Deep-Learning Assisted Prediction of Lithium-Ion Battery Thermal Runaway in Subsurface Mining Conditions Based on Numerical Data - SME Annual Meeting 2026

Using battery-powered equipment in underground mining can improve safety, advance automation, and reduce greenhouse emissions. In this context, rechargeable lithium- ion batteries (LIBs) have a competitive edge due to their high Coulombic efficiency, superior energy density, and extended cycling stability. However, several conditions must be met for their safe operation. Their failure can cause thermal runaway (TR), which poses a significant safety risk in underground mines. Despite studies having investigated this phenomenon using physics-based models, limited research has used deep learning to predict its implications. This study aims to bridge this gap by characterizing LIB TR under dynamic airflow through simulations and predicting temperatures using two of the most common recurrent neural networks: long short-term memory (LSTM) and gated recurrent units (GRU).