Prediction of long-term streamflow by using adaptive neuro-fuzzy inference system ((ANFIS)

Water resources are one of the most basic needs of living life. In order to sustain human life without any problems, a rational planning is required for the protection and use of existing water resources. At the beginning of the plans to be made, the potential of the water source to be used in the future should be determined. Therefore, river flow estimation is necessary to provide basic information on a wide variety of problems associated with the functioning of river systems. In this study, the daily flow values of Zamanti River-Değirmenocağı, Zamanti River-Ergenuşağı and Eğlence River-Eğribük stations in the Seyhan Basin in Turkey were investigated. has been made. Artificial intelligence methods have been used instead of traditional methods for some time due to their success in modelling complex nonlinear problems. Within the scope of the thesis, the Adaptive Neuro-Fuzzy Inference System (ANFIS) model was trained using Simulated Annealing (SA), Back Propagation (BP) and Hybrid Learning (HB) algorithms in order to make forward flow rate estimation from past flow measurement values and the results obtained from all models were compared. Mean Absolute Error (MAE), Root Mean Square Error (RMSE), Determination Coefficient (R2) and Mean Absolute Percentage Error (MAPE) evaluation criteria were used for comparison. After the analysis, it was concluded that HB and BP algorithms can be used more successfully and effectively than SA algorithm in training ANFIS parameters in nonlinear problems. Key Words: Neuro-Fuzzy, ANFIS, Hydrology, Streamflow, Prediction