Evaluating the Effects of Different Input Signals on Efficiency Coefficients of Artificial Neural Network Models for Intelligent Estimation of Flood Hydrographs

The estimation of flood hydrograph characteristics in natural rivers is of hydrologists interests. In this paper, the ability of neural networks for estimation of flood hydrograph to Shirindareh reservoir dam in Khorasan province is evaluated. Therefore, all flood hydrograph events of hydrometery station in upstream were collected and normalized. It is notable that the flood hydrometery was estimated 2, 3, 4 and 5 hours earlier using the flood discharges at 2, 3, 4 and 5 previous hours as model inputs respectively. In each pattern, four signals were selected for considering the effect of number of inputs for estimation of flood hydrograph. The results show that by increasing the estimation lag time, the accuracy of results decrease and in given lag time, by increasing the number of input, the accuracy of results increase. The results show that the amount of efficiency coefficients, which is the representation of goodness of flood hydrograph modeling,is increased from 0.79 for signal one to 0.91 for signal four.