Estimation of Inlet Flow and Water Reservoir Level of Dez Dam Using Fuzzy Systems and Artificial Neural Networks

One of important aspects of the management of river basins is the management of flood events occurring with its boundaries. The importance of flood management is highlighted by the magnitude of economic losses associated with floods. Flood forecasting is undoubtedly a challenging field of operational hydrology, and a huge literature has been developed in years. Since flood warning systems do not aim at providing an explicit knowledge of the rainfall-runoff process, and the main concern is making accurate and timely predictions at specific locations, a simple black-box model is then preferred in identifying a direct mapping between inputs and outputs. 
This study presents the development of artificial neural network (ANN) and fuzzy logic (FL) models for prediction of daily reservoir inflow. Furthermore, adaptive neuro-fuzzy inference system (ANFIS) technique was used to construct a water level forecasting system during flood periods. For comparison the results, a linear regression model was developed as the benchmark for comparison in flood forecasting. Dez dam reservoir was used as a case study. In this study, the precipitation, evaporation and discharge data of upstream stations were used to forecast the Dez dam reservoir inflow. The Talezang gauging station data was intended as the output data and the Dez dam reservoir inflow indication. The original data consists of 6 years (1993-1999) which 1363 input-output pairs were intended for training set and 313 and 314 data set were used for validation and test set, respectively. The emphasis in this study is to improve the short-term (1 to 4-day-ahead) forecast of reservoir inflows. In addition, The hourly water level of the reservoir and the upstream five stations for floods during 2001 to 2003 years were gathered to forecast the water level 3 to 60 hours ahead. To investigate whether ANFIS can deal with human knowledge, two ANFIS models were developed, one with a human decision of reservoir outflow as input variables, another without that.
Results demonstrated that ANNs can predict the reservoir inflow for 1-day-ahead, especially in training period with %93.9 correlation, RMSE=4.13 better than fuzzy systems (%93 correlation, RMSE=5.17) and regression model (% 89.7 correlation, RMSE=4.73). As the prediction time go ahead, fuzzy systems and regression model revealed the better results and ANN model results deteriorated. The results showed that the ANFIS can provide high accuracy and reliability for reservoir water level forecasting. In models with human decision as input variables, ANFIS demonstrated better results in short periods (1 to 12 hours).