Tidal Level Short-Term Prediction using Back-Propagating Artificial Neural Network (BP-ANN)

Abstract

This research paper presents a back-propagating artificial neural network (BPANN) model for short-term prediction of tidal levels in Penang, Malaysia. The model is trained using historical tidal data and meteorological parameters such as tides height within 3 hours, tides coefficient, and tides cycles for four months data were taken from website in May, June, July and August. The study aims to develop an accurate and efficient prediction model to aid in tidal energy management and forecasting for simulating tidal level value that closely matches the actual tide from the input-output relationships in the short-term tidal records through the unknown parameters determine by ANN. The prediction determination used modelling with artificial neural network (ANN) with the back-propagation method. The optimal predictions using ANN were obtained by conducting five input layers, five hidden layers, and one output layer (5-5-1). The results end with mean absolute percentages error (MAPE) for May, June, July and August was 1.76, 0.39, 0.26 and 0.07 respectively. ANN proved very effective in predictions tidal level.