Shahram Karimi-Googhari - Assistant Professor, Department of Water Engineering, Shahid-Bahonar University of Kerman, Kerman, Iran

Time series modeling for either data generation or forecasting of hydrologic variables is an important step in the planning and operational analysis of water resources. The capability of Artificial Neural Networks (ANN) in modeling of daily reservoir inflow forecasting was examined in a small tropical catchment. Cross-validation and pre-processing of data was considered as alternatives in modeling process. The model inputs were extracted using auto-, cross-, and partial auto-correlation functions. The results showed that the feed forward back-propagation neural networks are able to forecast extremely changeable daily reservoir inflows. Cross-validation of data improved the model performance indices. Transforming the data to normal distribution prior to training confirmed increasing significantly the model persistency and generalization in simulating an independence data set.

Data pre-processing, Neural networks, Forecasting, Reservoir inflow