Predictive modeling of discharge of flow in compound open channel using radial basis neural network

Abstract

Predicting the flow discharge in open channel is the main parameters in the flood management. The concept of the compound open channel is the accurate approach for modeling the natural streams. Several ways as analytical approaches and artificial intelligence methods have been proposed for predicting the discharge in rivers in term of compound open channel concepts. In this paper the single channel method (SCM), coherence method (COHM), and divided channel method (DCM) as common analytical approaches were used to predict the discharge in the compound open channel and in follow to achieve more accuracy in flow discharge prediction the radian basis neural network (RBF) was developed. The performance of RBF was compared with other types of transfer function governed on neurons of neural network. The results showed that the DCM with horizontal separated boundary among the subsections with correlation of determination (R2 = 0.76) is accurate through the analytical approaches. Assessing the results of the MLP model showed that this model with (R2 = 0.95) is a bit more accurate than the RBF (R2 = 0.85) and analytical approaches.