Wave estimation within a port using a fully nonlinear Boussinesq wave model and artificial neural networks

Abstract

An artificial neural network is developed to estimate the wave conditions within a port efficiently. Wave motions within the harbor subject to incident waves with varying combinations of offshore parameters (i.e., significant wave height, peak period, and peak direction) are simulated using the Boussinesq-type wave model to obtain a pre-calculated dataset. The outputs of the network are significant wave height and low-frequency wave height within the port. The neural network is trained on the dataset to ensure that efficient estimations for unknown cases with incident parameters similar to those in the pre-calculated cases can be provided. This is done to avoid the high computational cost of the wave model during real-time prediction. A k-fold cross-validation method is utilized to determine the optimum network among a variety of architectures. As a first verification step, the statistical results show that the network performs well when estimating unknown cases that are not included in the training process. Subsequently, the network is successfully validated against the observed in-situ data, proving the reliability of both the wave model and neural networks. Once the forecasting of offshore wave parameters is provided, the wave conditions within the port can be estimated quickly to realize the real-time prediction.