From gas generation to intelligent management: a gas-signal-driven framework for thermal runaway early warning in lithium-ion batteries

Thermal runaway (TR) in lithium-ion batteries poses a critical challenge to the safety and reliability of electric vehicles and large-scale energy storage systems. Conventional battery management systems (BMSs), which primarily rely on voltage and temperature signals, frequently exhibit delayed response and limited capability for early-stage fault detection. Gas signals, generated at the early stages of TR, provide rich physicochemical information on internal reactions and offer unique potential for timely warning and fault diagnosis. This review presents a gas-signal-driven early warning framework that systematically integrates mechanism understanding, signal sensing, and intelligent decision-making. The framework is designed to upgrade current BMS architectures without major hardware redesign. Specifically, this review (i) deconstructs multi-stage gas generation mechanisms, characteristic gas fingerprints across different battery chemistries, and their spatiotemporal evolution; (ii) critically evaluates state-of-the-art gas sensing technologies and their applicability to practical battery systems; and (iii) synthesizes intelligent early warning architectures, with an emphasis on multi-sensor fusion and AI-based methods to enhance detection reliability, extend warning lead time, and enable predictive safety interventions. In addition, key engineering challenges-including trace gas detection under realistic conditions, environmental interference, long-term sensor stability, and constraints for edge deployment-are systematically discussed. Future directions toward cloud-edge collaboration are also outlined. Beyond early warning, this infrastructure also holds promise for full-lifecycle management, including state estimation and maintenance scheduling.