Abstract
To predict the evolution of wave spectrum in real ocean, a machine leaning framework is developed by training a long short-term memory (LSTM) neural network on a physics-based third generation wave model (Simulating WAve Nearshore, SWAN). Considering the realistic ocean waves are usually mixtures of windsea and swells, the wave spectrum is partitioned using a watershed algorithm, such that the windsea and swells are analysed and predicted separately. Four parameters are selected to capture the wave spectrum of each systems, including the significant wave height Hs, peaked wave period Tp, mean propagation direction Qm and directional spreading width sq. The results demonstrate the machine leaning model can achieve accurate prediction of wave condition, the MAEPs (mean absolute error percentage) of 1-hour prediction are less than 5.9%, 3.3%, 3.5% and 3.3% for Hs, Tp, Qm and σθ respectively, and accurate prediction of wave spectra is achieved. Even for 10-hour prediction, satisfactory results are obtained, e.g. the MAEP of Hs is less than 15.5%. The effects of output size (i.e. prediction duration), input data size (i.e. number of delays), as well as different combinations of input features on predictions of wave condition are examined.