Assessment of Inflow Forecast Uncertainty in Optimal Reservoir Operation

In this paper, various types of stochastic dynamic programming models (SDP) and also a deterministic dynamic programming (DP) are presented and compared for multipurpose Dez reservoir dam located in southwest of Iran. The impact of inflow forecasting uncertainty in optimum reservoir operation is investigated through two types of stochastic dynamic programming models. SDP models are different based on hydrologic state variable and inflow conditional or non conditional probability assumptions. A simulation model is developed to investigate the achieved optimum policies in different models. So, average of different operation variables and also performance criteria such as reliability, resiliency and vulnerability are used to compare the results. In two SDP models which apply current inflow instead of previous time step inflow as hydrologic state variable, real time simulation is done with forecasted and observed inflows. The effect of forecasting accuracy and different forecasting methods in reservoir operation are also studied. In general, objective function is considered as minimizing the sum squared of two sided deviations from target release and storage. This research shows that lack of attention to inflow forecasting in models which need it and assuming known values could lead us to unreal and false results and mislead us in selecting type of model. On the other hand, it is noticed that forecasting accuracy plays an important role in optimum reservoir operation.