: https://doi.org/10.1038/s41586-025-09951-7
