Synergistic Multiphysics Coupling and Machine Learning Approach for High‐fidelity Simulation Modeling of an Electric Sport Utility Vehicle

In this study, a systematic investigation into vehicle‐level energy‐flow measurement, multiphysics coupled simulation, and machine‐learning driven prediction was conducted for an electric sport utility vehicle under diversified ambient temperatures. A high‐resolution experimental platform incorporating multisource measurement units was developed, and comprehensive world light vehicle test cycle experiments were performed at temperatures of −7°C, 23°C, and 35°C. Based on these experiments, a high‐fidelity simulation model was established, seamlessly coupling battery electrochemical‐thermal processes, motor electromagnetic characteristics, transmission system dynamics, and vehicle kinematics. This model accurately reproduced the dynamic energy transfer characteristics across different subsystems, demonstrating excellent conformity to real‐world test results and thus providing a robust virtual platform for parameter optimization and sensitivity analyses. Ensemble surrogate models were subsequently employed to reconstruct intricate nonlinear input–output relationships captured by the multiphysics simulations. Among evaluated methods, extreme gradient boosting with 40 estimators significantly outperformed random forest and adaptive boosting (AdaBoost), exhibiting superior prediction accuracy (RMSE = 1.056, R 2 = 0.959, MAE = 0.773) and negligible residual drift, thereby confirming its excellent generalization capability. Future work will aim to enhance model adaptability to diverse electric‐vehicle architectures and operating conditions through the enrichment of training datasets and the application of transfer‐learning strategies.