Abstract: Objectives Ship fuel consumption prediction is crucial for reducing operational costs and achieving energy-saving emission reduction, to this end, a ship fuel consumption prediction model based on an improved grey wolf algorithm and XGBoost is proposed to provide a reliable basis for ship energy efficiency management decisions. Methods This study develops an XGBoost-based fuel consumption prediction model using actual voyage data. To address the challenges of complex hyperparameter combinations and the inefficiency of manual tuning in the algorithm, we propose an improved Grey Wolf Optimizer for guided hyperparameter optimization of XGBoost. On the basis of retaining the advantages of the traditional algorithm, Tent Chaos mapping is introduced, horizontal and vertical crossover mechanisms as well as boundary checking mechanism are added during the iteration process to train the XGBoost model with the optimal hyperparameter combination. Methods The experimental results demonstrate that the prediction points of the proposed model are densely distributed near the ideal curve on the test set, indicating excellent fitting performance. The evaluation metrics (RMSE, MAE, and R²) all outperform the untuned original XGBoost model and other mainstream models, confirming its strong generalization capability and prediction accuracy. Conclusions The XGBoost model optimized by the improved Grey Wolf Algorithm demonstrates effective performance in ship fuel consumption prediction, providing reliable data support for energy-efficient operations and intelligent navigation of vessels.