GA-LSTM-ARMA Neural Network Based on Wavelet Transform for Ship Fuel Consumption Prediction

In order to predict the fuel consumption of ships more accurately, a GA-LSTM-ARMA prediction model based on wavelet transform is proposed. Firstly, the time series of fuel consumption is decomposed using wavelet transform, and the linear series is predicted by ARMA model, the nonlinear series is predicted by LSTM network rolling, then the above two series are stacked as the total time series prediction value. Finally, the GA algorithm is used to optimize the problem that the LSTM is prone to fall into the local optimal solution. The model was tested using the voyage data of the vessel "OCEAN FAVOUR", GA-LSTM-ARMA prediction model shows a significant accuracy improvement over LSTM, achieving a 12.5% error reduction.