Aplikasi Teknik Data Driven untuk Prediksi Debit Sungai Bulanan Studi Kasus Bendung Loning, Magelang

• Main Author: Joko Suryanto
• Format: Article
• Language: English
• Published: Syiah Kuala University 2016-10-01
• Series: Rona Teknik Pertanian
• Subjects: ANFIS; ARIMA; debit bulanan; FFNN; model data driven; Sungai Loning; data driven model; Loning River; monthly run off
• Online Access: http://jurnal.unsyiah.ac.id/RTP/article/view/5649
• ISSN: 2085-2614; 2528-2654

Abstrak. Model prediksi debit sungai sangat penting dalam perencanaan, desain dan manajemen sumberdaya air. Penelitian ini bertujuan untuk membandingkan akurasi prediksi debit bulanan sungai Loning di DAS Loning Magelang menggunakan pendekatan model data driven. Pembentukan model didasarkan pada model deret waktu debit bulanan menggunakan data debit bulanan antara Januari 1990 hingga Desember 2015. Tiga model data driven yaitu ARIMA, ANFIS dan FFNN digunakan untuk prediksi debit bulanan sungai Loning periode 2014 - 2015. Indeks error (RMSE dan MAPE) dan koefisien Nash-Sutcliffe efficiency (NS) digunakan untuk mengevaluasi akurasi prediksi model. Hasil penelitian diperoleh nilai RMSE, MAPE dan NS model FFNN adalah 8,422 lt/ dt, 22.79 % dan 0.709, untuk  model ANFIS 9,465 lt/dt, 25.62 % dan 0.633, sedangkan model ARIMA diperoleh  9,710 lt/dt, 27. 32 % dan 0.614.  Nilai indeks error tersebut mengindikasikan bahwa model FFNN lebih sesuai untuk simulasi dan prediksi debit bulanan sungai Loning dibandingkan model ANFIS dan ARIMA.   Application Data Driven Technique for Monthly Runoff Forecasting: A Case Study of Loning Dam, Magelang  Abstract. The development of runoff forecasting model is one of the most important aspects in water resources planning, design and management. This study aimed to compare the accuracy data driven models for simulation and forecasting the monthly runoff data of Loning river, Loning Watershed Magelang. The models were developed based on time series model, and monthly data collected over 26 year period from January 1990 to December 2015.  Three data driven models, ARIMA, ANFIS and FFNN models, were used for forecasting monthly runoff for period 2014 -2015. The index error, root mean square error (RMSE), mean absolute percentage error (MAPE) and Nash-Sutcliffe efficiency coefficient (NS) were employed to evaluate the performances of model developed. The RMSE, MAPE and NS indices were obtained as 8,422 lt/s, 22.79 % and 0.709 for FFNN model, as 9,465 lt/s, 25.62 % and 0.633 for ANFIS and as 9,710 lt/s, 27. 32 % and 0.614 for ARIMA model. The result indicated that FFNN model appear to be better than ARIMA and ANFIS model for simulation and forecasting the monthly runoff Loning river.