Discharge is one of the bases in a plan for a water resource management activity. Determination of the discharge value can be done through measurements directly in the field or through analysis. With the analysis of rainfall on river discharge can be an alternative in water structure planning, with the right methods the results of the discharge analysis will approach with real hydrological condition. In this study, the analysis was carried out using the Artificial Neural Network method to find out the results of the modeling of river discharge based on rainfall and evapotranspiration data and to know the comparison between model discharge and real discharge. To discover the compatible between modeling discharge with real discharge, calibration and learning of Networks has been done with 6 until 9 years learning data, and verification of the model discharge by 4 until 1 years from the rest of the calibration data. In network learning, epoch 500 until 2000 are used. The error test are Mean Square Error (MSE), Mean Absolute Error (MAE), Relative Error (Kr), Correlation Coefficient (R), Nash-Sutcliffe Efficiency (NSE). The results of comparison between model discharge using the Artificial Neural Network with real discharge, based on the test results discover that in the distribution of 7 year learning data and 3 year test data with epoch 1000 that compatible with criteria and had the best value, based on result of Nash-Sutcliffe Efficiency (NSE) value, and the Correlation Coefficient (R)
Copyrights © 2020