عنوان مقاله [English]
Accurate prediction of the flow in the river is an important element in the management of surface water resources, especially by adopting appropriate measures in the event of flooding, and drought. In fact, ensuring proper and accurate methods for predicting the river flow can be considered as one of the challenges in management approach and water resources engineering. In this paper for Intelligent Modeling, time series of monthly flows of a 26 years period (1985-2011) were used. To achieve the best ANN structure, different structures were compared with those described for the optimization of weight connectivity between different layers of artificial neural networks, genetic algorithms was used.
Learning rule which used in current study is Quick Learning, with advection Tan Axon function, objective function, Mean Square Error and type of Training is N times training. The results show that the best performance is for a condition which monthly precipitation input data, monthly flow, monthly temperatures was used for obtaining next month flow. Also, at each stage that one of the flow data or rainfall data entered as one of the inputs data network performance declined. As result, highest sensitivity is for monthly flow and Temperature has little effect on the estimated flow.
The results of the correlation coefficient 0.84 Indicate high accuracy of artificial neural networks in the estimation of monthly flow in Shor river basin.