Research on EMD-SVR Model with ADF Test for Short-Term Prediction of Ship Motion

Short-term prediction technology is crucial for ships' offshore operations. Combining Empirical Mode Decomposition (EMD) and Support Vector Regression (SVR) method has been confirmed to be an effective approach for short-term prediction of ship's motion. However, the boundary effect of the EMD algorithm leads to significant errors when the EMD-SVR model is used to forecast non-stationary data of ship motion. This paper employs the Augmented Dickey-Fuller (ADF) test algorithm to process non-stationary ship motion data. Data passing the ADF test will be directly predicted using the SVR algorithm, while data failing the ADF test will first undergo EMD processing. According to the experimental results, the rolling and pitching data of the ship can be accurately predicted by the EMD-SVR model with ADF test.