PREDICTION OF RIVER FLOW DISCHARGE USING HYBRID NEURAL NETWORK ALGORITHM

Successful river flow forecasting is major goal and essential procedure that is necessary in water resources planning and management. In accordance with the characteristics hydrological phenomena in a river watershed area, the river flow varies irregularly. Therefore, it is difficult to predict the magnitude of discharge in the river an exactly at a certain moment. The main purpose of the research is to develop of model to predict river flow discharge to monitor the dynamics of river flow fluctuations especially the sufficient information for Regional Water Supply Company (PDAM) Pekanbaru city continue so to operate serving the needs of drinking water for urban communities and as for the development of the flood early warning system (FEWS) in Siak river. The method of research approach used is using Self Organizing Map (SOM)- Artificial Neural Network (ANN) developed a hybrid system between algoritma SOM and ANN Backpropagation algorithm as branch of soft computing. The location of research on the of Station Automatic Water Level Record (AWLR) Pantai Cermin in Pantai Cermin Village, Kampar Residence, Riau Province. The data used in this research is secondary data of discharge sourced from River Region Agency (BWS III) Sumatera Pekanbaru, Riau province from year 2002 until 2007. The main results of the research proved the use of the model SOM-ANN better than model ANN as branch of soft computing that have advantages on recognized pattern data resulted in high prediction accuracy at river flow discharge in Siak River for one annual ahead (Qt+1) has a degree of very strong relationship with the value of coefficient of correlation respectively 0.9016 and 0.8589.