Tidal Level Forecasting during Typhoon Surge Using Functional and Sequential Learning Neural Networks
This note presents the application of the functional network (FN) and the sequential learning neural network (SLNN) for accurate prediction of tides during surge using short-term observation. Based on 34-day observed data, the proposed functional network model can predict the time series data of hourly tides directly, using an efficient learning process that minimizes the error. In the functional network, a simple equation in the form of a finite-difference equation is derived, using the tidal levels at two previous time steps. The sequential learning neural network uses one hidden neuron to predict the current tidal level. Hourly tidal data for the Typhoon Herb, measured at Taichung Harbor along the Taiwan coastal region, is used for testing the capacity of the functional network and sequential neural network models. Numerical results demonstrate that the proposed models can predict the tidal level during typhoon surge with a high correlation coefficient, based on 1-month hourly data.
Journal of Waterway, Port, Coastal & Ocean Engineering, SCE
Civil Geotechnical Engineering