Suggested Citation:
Jing, Hao, Shiqi Ou, Zhilong Lv, Haifeng Guo, Andrew Burke, Jingyuan Zhao (2026)

Battery Safety: Mechanisms, Monitoring, and Machine Intelligence

Batteries constitute the foundation of electronic devices and electrified transportation. Nevertheless, aging and sudden faults can precipitate thermal runaway, making battery safety a global concern. This review elucidates failure triggering and evolution from the perspectives of multiphysics coupling and multiscale failure propagation, with emphasis on chemistry-specific heterogeneity in next-generation battery systems. To mitigate these risks, intrinsic-safety materials and structural designs are systematically examined, together with a graded evaluation of their maturity. System-level active protection is further discussed, highlighting the role of cloud-based Battery Management Systems in data governance and cloud-edge collaborative monitoring and control. Building on this architecture, an artificial intelligence-empowered monitoring and control framework is synthesized across four dimensions: (1) perception, which uses multimodal fusion to overcome the limitations of single-variable monitoring and enable holistic mapping of internal states; (2) algorithms, which adopt data-efficient paradigms such as self-supervised learning to address data scarcity in extreme fault scenarios; (3) mechanisms, which integrate physics-informed neural networks and digital twins to enhance interpretability and physical consistency; and (4) deployment, which leverages edge computing and federated learning to enable cloud-edge collaboration and swarm intelligence under privacy constraints. Finally, this review outlines prospects for next-generation safety testing standards, autonomous closed-loop safety management, self-healing technologies, and cross-domain safety management.