Modeling liquid immersion-cooling battery thermal management system and optimization via machine learning

An efficient battery thermal management system (BTMS) is essential to ensure the optimal performance and safe operation of lithium-ion batteries. This study proposed a BTMS that submerged 10 large-format prismatic cells in a dielectric liquid. First, we compared the performance of flow dielectric immersion cooling (FIC) to immersion cooling with no inlet flow (ICNF). The FIC demonstrated a substantial improvement, reducing the maximum temperature by 8.3% compared to ICNF. The investigation revealed that various factors, including battery spacing, vertical spacing, inlet velocity, and the number of inlets, significantly impact battery pack performance and power consumption. To optimize the proposed BTMS, an Artificial Neural Network (ANN) model was employed to train the numerical database of 8080 cases and identify the optimal combination that balances FIC BTMS cooling efficiency and energy consumption. After the machine learning prediction, the maximum temperature and temperature difference decreased by 7.7% and 86%, respectively, compared to the base case with the same pumping power consumption. These findings can deepen our understanding of battery immersion cooling technology and offer novel insights for BTMS optimization via machine learning methods.