Application of artificial neural network for the prediction of thermal runaway in lithium-ion batteries

This paper investigates the electrochemical conditions influencing thermal runaway (TR) in lithium-ion batteries. TR is a critical occurrence marked by an escalation in temperature triggered by exothermic reactions that may lead to fumes, fire, or explosions in lithium-ion batteries. This study develops an artificial neural network (ANN) to predict thermal runaway in a lithium-ion battery pack. Three models were considered: Layer recurrent- NN, Elman- NN, and FF- NN. These models are trained, tested, and verified using data sets obtained from a COMSOL Multiphysics model. The Layer recurrent -NN model is found to perform better because it provided a lower mean squared error (MSE) and root mean squared error (RMSE) of 0.840196 and 0.916622, respectively, because the predicted temperature values are closer to the simulated temperature. Furthermore, the Layer recurrent-NN model was optimised to determine the best network configuration. The optimised configuration consists of 8 hidden neurons, a logistic hidden activation function, and an identity output activation function. The effect of various parameters, such as state of charge, voltage, and current, on the battery temperature was also investigated.