Dynamical Modeling of Temperature and Smoke Evolution in a Thermal-Runaway Event of a Large-Format Lithium-ion Battery in a Mine Tunnel

TL;DR

This paper addresses the challenge of predicting temperature and smoke evolution during thermal runaway of large-format lithium-ion batteries in underground tunnels, where high-fidelity CFD is prohibitively expensive. It proposes a reduced-order approach based on grey-box, data-driven state-space models trained on Fire Dynamics Simulator data for two battery capacities (60 Ah and 243 Ah) and multiple airflow conditions, using a node-wise cascading structure to capture spatial coupling. The results show that the ROMs reproduce dominant transient trends in temperature and smoke with reasonable accuracy and provide uncertainty assessments, enabling faster safety analyses. The framework offers a practical, scalable tool for mine safety planning and can be extended to larger network layouts through ensemble CFD-ROM strategies.

Abstract

Large-format lithium-ion batteries (LIBs) provide effective energy storage solutions for high-power equipment used in underground mining operations. They have high Columbic efficiency and minimal heat and emission footprints. However, improper use of LIBs, accidents, or other factors may increase the probability of thermal runaway (TR), a rapid combustion reaction that discharges toxic and flammable substances. Several such incidents have been documented in mines. Since repeatable TR experiments to uncover the transient-state propagation of TR are expensive and hazardous, high-fidelity models are usually developed to mimic the impact of these events. They are resource-intensive and are impractical to develop for many scenarios that could be observed in a mine. Therefore, dynamic models within a reduced-order framework were constructed to represent the transient-state combustion event. Reduced order models (ROMs) reasonably replicate trends in temperature and smoke, showing strong alignment with the ground-truth dataset.

Figures (14)

• Figure 1: Illustration of the structure of the manuscript showing key focus for each section
• Figure 2: Geometry of the study domain: The first illustration depicts the battery alongside a tunnel cross-section with dimensions of 9.0 m x 5.5 m; the second one shows the direction of airflow with the temperature decaying along it
• Figure 3: Summary of employed method: FDS provides the high fidelity ground truth data using the flow volume and boundary conditions to provide the ground-truth data, while SSMs present an efficient pathway to capture the dynamics using the input data and internal states
• Figure 4: HRR of the battery capacities under consideration, including 243 Ah and 60 Ah LIBs, presented as a fraction of the maximum area-averaged heat release rates
• Figure 5: Schematic of the underground mine setup for dynamical modeling purposes. The lithium-ion battery is placed $\sim$ 5.0 m from the mine inlet. Temperature and smoke are monitored at nodes below and downstream of the LIB