Application of Artificial Neural Network (ANN) for Reservoir Water Level Forecasting

Future water level forecast helps in knowing the water storage capacity of reservoir which can further be employed for irrigation, water supply, hydropower, etc. These soft computing techniques can map an input-output pattern without the prior knowledge of the criteria that influence the forecast procedure. Amongst the all, the artificial neural network (ANN) is one of the most accurate models that is used in water resource management. The principal inputs that are used to compute water level at t+1 time are: Inflow, water level at t time and released water. The statistics parameters room mean square error (RMSE), coefficient of correlation (R), coefficient of determination (R2) and discrepancy ratio (D) were used to get best model out of three alternatives. The objective is to define the ANN model by applying different types of network tools like Feed Forward distributed time delay, layer recurrent and NARX on a set of input data to forecast the 10 day ahead water level for the case study, Sukhi Reservoir Project, Gujarat, India. The study reveals that, among the three algorithms applied, ANN using Feed Forward distributed time delay is an appropriate predictor for real-time Water Level forecasting of study area.