Muhammad Shahid
, Xia Chengjun, Liu Yicheng
Based on the pressing need for fast, reliable estimates of capacity and State‑of‑Health (SOH) in electric‑vehicle lithium‑ion batteries, this study set out to test whether Kolmogorov‑Arnold Networks (KANs) can outperform the multilayer perceptron (MLP) models that dominate current battery‑management research. Through the NASA AMES PCoE dataset—which tracks four cells (B0005, B0006, B0007 and B0018) from initial charge to end‑of‑life—we trained shallow and deep KAN architectures alongside depth‑matched MLP baselines. Each model received 24 historic cycles and was asked to forecast the future 10 cycles. KANs performance was gauged with mean‑absolute error (MAE) and root‑mean‑square error (RMSE). The experiments show a consistent edge for KANs. On battery B0005, the deep KAN cut capacity‑forecast MAE to 0.014 and RMSE to 0.015, versus 0.023 and 0.030 for the deep MLP. Across all four cells, the deep KAN lowered capacity MAE by an average of 36 % and SOH RMSE by 30%. Shallow KANs also surpassed their MLP counterparts, though by smaller margins, confirming that the spline-based adaptive links inherent to KANs not merely by increasing depth but also in driving the improvement. Combined, these results demonstrate that KANs deliver more accurate, more data‑efficient and easier‑to‑interpret forecasts than conventional MLPs. The specific gains—up to 0.009 absolute MAE and 0.015 absolute RMSE on individual batteries—suggest that replacing MLP blocks with KANs can immediately enhance real‑time battery‑management systems without inflating model size or computational cost.
Muhammad Shahid Xia Chengjun Liu Yicheng “Li-Ion Battery Capacity and State-of-Health Prediction through Novel Kolmogorov-A Vol. 12 Issue 05 PP. 66-72 May 2025. https://doi.org/10.34259/ijew.25.12056672.
[1] M. Muratori et al., “The rise of electric vehicles—2020 status and future expectations,” Prog. Energy, vol. 3, no. 2, p. 22002, 2021.
[2] E. Fan et al., “Sustainable recycling technology for Li-ion batteries and beyond: challenges and future prospects,” Chem. Rev., vol. 120, no. 14, pp. 7020–7063, 2020.
[3] P. H. Camargos, P. H. J. dos Santos, I. R. dos Santos, G. S. Ribeiro, and R. E. Caetano, “Perspectives on Li‐ion battery categories for electric vehicle applications: a review of state of the art,” Int. J. Energy Res., vol. 46, no. 13, pp. 19258–19268, 2022.
[4] J. Urquizo and P. Singh, “A review of health estimation methods for Lithium-ion batteries in Electric Vehicles and their relevance for Battery Energy Storage Systems,” J. Energy Storage, vol. 73, p. 109194, 2023.
[5] S. BS, S. Hampannavar, and B. Bairwa, “Applications of battery management system (bms) in sustainable transportation: A comprehensive approach from battery modeling to battery integration to the power grid,” World Electr. Veh. J., vol. 13, no. 5, p. 80, 2022.
[6] S. Shen, B. Liu, K. Zhang, and S. Ci, “Toward fast and accurate SOH prediction for lithium-ion batteries,” IEEE Trans. Energy Convers., vol. 36, no. 3, pp. 2036–2046, 2021.
[7] J. Li, K. Adewuyi, and others, “A single particle model with chemical/mechanical degradation physics for lithium-ion battery State of Health (SOH) estimation,” Appl. Energy, 2018.
[8] O. Demirci, S. Taskin, E. Schaltz, and B. A. Demirci, “Review of battery state estimation methods for electric vehicles-Part II: SOH estimation,” J. Energy Storage, vol. 96, p. 112703, 2024
