Physics-Constrained Deep Learning for Interpretable Anomaly Detection in Large Battery Packs with Limited Monitoring Granularity
##plugins.themes.bootstrap3.article.main##
##plugins.themes.bootstrap3.article.sidebar##
Abstract
Condition monitoring of large-scale lithium-ion battery packs is often constrained by limited measurement granularity, as operational systems typically rely on aggregated pack-level signals rather than detailed cell-level data. While numerous methods for anomaly detection and health estimation have been developed for individual battery cells, pack-level modeling remains relatively underexplored. Moreover, the impact of reduced monitoring granularity on battery health assessment is still insufficiently understood.
In this work, we present PILSNet, a Physically Interpretable Latent State Network that enables the inference of physically interpretable aging indicators from current-voltage time series. The model combines a convolutional neural network with an equivalent circuit model (ECM), which acts as a constraint to enforce latent variables corresponding to internal resistance and capacity. We apply this hybrid framework to detect abnormal aging behavior in battery packs, achieving both high anomaly detection performance and interpretability of the underlying degradation processes.
Using a physics-based battery simulation framework, we conduct a systematic study of the effect of monitoring data granularity by comparing anomaly detection performance at the cell-level and under aggregated pack-level measurements. The results show that reduced monitoring granularity leads to a significant decrease in anomaly detectability, particularly under realistic scenarios with varying operating conditions and incomplete degradation trajectories. The proposed hybrid model mitigates this performance loss and consistently outperforms purely data-driven and feature engineering-based baselines, especially under constrained data conditions.
In addition to improved detection performance, the inferred latent variables provide direct insight into degradation mechanisms, enabling a pathway from anomaly detection toward fault diagnostics. Beyond battery systems, this work highlights the broader importance of systematically analyzing the relationship between monitoring design and achievable performance in complex systems.
How to Cite
##plugins.themes.bootstrap3.article.details##
battery packs, anomaly detection, data granularity, limited monitoring granularity, condition monitoring, lithium-ion battery pack degradation, physics-informed deep learning, battery degradation, interpretable AI, data scarcity
Bajarunas, K., Baptista, M. L., Goebel, K., & Chao, M. A. (2024). Health index estimation through integration of general knowledge with unsupervised learning. Reliability Engineering & System Safety, 251, 110352.
Bajarunas, K., Navidi, S., Dimitrios, Z., Hu, C., Goebel, K., & Chao, M. A. (2025). Degradation-dynamics-informed learning for future-aware battery health prognostics. Available at SSRN 5928714.
Barreras, J. V., Raj, T., Howey, D. A., & Schaltz, E. (2017). Results of screening over 200 pristine lithium-ion cells. In 2017 IEEE Vehicle Power and Propulsion Conference (VPPC) (pp. 1–6).
Baumhöfer, T., Brühl, M., Rothgang, S., & Sauer, D. U. (2014). Production-caused variation in capacity aging trend and correlation to initial cell performance. Journal of Power Sources, 247, 332–338.
Biggio, L., Bendinelli, T., Kulkarni, C., & Fink, O. (2023). Ageing-aware battery discharge prediction with deep learning. Applied Energy, 346, 121229.
Bosman, H. H., Iacca, G., Tejada, A., Wörtche, H. J., & Liotta, A. (2017). Spatial anomaly detection in sensor networks using neighborhood information. Information Fusion, 33, 41–56.
Che, Y., Deng, Z., Li, P., Tang, X., Khosravinia, K., Lin, X., & Hu, X. (2022). State of health prognostics for series battery packs: A universal deep learning method. Energy, 238, 121857.
Chen, C.-H., Planella, F. B., O’Regan, K., Gastol, D., Widanage, W. D., & Kendrick, E. (2020). Development of experimental techniques for parameterization of multiscale lithium-ion battery models. Journal of The Electrochemical Society, 167(8), 080534.
Guo, F., Couto, L. D., Mulder, G., Trad, K., Hu, G., Capron, O., & Haghverdi, K. (2024). A systematic review of electrochemical model-based lithium-ion battery state estimation in battery management systems. Journal of Energy Storage, 101, 113850.
Guo, F., Xu, K., Zhang, Z., Zhou, H., Chen, G., Hu, J., ... Mo, H. (2025). Battery SOH prediction under different conditions via MBLSTM and iTransformer with anomaly detection and explainability. IEEE Open Journal of the Computer Society.
Kuzhiyil, J. A., Damoulas, T., & Dhammika Widanage, W. (2024). Neural equivalent circuit models: Universal differential equations for battery modelling. Applied Energy, 371, 123692.
Lambert, P., Drummond, R., Ross, J. P., Tredenick, E. C., Howey, D. A., & Duncan, S. R. (2026). Detecting faulty lithium-ion cells in large-scale parallel battery packs using current distributions. Communications Engineering, 5(1), 17.
Li, H., Heleno, M., Zhang, W., Sun, K., Garcia, L. R., & Hong, T. (2025). A cross-dimensional analysis of data-driven short-term load forecasting methods with large-scale smart meter data. Energy and Buildings, 115909.
Liu, S., Li, K., & Yu, J. (2025). Battery pack condition monitoring and characteristic state estimation: Challenges, techniques, and future prospectives. Journal of Energy Storage, 105, 114446.
Lüscher, M. F., Zgraggen, J., Guo, Y., Notaristefano, A., & Huber, L. G. (2024). Data scarcity in fault detection for solar tracking systems: The power of physics-informed artificial intelligence. In PHM Society European Conference (Vol. 8, pp. 8–8).
Machlev, R. (2024). EV battery fault diagnostics and prognostics using deep learning: Review, challenges & opportunities. Journal of Energy Storage, 83, 110614.
