2023 Volume 79 Issue 18 Article ID: 23-18061
Significant wave height forecasting is needed to determine whether or not a ship is navigable and offshore construction work is feasible. The wave forecast is usually carried out by using global or regional operation numerical ocean surface wave model. In addition, the artificial neural network has been also used for wave forecasting. However, almost all neural network wave forecasting has been carried out on the specific location and the 2-dimensional wave forecasting has been not so much performed. In this study, we attempted to forecast a two-dimensional wave field using convolutional LSTM, using numerical values of a third-generation wave estimation model SWAN as training data. The results shows that 2 dimensional wave forecasting in the lead time of 1~3 hours are conducted in high precision, while the wave forecast in the lead time over 12 hours were not carried out with high precision.
