Next-generation thermal runaway monitoring in batteries using real-time gas sensing and AI-driven diagnostics

Abstract

Electric vehicles (EVs) are central to a sustainable future worldwide, with batteries as their core component, including Li-ion, solid-state, NiMH, and Lead-acid batteries, each with its own advantages and disadvantages. However, thermal runaway (TR) in these batteries poses a major safety challenge, hindering large-scale EV adoption. TR is an irreversible series of exothermic reactions that can cause explosions, fires, and the release of toxic gases. Conventional battery management systems (BMS) that rely on voltage, current, and temperature often fail to detect early-stage TR events because they respond slowly to internal degradation. Early-stage TR events release gases like H₂, CO, CO₂, CH₄, and VOCs from electrolyte decomposition, often minutes before temperature increases, voltage drops, or smoke, making gas sensors promising for early detection. This positions gas sensing as a promising early-warning modality for next-generation battery safety systems. Nonetheless, deploying these sensors commercially faces challenges such as cross-sensitivity, drift, complex gas mixtures, and variations in battery design. In this context, artificial intelligence (AI) has become a transformative tool that can improve the accuracy, reliability, and commercial viability of gas-sensor-based TR diagnostics. AI models aid in signal interpretation, sensor drift correction, and integrating gas with electrical, thermal, impedance, and mechanical signals. Machine learning, like DFT studies, speeds up the discovery of new materials for better gas sensors with improved selectivity and stability. This review evaluates TR precursors in EV batteries, focusing on gas signatures and their integration into AI battery management for predictive safety, data fusion, and scalability. Graphical abstract