[EN] The battery electric vehicle is the leading technology for reducing greenhouse gas emissions using clean and renewable energy. However, concerns due to battery thermal runaway are becoming more severe as the battery energy density increases. Fast-calculation models capable of predicting the heat released during the thermal runaway phenomenon can help to develop safety mechanisms according to the battery chemistry. The current study assesses the battery thermal runaway variability for two different battery chemistries, nickel cobalt aluminium oxides and nickel manganese cobalt oxides, for 3 different states of charge (100%, 80% and 50%), two different battery sizes (18,650 and 21,700), and two different battery health (pristine and aged). The tests are performed in the accelerating rate calorimeter using the heat-wait-seek protocol and repeated 5 times (each battery condition) for statistical analysis of the main thermal runaway parameters. A model using the Arrhenius equation was developed, calibrated, and validated. The model was developed considering 5 steps during temperature evolution to the reliable prediction of thermal runaway characteristics, considering inputs as states of charge, capacity fade (solid electrolyte interface growth), energy density, battery end mass and initial voltage. The experimental tests show that temperature rise rate, when the exothermic is detected, and battery end mass play an important role in the self-heating duration and maximum temperature, respectively, which are key parameters to understanding scattering behaviour. Considering these effects during modelling, the model can forecast the primary features of a thermal runaway, including maximum temperature, onset temperature, and duration of the whole battery thermal runaway process, all within the average difference of no more than 3%. For this reason, the model proposed seems to be a suitable tool for battery safety mechanism design as it considers the state of charge, energy density and ageing effects.

The authors acknowledge the Vicerrectorado de investigacion de la Universitat Politecnica de Valencia for supporting this research through Programa de Ayudas de Investigacion y desarrollo (PAID-01-22). This research is part of the projects TED2021-132220B-C21 and TED2021-130488 A-I00, funded by the MCIN/AEI/10.13039/501100011033 and the European Union "NextGenerationEU"/PRTR. This research is part of the project PID2021-124696OB-C21, funded by Ministerio de Ciencia e Innovacion, Agencia Estatal de Investigacion and FEDER (MCIN/AEI/1 0.13039/501100011033/FEDER, UE).