2021 Volume 77 Issue 2 Pages I_751-I_756
In this study, the prediction of significant wave heights and periods using a neural network including LSTM layer was performed for five ports located on the Sea of Japan. The prediction of oceanic wave by using LSTM can contribute to determine whether maritime construction could be performed or not. In order to conduct the high-precision prediction, meteorological data (wind speed, sea level pressure, temperature, and relative humidity) from GPV-MSM were used as input data and their performances were evaluated considering different combinations of input parameters and prediction length. As a result, accuracy of the wave prediction is improved by including the pressure and temperature in addition to the wind speed term apparently influencing the development of oceanic waves. As a summary, it can be said that it is possible to predict significant wave heights and periods 2~24 hours ahead at several locations by using LSTM.
