Prediction of River Runoff Using Fuzzy Theory and Direct Search Optimization Algorithm Coupled Model

Abstract

The correct prediction of daily runoff is considerable importance in flood control. The aim of this study is to evaluate the ability of a combined model, fuzzy inference system–direct search optimization algorithm (FIS-DSOA), to forecast the daily runoff in downstream of Taleghan river using data of rain gauge and hydrometric stations which are located upstream of the river. Initially, the measured daily data (6 years from 2008 to 2013) related to rain gauge and hydrometric stations located in Taleghan river upstream (as input variable) and Siahdasht hydrometric station located in downstream of river (as output variable) were collected. Then the structure of proposed FIS-DSOA model was developed. In the proposed model, the variation range and type of membership functions associated with the input and output variable and the type of rules governing time series were taken as decision variable. Also, the objective function of hybrid model is to minimize root-mean-square error between observed and simulated daily runoff. The forecasts of the FIS-DSOA model are tested using the five statistical indicators (i.e., correlation coefficient, efficiency coefficient, mean absolute error, MSE and index of agreement), and the results are compared with those of the ANNs, FIS and ANFIS models. The comparison results revealed that the FIS-DSOA performs with EFF = 0.83, MAE = 0.75 \({{m}^{3}/{s}}\), MSE = 26.1 \({{m}^{3}/{s}}\) and IOA = 0.92 better than the non-hybrid models in daily runoff discharge prediction. It is concluded that the proposed model significantly improved the accuracy in daily runoff discharge forecasting by combining the capabilities of ANFIS and DSOA models.