Wave Prediction in Korea Waters Based on ML
Waves are a complex phenomenon that occurs in marine and coastal areas and affects many areas, including maritime transportation and, the stability of offshore structures. In particular, accurate wave prediction is essential for ensuring the safety of ships and offshore operations at sea, and resource management. In this study, we use machine learning methods of Recurrent Neural Network (RNN), Long Short Term Memory (LSTM), and Gated Recurrent Unit (GRU), which are specialized in nonlinear data processing, to predict the wave of the East Sea, West Sea, South Sea, and Jeju Coastal Sea. We also evaluate and compare the performance of machine learning models and present an optimized algorithm for each sea area. We use oceanographic buoy data from 2012 through 2021 provided by the Korea Hydrographic and Oceanographic Agency (KHOA). To find the optimal parameters, we vary the size of the window size that the model is trained and evaluate the performance by comparing the Mean Absolute Error (MAE) between the actual and predicted values. The results show that GRU has the best performance and differences between the actual and prediction are within 0.161 m of significant wave height and 0.491 s of significant wave period.
DOI : https://doi.org/10.26748/KSOE.2023.040