Navidi, S., Bajarunas, K., Chao, M. A., & Hu, C. (2025). Forecasting battery capacity for second-life applications using physics-informed recurrent neural networks. eTransportation, 25, 100432.
Qi, Q., Liu, W., Deng, Z., Li, J., Song, Z., & Hu, X. (2024). Battery pack capacity estimation for electric vehicles based on enhanced machine learning and field data. Journal of Energy Chemistry, 92, 605–618.
Seigneur, H. (2022). LCOE reduction through proactively optimized monitoring of PV systems: Final technical report (Tech. Rep.). University of Central Florida, Orlando, FL, United States.
Sulzer, V., Marquis, S. G., Timms, R., Robinson, M., & Chapman, S. J. (2021). Python Battery Mathematical Modelling (PyBaMM). Journal of Open Research Software, 9(1).
Thingvad, M., Calearo, L., Thingvad, A., Viskinde, R., & Marinelli, M. (2020). Characterization of NMC lithium-ion battery degradation for improved online state estimation. In 2020 55th International Universities Power Engineering Conference (UPEC) (pp. 1–6). doi: 10.1109/UPEC49904.2020.9209879
Tranter, T., Timms, R., Sulzer, V., Planella, F., Wiggins, G., Karra, S., ... others. (2022). liionpack: A Python package for simulating packs of batteries with PyBaMM. Journal of Open Source Software, 7(70).
Wang, H., & Zhao, H. (2026). Differentiable physics information neural network with equivalent circuit model constraints for lithium-ion battery parameter identification. Journal of Energy Storage, 155, 121685.
Wang, T.-T., Liu, K.-Y., Peng, H.-J., & Liu, X. (2025). Interpretable machine learning for battery prognosis: Retrospect and prospect. Advanced Energy Materials, 15(48), e03067.
Weng, C., Feng, X., Sun, J., & Peng, H. (2016). State-of-health monitoring of lithium-ion battery modules and packs via incremental capacity peak tracking. Applied Energy, 180, 360–368.
Wu, J., Fang, L., Dong, G., & Lin, M. (2023). State of health estimation for lithium-ion batteries in real-world electric vehicles. Science China Technological Sciences, 66(1), 47–56.
Wüest, M., Plonczak, A., & Goren Huber, L. (2026). Detecting anomalies under data scarcity: Data granularity tradeoffs in large battery packs. In 2026 IEEE Swiss Conference on Data Science and AI (SDS) (pp. 91–98). doi: 10.1109/SDS70563.2026.00020
Xu, W., Mao, R., Han, P., Yuan, N., Li, Y., Guo, Y., & Zhang, H. (2026). Graph neural network modeling of lithium-ion battery capacity with physics-guided features. Journal of Energy Storage, 155, 121627.
Yang, S., Zhang, C., Chen, H., Wang, J., Chen, D., Zhang, L., & Zhang, W. (2024). A hierarchical enhanced data-driven battery pack capacity estimation framework for real-world operating conditions with fewer labeled data. Journal of Energy Chemistry, 91, 417–432.
Yang, X.-G., Leng, Y., Zhang, G., Ge, S., & Wang, C.-Y. (2017). Modeling of lithium plating induced aging of lithium-ion batteries: Transition from linear to nonlinear aging. Journal of Power Sources, 360, 28–40.
Zha, D., Bhat, Z. P., Lai, K.-H., Yang, F., Jiang, Z., Zhong, S., & Hu, X. (2025). Data-centric artificial intelligence: A survey. ACM Computing Surveys, 57(5), 1–42.
Zhalehdoost, A., & Taleai, M. (2025). Unravelling the importance of spatial and temporal resolutions in modeling urban air pollution using a machine learning approach. Scientific Reports, 15(1), 27708.
Zhang, H., Li, Y., Zheng, S., Lu, Z., Gui, X., Xu, W., & Bian, J. (2025). Battery lifetime prediction across diverse ageing conditions with inter-cell deep learning. Nature Machine Intelligence, 7(2), 270–277.
Zhang, L., Xia, B., & Zhang, F. (2024). Adaptive fault detection for lithium-ion battery combining physical model-based observer and BiLSTMNN learning approach. Journal of Energy Storage, 91, 112067.
Zhang, X., Liu, P., Lin, N., Zhang, Z., & Wang, Z. (2023). A novel battery abnormality detection method using interpretable autoencoder. Applied Energy, 330, 120312.
Zhao, D., Zhou, Z., Zhang, P., Zhang, Y., Feng, Z., Yang, Y., & Cao, Y. (2023). Health condition assessment of satellite Li-ion battery pack considering battery inconsistency and pack performance indicators. Journal of Energy Storage, 60, 106604.
This work is licensed under a Creative Commons Attribution 3.0 Unported License.
The Prognostic and Health Management Society advocates open-access to scientific data and uses a Creative Commons license for publishing and distributing any papers. A Creative Commons license does not relinquish the author’s copyright; rather it allows them to share some of their rights with any member of the public under certain conditions whilst enjoying full legal protection. By submitting an article to the International Conference of the Prognostics and Health Management Society, the authors agree to be bound by the associated terms and conditions including the following:
As the author, you retain the copyright to your Work. By submitting your Work, you are granting anybody the right to copy, distribute and transmit your Work and to adapt your Work with proper attribution under the terms of the Creative Commons Attribution 3.0 United States license. You assign rights to the Prognostics and Health Management Society to publish and disseminate your Work through electronic and print media if it is accepted for publication. A license note citing the Creative Commons Attribution 3.0 United States License as shown below needs to be placed in the footnote on the first page of the article.
First Author et al. This is an open-access article distributed under the terms of the Creative Commons Attribution 3.0 United States License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.