Many-objective optimization for overall performance of an electric sport utility vehicle under multiple temperature conditions based on natural gradient boosting model

In this study, significant efforts have been devoted to optimizing energy consumption and driving range of an electric sport utility vehicle (ESUV). Vehicle energy flow tests were conducted at different temperatures, and integrated simulation models were built and validated. Subsequently, Natural Gradient Boosting (NGBoost) model was constructed based on extensive datasets obtained by parallel simulation framework, and many-objective optimizations of ESUV design parameters were performed. The results show that energy consumption per 100 km under high and low temperatures increases by 19.3 % and 34.9 % compared to room temperature, while effective work of half-axis decreases by 12.9 % and 28 %, suggesting extreme environmental temperature significantly affects ESUV performance. Mean square errors (MSEs) of NGBoost model predictions are less than 0.12, indicating its reliable predictive capability for unknown data. Many-Objective Random Walk Grey Wolf Optimizer (MORW-GWO) algorithm achieves optimal or suboptimal results on 10 test problems, fully demonstrating its efficient convergence and excellent robustness. Battery recovered energy, effective work of half-axis and electricity consumption per 100 km of final optimization scheme are increased by 20.2 %, 9.0 % and 13.0 %, with the largest overall percentage improvement of 17.3 %. These findings can provide data support and directional guidance for optimization designs and performance improvements of ESUVs.