Thermal Runaway in Batteries: A Database-Driven Literature Review and Exploratory Statistical Analysis

Thermal runaway (TR) in batteries remains a key safety challenge, yet its prediction is hindered by strongly coupled physics and many interdependent influencing factors. This review bridges the gap between mechanistic TR overviews and narrowly scoped experimental studies by conducting a broad database-driven review of published TR experiments. Therefore, the largest publicly available TR database to date is curated. It comprises 1703 tests from 257 papers and 203 variables describing cell properties, test conditions, and TR outcomes. Descriptive and pairwise inferential methods are applied to identify recurring patterns reported across the literature and to enable structured description of observed trends. Cathode chemistry, specific energy, and state of charge (SOC) emerge as the key associates of characteristic TR temperatures, with oxygen release from nickel-rich cathodes significantly amplifying TR severity. Aging-related effects strongly depend on the specific aging history and remain insufficiently characterized. Relative mass loss can reach 90% and is linked to the severity of TR reactions and the associated gas generation. On average, vent gas volume scales at 1.7 L/Ah, but capacity-normalized volume varies significantly with cell chemistry and SOC. H2, CO, and CO2 dominate vent gas compositions, with dependence on chemistry, SOC, and overall explosivity, while toxic and condensable species are clearly under-reported. The influence of abuse type and test setup on measured TR characteristics is highlighted, and emerging battery technologies are discussed. The database and derived trends provide a basis for benchmarking cell safety, informing pack-level design and modeling, suggesting future research directions, and supporting the development of standardized TR test protocols.