Abstract
The broader application of lithium-ion batteries (LIBs) is constrained by safety concerns arising from thermal runaway (TR). Accurate prediction of TR is essential to comprehend its underlying mechanisms, expedite battery design, and enhance safety protocols, thereby significantly promoting the safer use of LIBs. The complex, nonlinear nature of LIB systems presents substantial challenges in TR modeling, stemming from the need to address multiscale simulations, multiphysics coupling, and computing efficiency issues. This paper provides an extensive review and outlook on TR modeling technologies, focusing on recent advances, current challenges, and potential future directions. We begin with an overview of the evolutionary processes and underlying mechanisms of TR from multiscale perspectives, laying the foundation for TR modeling. Following a comprehensive understanding of TR phenomena and mechanisms, we introduce a multiphysics coupling model framework to encapsulate these aspects. Within this framework, we detail four fundamental physics modeling approaches: thermal, electrical, mechanical, and fluid dynamic models, highlighting the primary challenges in developing and integrating these models. To address the intrinsic trade-off between computational accuracy and efficiency, we discuss several promising modeling strategies to accelerate TR simulations and explore the role of AI in advancing next-generation TR models. Last, we discuss challenges related to data availability, model scalability, and safety standards and regulations. •Thermal runaway mechanism is elucidated from multiscale perspectives of electrode, cell, module, and system.•Multiphysics modeling framework is introduced based on thermal, electrical, mechanical, and fluid dynamics models.•Promising modeling strategies for accelerating thermal runaway simulations are outlined and envisioned.•Machine learning can break inherent contradictions between accuracy and efficiency in thermal runaway modeling.•Perspectives guide future thermal runaway model development toward higher accuracy, efficiency and scalability.