O. BOZORG HADDAD - Iran university of science and Technology, Tehran, Iran
F. SHARIFI - Iran university of science and Technology, Tehran, Iran
S. ALIMOHAMMADI - Water Engineering Department, Shahid Abbaspoor University; Water Resources Expert, Moshanir Power Engineering Consultant, Tehran

Providing stream flow forecasting models is one of the most important problems in water resources planning and management. Traditional models in this field have been developed in the form of regression models, and time series models. Nowadays, Artificial Neural Networks (ANNs) are also used besides the classic methods. In this study, the ability of ANNs in stream flow forecasting has assessed. For this purpose, the monthly Inflow of Karoon 5 reservoir in Iran is selected. A 43-year monthly time series of inflow is available that it has been used in modeling process. 80% of data were used to develop the models and the rest of data were utilized to test the models. A Multi Layer Perceptron (MLP) with Back Propagation (BP) algorithm was applied to forecast the amount of monthly stream flow and numerous alternatives were tested to find the most suitable model. The results showed that although all 12 past months perform the best results, the combination of 1, 6 and 12 months ago has the same results as well. So the preferable option for forecasting is the second one because of the less time in training the networks with the same results.

Artificial neural network, Stream flow forecasting