A Neural Network-Based Regression Study for a Hybrid Battery Thermal Management System under Fast Charging

Fast charging is significant for the driving convenience of an electric vehicle (EV). However, this technology causes lithium (Li)-ion batteries' massive heat generation under such severe current rates. To ensure the thermal performance and lifespan of a Li-ion battery module under fast charging, an artificial neural network (ANN) regression method is proposed for a hybrid phase change material (PCM)-liquid coolant-based battery thermal management system (BTMS) design. Two ANN regres-sion models are trained based on experimental data considering two targets: maximum temperature (Tmax) and temperature standard deviation (TSD) of the hybrid cooling-based battery module. The regression accuracy reaches 99.942% and 99.507%, respectively. Four sets of experimental data are employed to validate the reliability of this method, and the cooling effect (Tmax and TSD) of the hybrid BTMS are predicted using the trained ANN regression models. Comparison results indicate that the deviations between the predicted value and the experimental value are acceptable, which prove the accuracy of the ANN regression models. This proposed method combines regression modelling with experimental tests to achieve the desired design efficiency and control, which can be utilized for efficient BTMS design, especially with more complex factors such as the future fast -charging requirements.

Automated summary

This study proposes a hybrid phase change material-liquid coolant-based battery thermal management system (BTMS) design using an artificial neural network (ANN) regression method to ensure the thermal performance and lifespan of a Li-ion battery module under fast charging. The accuracy of the regression models is validated through experimental data, and the predicted cooling effect matches well with the experimental results, indicating the accuracy and reliability of the ANN regression models.