FUNDAMENTAL STUDY ON TWO-DIMENSIONAL WAVE PREDICTION USING DEEP GENERATIVE MODEL

Abstract

 Wave predicting is in high demand for process control in the implementation of offshore construction work, such as the construction of offshore wind power generation facilities. A number of deep learning models have been developed in recent years for wave prediction. However, many of these models are pinpoint wave predictions at NOWPHAS. When applied to construction work, those models cannot be used when the work is far from NOWPHAS observation points or when the work vessels or materials are moved long distances. When predicting multiple locations using models from previous studies, it is necessary to create the model for each location, which is labor-intensive. In this study, we developed the two-dimensional wave prediction method using the deep generative model, which is a type of deep learning, with the aim of accurately predicting waves at multiple locations with a single model.