Significant Wave Height Prediction Based on MSFD Neural Network

Abstract

Due to the complicated behavior of the ocean wave, significant wave height (SWH) prediction is a difficult field in physical oceanography. In this paper, a novel neural network model, based on multiple sine functions decomposition (MSFD), is exploited to achieve the prediction of SWH. Different from traditional models built on physical processes of wave generation and dissipation, the method presented in this paper predicts and analyzes SWH from a mathematical statistical perspective. In particular, the variation rules of the SWH are learned by decomposing the mapping from time to SWH into a plurality of sine functions, and then the new data are predicted by linear combination of these sine functions. Correlation analysis and error between the forecast data and the actual data indicate that the MSFD neural network performs well in predicting SWH data.