Temperature Prediction of Lithium-Ion Batteries Based on Physical Information and Deep Neural Network

Accurately predicting the temperature of lithium-ion batteries is a key technology for battery management systems. A deep neural network is constructed for temperature prediction of lithium-ion batteries based on their dynamic as well as time-dependent characteristics. The model can extract the potential high-dimension features of the data and appropriately reduce their dimensionality to reduce the model complexity while capturing the long-term dependence of temperature through the layer of long short-term memory cells. In addition, the heat generation rate is calculated in real-time through the open circuit voltage, terminal voltage and current of the lithium-ion battery, thus providing additional physical information input to the deep neural network. The results show that the method has better temperature prediction performance compared to other methods.