Analysis of the Effect of Thermal Runaway Initiation Conditions on the Severity of Thermal Runaway—Numerical Simulation and Machine Learning Study

The main goal of this work is to understand the effect of thermal runaway initiation conditions on the severity of thermal runaway (TR) of Graphite—NMC (111) cells. A coupled electrical-thermal model is developed, which includes the initial energy input, the chemical decomposition processes of the anode, cathode and the electrical energy released by an internal short circuit. 780 different thermal runaway events are simulated and the output is analysed by machine learning techniques such as principal component analysis and clustering. It was found that TR events form 5 clusters between no thermal runaway and severe thermal runaway. Sensitivity analysis is applied on the 39 input invariants and the triggering energy input, resistance ratio, the heat convection coefficient, the ratio of activation energy of oxygen liberation and electrolyte evaporation are found to be the most important parameters. The later one determines the amount of electrolyte combustion. The probability of thermal runaway is calculated taking into account the most important parameters and their interactions. Finally, a combination of initiation parameters is suggested, which most likely results in a repeatable and reproducible outcome.