Computer Simulation and Predictive Model-Based Coolant Flow Optimization for a Hybrid Battery Cooling System Using High-Capacity LiFePO4 Batteries

Thermal management of the battery in electric vehicles is a major problem, as higher temperature or improper thermal distribution will shorten the service life, the safety and the performance of the battery. This paper investigates the hybrid LiFePO4 8-cells battery thermal management system using liquid cooling and PCM cooling by CFD simulation. Three configurations of the channel in the liquid cooling part are used, namely horizontal channel, sinusoidal channel and serpentine channel, and transient fast-charging process is applied. At the same time, three coolants are studied, Water-Glycol, Al2O3 and ZnO-H2O nanofluids to investigate the effect on cooling performance. In the hybrid system, the PCM is paraffin wax which can absorb latent heat during phase transition, with latent heat capacity as 210kJ/kg. Coolant flow rates of 0.2 m/s to 1.0 m/s are studied to find optimal velocity. Results indicate that the serpentine channel performed best due to sufficient mixing of coolant, and superior convective heat transfer coefficient. Among the three coolants, ZnO-H2O nanofluid has best cooling performance and reduced maximum battery temperature down to around 302.6K. The hybrid system exhibited better cooling performance than either liquid cooling or PCM cooling alone. Maximum battery temperature and non-uniformity was reduced to 303.9K respectively in the hybrid system. Moreover, for the coolants flow velocities the results indicate that cooling performance improves with increasing coolant velocity; however, diminishing thermal benefits are observed beyond approximately 0.7 m/s. To improve thermal management under varying battery operating conditions, a polynomial-regression-based surrogate model was developed using CFD-generated data and coupled with a Genetic Algorithm optimization framework. The developed model successfully predicted the coolant velocity required for different battery heat generation rates with a coefficient of determination (R²) of 0.960 and a root mean square error (RMSE) of 0.027. The results show that the required coolant velocity increases from approximately 0.10 m/s to 0.63 m/s as the volumetric heat generation rate increases from 5000 W/m³ to 50000 W/m³. The proposed framework provides a computationally efficient approach for predicting coolant velocity requirements under varying battery thermal loads and supports the development of intelligent battery thermal management systems of high-capacity electric vehicle battery systems