Electrode-Resolved Impedance and Thermal Runaway Risk Prediction via Experimental and Physics-Informed Transformer Model
ECS Meeting Abstracts · 07.07.2026 · MA2026-01 (1) · 126-126
DOI: 10.1149/ma2026-011126mtgabs ↗Kurz gefasst
Fachartikel in der Zeitschrift ECS Meeting Abstracts, erschienen am 07.07.2026. Verfasst von Seyed Reza Amini Niaki und 4 weiteren Autor:innen.
Abstract (Originalsprache)
Thermal runaway (TR) in lithium-ion batteries remains a critical safety concern, yet it is difficult to diagnose it early solely from pack-level voltage and temperature. Electrochemical impedance spectroscopy (EIS) encodes rich information on charge-transfer kinetics, transport, and interfacial degradation, but these effects are convolved at the full-cell level and not directly linked to TR risk. In this work, we develop a physics-informed transformer-based neural network that deconvolves full-cell EIS into electrode-resolved impedance and uses it to analyze degradation pathways associated with TR. Experimentally validated based on our own measurements (4680, 24 Ah type batteries), the COMSOL Multiphysics framework 1 couples a pseudo-two-dimensional (P2D) electrochemical model with a thermal model to simulate cell-level and electrode-level impedance over a wide range of operating conditions. Once calibrated to experimental full-cell EIS at multiple temperatures (‒14 °C, 2 °C, 23 °C and 42 °C) and states-of-charge (100%, 80% and 25%), the model is used to generate multiple EIS spectra, while systematically varying safety-critical parameters, such as solid-electrolyte interphase (SEI) layer thickness, negative- and positive-electrode exchange current densities, negative-electrode pore/contact resistance, and a lithium-plating kinetics term active at sub-zero temperature. These generated EIS spectra form the training data for a customized Transformer that receives only full-cell impedance (real and imaginary components), temperature, and state-of-charge, and predicts the corresponding anode and cathode impedances. A feature-tokenization scheme and a shape-matching loss ensure that the model respects both pointwise values and the geometric structure of electrode Nyquist/Bode arcs. The Transformer accurately predicts electrode-level spectra across all parameter sweeps and temperatures, reproducing not only the change in arc magnitude but also the correct shift in characteristic frequencies for variations in SEI thickness, negative- and positive-electrode exchange current densities, negative-electrode pore/contact resistance, and the lithium-plating at −14 °C. By combining the Transformer’s electrode-resolved predictions with the underlying thermal model, we map how specific impedance changes correlate with entry into self-accelerating thermal regimes, defined by rapidly rising temperature and positive feedback between heat generation and reaction kinetics. Conditions that produce enlarged mid-frequency anode arcs and slowed charge transfer (thick SEI, reduced exchange current density, or active plating kinetics) consistently align with TR-prone simulations, whereas spectra dominated by small, high-frequency charge-transfer arcs correspond to thermally benign operation. This integrated experimental-multiphysics–Transformer framework demonstrates how a single EIS measurement can be transformed into a non-intrusive, physics-aware indicator of TR risk for next-generation high-energy battery systems. Acknowledgment Authors gratefully acknowledge financial support from the Department of War (DoW) Office of Naval Research (ONR) and Air Force Office of Scientific Research through the Defense Established Program to Stimulate Competitive Research (DEPSCoR) under award number FA9550-24-1-0164. References [1] B. R. Das Goswami, Y. Abdisobbouhi, H. Du, F. Mashayek, T. A. Kingston, and V. Yurkiv, “Advancing battery safety: Integrating multiphysics and machine learning for thermal runaway prediction in lithium-ion battery module,” J Power Sources , vol. 614, 235015 (2024), doi: 10.1016/j.jpowsour.2024.235015.
Zitieren
APA
Niaki, S. R. A., AbdiSobbouhi, Y., Alahmad, Q., Kingston, T. A., & Yurkiv, V. R. (2026). Electrode-Resolved Impedance and Thermal Runaway Risk Prediction via Experimental and Physics-Informed Transformer Model. ECS Meeting Abstracts, MA2026-01(1), 126-126. https://doi.org/10.1149/ma2026-011126mtgabs
Vancouver
Niaki SRA, AbdiSobbouhi Y, Alahmad Q, Kingston TA, Yurkiv VR. Electrode-Resolved Impedance and Thermal Runaway Risk Prediction via Experimental and Physics-Informed Transformer Model. ECS Meeting Abstracts. 2026;MA2026-01(1):126-126. doi:10.1149/ma2026-011126mtgabs
BibTeX
@article{Niaki2026,
author = {Niaki, Seyed Reza Amini and AbdiSobbouhi, Yasaman and Alahmad, Qusai and Kingston, Todd A. and Yurkiv, Vitaliy R.},
title = {Electrode-Resolved Impedance and Thermal Runaway Risk Prediction via Experimental and Physics-Informed Transformer Model},
journal = {ECS Meeting Abstracts},
year = {2026},
volume = {MA2026-01},
number = {1},
pages = {126--126},
doi = {10.1149/ma2026-011126mtgabs}
}
EUVIMED