Evaluating Artificial Intelligence (AI) models in monthly reservoir inflow forecasting (Case study: Dez dam, Iran)

In the present study three AI techniques (ANFIS, GP, and ANN) have been used to forecast the inflow into Dez reservoir in the southwest of Iran. In order to develop a suitable model of time series for forecasting inflows, the models have been used considering pervious inflows and cycle terms in the input vector. To evaluate the model performance, root mean square error, mean absolute error, correlation coefficient and Nash-Sutcliffe coefficient of efficiency have been employed. Results showed that the ANFIS has the best performance in forecasting inflow time series into Dez dam reservoir. The GP and ANN are in the second and third ranks, respectively. According to the results, in all of the AI methods (ANFIS, GP, and ANN), the model with cycle terms had better performed when comparing to those models which are not considering the periodic nature.

Keywords:

Artificial Intelligence models , Cycle Term , Reservoir Inflow , Dez dam