Cell temperature prediction in the refrigerant direct cooling thermal management system using artificial neural network

Lithium-ion battery for electric vehicles is highly sensitive to operating temperature. Accurate prediction of individual cell temperature in the battery pack under different conditions is essential for designing thermal management system of battery pack. With the increasing application of artificial intelligence in a variety of disciplines, it appears to be effective to investigate artificial intelligence approaches to evaluate various types of battery thermal management systems. The aim of this paper is to establish an artificial neural network model for prediction cell temperature of a battery pack equipped with the refrigerant direct cooling thermal management system. The inputs of the model are ambient temperature (30, 35, and 40 degrees C), compressor speed (3000, 3500, 4000, 4500, and 5000 rpm), discharge rate (0.5, 1, and 2C), and state of charge (100%-0%). The outputs of the model are the minimum cell temperature, the maximum cell temperature, and the temperature difference between them. The numerical simulation model of the battery pack thermal management system is established, and discretization is applied to the battery pack to obtain temperature variation data for cells under different operating conditions. This data is then used for the training, validation, and testing phase of the neural network model. The results indicate that compared to the Back Propagation neural network, the Elman neural network demonstrated superior predictive performance, effectively capturing the variations in the thermal characteristics of the cells. Under sample conditions, the maximum prediction error of the Elman neural network does not exceed 0.5 degrees C, with maximum MSE and minimum R-squared of 0.0363 and 98.79%, respectively. Under unfamiliar conditions, the maximum prediction error is only 0.94 degrees C, with maximum MSE and minimum R-squared of 0.2275 and 94.48%, respectively. Consequently. this model is suitable for predicting cell temperatures in the refrigerant direct cooling thermal management system for battery packs.