Regional ocean wave height prediction using sequential learning neural networks

Abstract

Wave height prediction is a critical factor in the efficient operation of many offshore and coastal engineering activities. The classical numerical solutions to this problem, based on the wave energy-balance equation, involves complex implementation with higher computational powers. In recent years, machine learning approaches are being widely used for the prediction of wave heights. However, these approaches involve batch learning algorithms that are not well-equipped to address the demands of continuously changing data stream. In this paper, we conduct a study to predict the daily wave heights in different geographical regions using sequential learning algorithms, namely the Minimal Resource Allocation Network (MRAN) and the Growing and Pruning Radial Basis Function (GAP-RBF) network. The study is conducted using data collected from 13 stations across three geographically distinct regions, viz., the Gulf of Mexico, the Korean region and the UK region, for the period between Jan 1, 2011 and Aug 30, 2015. The data is chosen such that the study covers a wide range of geographical terrains and locations, a wide range of wind speed and wave heights. We compare the performance of MRAN and GAP-RBF with Support Vector Regression (SVR) and Extreme Learning Machine (ELM). The performance study results show that the MRAN and GAP-RBF outperform the SVR and ELM with minimal network resources, in the daily wave height prediction. They also predict the significant wave heights accurately. Performance comparison between MRAN and GAP-RBF shows that MRAN outperforms GAP-RBF with minimal architecture.