A fuel consumption prediction model for wind-assisted ship based on operational data

• Features influencing wind-assisted ship energy efficiency is explored based on data characteristic analysis. • A GA-CNN-LSTM-WBM-based fuel consumption prediction model for wind-assisted ship is proposed. • The distribution of prediction errors and the robustness in case of data perturbation are discussed. • The established prediction model can accurately predict the fuel consumption of the wind-assisted ship under various conditions. • The established prediction model and method lay solid foundation for the energy efficiency optimization. Establishing an accurate ship fuel consumption (SFC) prediction model is significant for achieving ship energy efficiency optimization and enhancing green transitions of the shipping industry. A GA-CNN-LSTM-WBM-based SFC prediction model for the wind-assisted ship is constructed based on operational data. Firstly, appropriate input features are selected by analyzing characteristics of ship operational data and navigational environmental data. Subsequently, a GA-CNN-LSTM-WBM-based SFC prediction model is established by combining the mechanism-based model with the data-driven model. Two case studies demonstrate that the prediction error of the proposed SFC prediction model is less than 0.77% and 2.25%, which is much more accurate than other prediction models. In addition, the proposed method can effectively enhance the prediction accuracy and robustness of the fuel consumption model for the wind-assisted ship, thus providing a scientific basis for the energy efficiency optimization and emission reduction of the wind-assisted ship.