Future flow prediction and flow modelling with artificial neural network using diffirent parameters for Aksu River in Giresun

In this study, a flow modelling has been made in three different current observatory stations located the Aksu River Basin in Giresun, a province in Eastern Black Sea Basin. The flow data provided by General Directorate of Hydraulic Works, the precipitation and evaporation data obtained from Turkish State Meteorological Service are used for the modelling. artificial neural network (ANN) method is preferred for the flow modelling that will be created. The multi layered artificial neural network (ML-ANN) method which is selected in the architecture of ANN model has been tested in 14 different models with various combinations of the specified parameters mentioned before. The estimated values from the experiments, are evaluated according to the error coefficients of RMSE, r, RH, E and MAE. The most successful input combinations for each station were tested by Multiple Linear Regression Analysis (MLR) model. During the design phase of structures such as predicted flow results flood protection facilities, hydroelectric plants and water treatment plants, the predicted flow results are compared with the observation values for the most commonly used flow rate which were determined according to the possibility of exceedance. As a result, ML-ANN method was given better results than MLR method and this study shown that ML-ANN method can be used for prediction of future flow datas.