USING AN INTEGRATED FUZZY INFERENCE SYSTEM AND ARTIFICIAL NEURAL NETWORK TO FORECAST DAILY DISCHARGE

Given the nonlinearity and uncertainty in the rainfall-runoff process, estimating or predicting hydrologic data often encounters tremendous difficulty. This study applied fuzzy theory to create a daily flow forecasting model. To improve the time-consuming definition process of membership function, which is usually concluded by a trial-and-error approach, this study designated the membership function by artificial neural network (ANN) with either a supervised or unsupervised learning procedure. The supervised learning was processed by the adaptive network based fuzzy inference system (ANFIS), while the unsupervised learning was created by fuzzy and self-organizing map (SOMFIS). The results indicate that the ANFIS method under increment flow data could provide more precise results for daily flow forecasting.