Neural network with multi-trend simulating transfer function for forecasting typhoon wave

Abstract

This study develops an NN typhoon wave model to accurately and efficiently calculate wave heights at a point of interest. Multi-trend simulating transfer functions were first introduced to exemplify the relationship between wave heights and each conceivable input factor by regressive fitting. The proposed NN-MT model can accurately forecast wave peak with an error of less 1.2 m and with time delay within 3 h and can be extended to cover the station besides the original station of interest. (C) 2005 Elsevier Ltd. All rights reserved.