Thermal characteristic analysis and optimization research of new energy vehicle battery thermal management system based on numerical simulation

To address the issues of excessive temperature rise, uneven temperature distribution, and increased cooling energy consumption in new energy vehicle power batteries under high-rate charging and discharging and complex operating conditions, this paper proposes a multi-physics coupled modeling and optimization method for battery thermal management systems based on numerical simulation. First, an electro-thermal-fluid coupling framework is established, including SOC estimation, an electrical model, a Bernardi heating model, an equivalent thermal network, and a fluid-structure interaction heat transfer model, to describe the battery’s heat generation, solid-state heat conduction, and convective heat transfer processes of the cooling medium. Second, temperature, velocity, and pressure field analyses are conducted for both air-cooled and liquid-cooled structures to reveal the influence mechanisms of flow channel structure, flow distribution, and local heat transfer capacity on battery thermal uniformity. Based on this, nonlinear model predictive control and the NSGA-II multi-objective optimization method are introduced, with the highest temperature, maximum temperature difference, and cooling energy consumption as optimization objectives, to obtain a Pareto compromise solution between thermal safety and energy consumption. Finally, the effectiveness of the model is verified by comparing experimental temperature data with the model prediction results. The results show that the proposed method can effectively capture the battery’s thermal response characteristics and provides a theoretical basis for the structural design and operational optimization of power battery thermal management systems. The numerical results demonstrate a 6.7% reduction in maximum battery temperature and a 15.3% reduction in cooling energy consumption after optimization. The reduced-order model achieves an R2 of 0.987 with an RMSE of 0.38 K and provides an 18.0× computational speed-up compared with the full-order model, while experimental validation achieves an R2 of 0.986 with an RMSE of 0.41 K. The NSGA-II optimization converges at approximately 150 generations with a final hypervolume of 0.965, confirming the effectiveness and computational efficiency of the proposed framework.