Published October 6, 2026
| Version v1
Publication
Open
Thermal runaway in large format semi-solid-state cells: Analysis of abuse test results and machine-learning-based core temperature prediction
Authors/Creators
Description
This research paper investigates the behaviour and thermal runaway (TR) conditions of large format semi-solid-state battery (SSSB) cells under several abuse tests like forced discharge test, overcharge test, external overheating test, external short-circuit test, nail penetration test, and crush test using the IEC 62660–2/3 (2022) standards and conditions beyond the IEC 62660–2/3 (2022) standard. Based on the analysis of the abuse test results and the rating from the hazard severity level, it was revealed that SSSB cells exhibit improved thermal stability when compared to traditional lithium-ion battery cells due to the Lithium Phosphorous Sulphide coating of the Nickel-rich cathode which delays the release of oxygen. Furthermore, machine learning (ML) models (Random Forest, Extreme Gradient Boost, and Multiple Linear Regression) were trained to predict the core temperature for operating temperature range not more than 50 °C. For clarity, the predicted core temperature is the predicted center surface temperature. This core temperature is practically relevant because the tab regions are more accessible for sensing in battery modules/systems than placing additional sensors at the center region of every pouch cell. The random forest model best predicted the core center surface temperature of the SSSB cell with a root mean square error of 2.9 °C when compared to the experimental center surface temperature (T5), demonstrating the robustness of the model within the tested experimental domain. The results indicate that tabs' temperature measurements can serve as effective predictive indicators for the thermal state of the central surface region, offering a potential pathway for non-invasive thermal monitoring t and safety enhancement in SSSB systems.