THE STUDY OF FEED-FORWARD BACK PROPAGATION ARTIFICIAL NEURAL NETWORK AND RADIAL BASIS FOR PREDICTING THE RIVERS FLOWBehrooz Yaghobil. Saeid Shahanlou and Farshid Salmani
The necessity to predict the flow in the rivers, in managing correctly the water resources for agricultural utilities, drinking water, industries, the input flow currents into the dam reservoirs, organizing the Pvers, the flood alarming systems and etc have always made the hver engineers to designing some models to be prone to predicting the flow with high capabilities and less Fount trror. Therefore, the novel models of Artificial Neural Networks (ANNs) with their capabilities to make models of non-linear phenomenon not only are able to meet such needs without having the various parameters etc but it has also gone beyond its other traditional counterparts, like Regersiom and the Time Sequence and it has also attracted the attention of the water engineers to itself. Thus aerological and hydrometric statistics (rainfall, discharge and transpi-ration) of the 23 wateryear in a monthly time period on the Gamasiab River are used in this paper and also MATLAB software, 7.8 in addition and the neural network is used to made models prone to predict the flow currents. Some various patterns of data %ere considered the network input parameters. In addition, two models of the neural systems, feed-forward back propagation and radial basic are used for modeling. The comparison deducted from the acquired results indicates the high capability of Rbf network in all of the patterns in predicting the flow and also the feed-forward back propagation in the patterns using the dneharF,e input parameters and the upstream stations rainfall had acceptable results.
Enter your contact information below to receive full paper.