Model Hidrologi Runtun Waktu untuk Peramalan Debit Sungai Menggunakan Metode Gabungan Transformasi Wavelet - Artifical Neural Network (Studi Kasus : Sub DAS Siak Hulu)

Softcomputing method has been widely used as hydrological analysis model, one of them for streamflow forecasting. Hybrid model of Wavelet Transformation - Artificial Neural Network (WANN) is one of the softcomputing method that can predict streamflow. WANN models need to be tested reliability of the Siak Hulu sub-watershed given the importance of streamflow information to generate management, planning, and useing water resources and sustainable accurately. WANN models in this study used AWLR data converted into discharge data at Siak Hulu sub-watershed from January 2002 to December 2012. First, the data were decomposed and reconstructed using Wavelet Transformation. Then, ANN constructed forecasting model use backpropagation algorithm. WANN models were constructing use 12 forecasting scheme, and then it compared to obtain the best model of each forecasting scheme. The results of statistical analysis on coefficient of correlation (R) showed that the 12 schemes of WANN models developed by the process of calibration, testing, validation and simulation were categorized as very strong correlation. As for the process simulation models from 2011 to 2012, WANN models db5 level 2 for lead time one day scheme resulted the best correlation coefficient (R) is 0.9501752.