Short-term wave elevation forecast by machine learning model based on support vector machine using frequency decomposition 基于滤波分频原理的机器学习方法短期海浪预报与分析

This study presents an improve Support Vector Machine (SVM) model for short-term wave elevation forecast based on the wave tank tests. A new method of frequency decomposition for prediction that combines frequency domain and time domain properties is introduced. Bandpass filtering technique is used to effectively make the sub-frequency decomposition of the original time history data. The corresponding time signal of each sub-frequency band is obtained by combining different filtering parameters, which realizes the combination of time domain and frequency domain and reduces the complexity of the original time history of wave elevation. According to the optimal prediction model, the prediction and superposition synthesis of each sub-frequency band are carried out. The SVM prediction method combined with the idea of frequency decomposition is demonstrated effective, which extends the wave forecast duration.