Operation of Roseires and Sennar Dams Using Artificial Neural Network

Abstract

This study, describes the operation of Roseires and Sennar dams during the dry season when the demand is greater than the available water at El Deim station and tries to find out the relationship between water levels, amount of evaporation losses, predicted inflow discharges at El Deim station during the recession period and reservoirs contents of Roseires & Sennar dams. Artificial neural network has been used to establish three models: Model -1 : To predict inflow discharges at El Deim station . Model -2 : To predict reservoir water level knowing reservoir content and year of application. Model -3 : To find out the numeric relationship between reservoir water level & Evaporation losses. Artificial Neural network using NeuroShell2 software, which provides a quick and flexible means of creating models for prediction, has been shown to perform well in comparison with conventional methods. Different scenarios of different network structures have been tried, the best architecture for the three models is the three connected layer with specified hidden neurons, but the hidden layer in the first two models consist of two slabs. The results obtained from model-1, model -2 and model -3 are used in a computer program to establish the emptying program of Roseires & Sennar reservoirs using M.S Excel software. By assuming different areas of Wheat crop - its planting season start on 2nd period of October - the emptying of the two reservoirs has been run taking into considerations the following aspects: i. At the beginning of the empting program, the two reservoirs of Sennar & Roseires dams are at their predetermined maximum water levels. ii. Irrigation demands include projects upstream Sennar dam, Gezira & Managil schemes and minimum downstream requirements. iii. At the end of emptying program, the water level of the two reservoirs should not drop below the minimum water level of both reservoirs respectively. Different abstraction ratios for Sennar & Roseires have been tried to satisfy the above mentioned considerations. The best ratio found is 1(Sennar): 4.4 (Roseires).