Artificial Neural Network for Forecasting Wave Heights along a Ship's Route during Hurricanes

Abstract

A data-driven prediction model using numerical solutions is proposed for forecasting wave heights along shipping routes during hurricanes. The developed model can be used to determine the wave heights on a ship’s trajectory, considering a short time step of a ship’s operation. This research used an artificial neural network (ANN) multilayer perceptron model (ANN-based) to build a data-driven prediction model. A quadtree-adaptive model was used as the numerical simulation–based model (NUM-based). The proposed NUM-ANN model is an ANN-based prediction model that incorporates precomputed numerical solutions to determine the wave heights at sample points on the shipping line where buoy measures are absent. The NUM-ANN model is highly efficient because the input–output patterns used to formulate it can be generated in advance through numerical models. A shipping line through the Caribbean Sea and the Gulf of Mexico was used for simulation. The 2005 Category 5 hurricanes Katrina and Rita were used for testing. Three buoys and three sample points on the ship trajectory were applied for modeling the wave heights. The results revealed that (1) for shipping-line buoys, the predictions made using the NUM-based and ANN-based models are satisfactorily consistent with the observed data; and (2) for the sample points, the predictions made using the NUM-ANN model are highly consistent with simulations made using the NUM-based model. Therefore, ANN-based prediction models can be regarded as reliable, and the NUM-ANN model can be effectively used in the real-time forecast of wave heights.