PREDIKSI POTENSI DEBIT BERDASARKAN DATA HUJAN MAKSIMUM BULANAN DENGAN METODE JARINGAN SYARAF TIRUAN BACKPROPAGATION DI DAS ALANG

Abstract

The data flow is important information in the management of water resources. Water resources management has many aspects such as flood controlpurposes, and so on electrical energy potential. For water resources management and watershed planning Alang long-term infrastructure, flow of dataneeded in the future. So we need an approach to the provision of discharge data with neural network models. The purpose of this study is (1) Determinethe coefficient of ANN parameters, (2) Determine the discharge prediction years 2013-2016 and (3) Determine the reliability of the model.This research is descriptive quantitative research, where data used are secondary data. The secondary data used were obtained from the office. Stages ofthe research is to collect data year 2001-2012 rainfall and discharge as well as topographic maps. Perform calculations using the area rain Thiessenpolygon method. Results rainfall areas converted into discharge using the Rational method with the help of software Backpropagation ANN Matlab(R2010b). Then perform simulations until the results obtained are at the limits set and simultaneously obtain discharge predictions. Furthermore, totest the reliability of the model.The results showed that the ANN parameters : Period = 4 years, Hidden Layer = 2 pieces (2 each neuron), Epoch = 150000, Goal Momentum =0.6 and = 0.02. Then for discharge predictions for the year 2013-2016 Alang DAS can be seen in table 5. Reliability models 58.17% derived fromthe analysis of reliability. The model has achieved 58.17% reliability and 95% Confidence qualify, but the parameters of the model need to be modifiedto apply to other watersheds.