Multi-objective optimization method for thermal management control parameters of electric vehicles based on orthogonal analysis and GPR

To reduce the reliance on empirical calibration in electric vehicle thermal management systems, a multi-objective optimization study of control parameters was conducted under high-temperature simultaneous cooling conditions. Control parameter offsets were introduced as decision variables, and three critical offsets were identified through orthogonal experiments and ANOVA. The battery overtemperature time ratio and the cabin supply air temperature deviation time ratio were selected as objectives, with the maximum battery temperature difference constrained below 2.0 °C and the remaining state of charge (SOC) required not to fall below the baseline value. A Gaussian Process Regression (GPR) surrogate model was established, and NSGA-II was employed for multi-objective optimization. Physical analysis shows that parameter significance is governed by dynamic heat loads, refrigerant distribution coupling, and PI feedback compensation. Under the combined NEDC and high-speed driving cycle, the optimized parameters reduced the supply air temperature deviation time ratio by 22.32%, while the maximum battery temperature difference remained stable at 1.8 °C. The remaining SOC increased by only 0.1 percentage point, indicating no observable energy penalty, at the cost of only a marginal 1.02% relative increase in the battery overtemperature time ratio. Validation under the WLTC cycle confirmed the robustness of the optimized parameters within the tested operating conditions. The proposed offline optimization framework can reduce the number of calibration simulations required and provides a theoretical basis for efficient development of integrated thermal management systems incorporating a refrigeration circuit.