A Study of Watershed Hydrological Models Constructed by Hybrid Artificial Neural Networks

• Main Authors: Jin-Ming Liang, 梁晉銘
• Other Authors: Fi-John Chang
• Format: Others
• Language: zh-TW
• Published: 2002
• Online Access: http://ndltd.ncl.edu.tw/handle/89698403680143564559

博士 === 國立臺灣大學 === 生物環境系統工程學系暨研究所 === 90 === A radial basis function neural network (RBFNN) is proposed to develop a hydrological model for flood forecasting and water stage data reconstruction. For faster training speed, the RBFNN employs a hybrid two-stage learning scheme. During the first stage, unsupervised learning, fuzzy min-max clustering is introduced to determine the characteristics of the nonlinear radial basis functions. In the second stage, supervised learning, multivariate linear regression is used to determine the weights between the hidden and output layers. The rainfall-runoff relation can be considered as a linear combination of some nonlinear radial basis functions. Rainfall and runoff events of the Lanyoung River and the Tanshui River under tidal effect collected during typhoons are used to train, validate and test the network. In addition, the RBFNN is also used for the water stage data reconstruction in the Kaoping River. The results show that the RBFNN can be considered as a suitable technique for predicting water stage and flood flow, and reconstructing water stage data.