THE INFLUENCE OF THE FORECAST MODEL ERROR ON OPTIMIZING THE STORAGE FUNCTION CONTROL OF THE RESERVOIR USING

Abstract

One of the ways to manage drought is to use optimization when managing the storage function of reservoirs. Optimizing reservoir management itself requires inflow forecasts. A lot of forecasting models have been created in the world, which according to many authors can be used for the given purpose. The models themselves are generally assessed according to classical criteria. The article therefore attempts to determine whether the quality of the forecast model for the mentioned application can be assessed as a whole (the model must be able to predict the entire interval of occurrence of flows) or whether the local quality of the model is sufficient (the model can predict a defined interval of occurrence of flows with good quality). For this purpose, a fictitious forecasting model was built that makes predictions based on real flow series and white noise. The predictions created in this way are then used to control the real reservoir using the deterministic evolution method. The achieved results are then compared with the results when a directly real series was used for control. The results themselves may be somewhat surprising, as certain combinations of the fictitious model predictions achieved better results than the results obtained using the real values according to the chosen criterion.