A study of the thermal runaway of lithium-ion batteries: A Gaussian Process based global sensitivity analysis

A particular safety issue with Lithium-ion (Li-ion) cells is thermal runaway (TR), which is the exothermic decomposition of cell components creating an uncontrollable temperature rise leading to fires and explosions. The modelling of TR is difficult due to the broad range of cell properties and potential conditions. Understanding the effect that thermo-physical and heat transfer characteristics have on the TR abuse model output is essential to develop more accurate and robust TR models. This study uses global sensitivity analysis (GSA) to investigate the effect of the cell parameters on the outcome of TR events. Using a Gaussian Process (GP) surrogate model to calculate the Sobol’ indices, it is shown that the emissivity value is the dominant thermo-characteristic throughout the overall abuse scenario. Further analysis, investigating three key TR features shows the conductivity coefficient to be the most important with respect to the maximum temperature reached during TR. Results demonstrate that researchers can confidently estimate some thermo-characteristics but require accurate characterisation of the emissivity and conductivity coefficient to ensure robust predictions. Given the importance of battery technology to aid in global de-carbonisation, these findings are key to increasing their safe design and operation. •A global sensitivity analysis investigates the thermal runaway in LiFePO4 cells.•Calculated time-dependent Sobol’ indices using a Gaussian Process surrogate model.•TR onset time during an abuse scenario is dominated by the emissivity value.•The conductivity coefficient has the largest impact on the severity of TR.•Well characterising important variables lead to more robust and accurate TR models.