A ship motion forecasting approach based on empirical mode decomposition method hybrid deep learning network and quantum butterfly optimization algorithm

Ship motion (SHM) forecasting value is an important parameter for ship navigation and operation. However, due to the coupling effect of wind, wave, and current, its time series has strong nonlinear characteristics, so it is a great challenge to obtain accurate forecasting results. Therefore, considering the strong nonlinear of SHM time series, firstly, this paper decomposes the original time series into multiple intrinsic mode functions (IMF) using empirical mode decomposition (EMD) technology and then establishes a hybrid deep learning network for each IMF based on convolutional neural network (CNN) and gated recurrent unit (GRU) according to the characteristics of SHM time series. On this basis, the EMD-CNN-GRU (ECG) hybrid forecasting model of SHM is constructed by integrating a component forecasting model. Secondly, considering the difficulty of hyper-parameters selection of ECG model, this paper improves the butterfly optimization algorithm (BOA) based on quantum theory, designs the quantum coding rules of butterfly spatial position, establishes the optimization process of butterfly algorithm based on quantum coding, and then proposes the quantum butterfly optimization algorithm (QBOA). Finally, a hybrid forecasting approach integrating ECG and QBOA is proposed, namely ECG & QBOA. To evaluate the feasibility and performance of the proposed approach. A prediction experiment was carried out with the SHM data of a real ship. The results indicate that, compared with the other comparison models selected in this paper, ECG-based models have significant higher forecasting accuracy (with MAPE values of 10.86% and 12.69% in two experiments, respectively, and with significant accuracy improvement of at least 10% than other compared models), and the QBOA has obtained more appropriate hyper-parameters combination of ECG model.