Desenvolvimento de um sistema simplificado para previsões probabilísticas de afluências ao reservatório da UHE Itaipu

Abstract

The inflows forecast to a hydroelectric plant is extremely important for its reservoir operation, since it allows planning its production according to water availability and also plays a fundamental role in dam safety management. Thus, a hydrological quality forecast is essential for a good planning of the available resources management, in order to employ them efficiently, focusing on operational safety and power plant production, associated with environmental and social responsibility. In this sense, hydrological modeling has been increasingly used as a tool to support decision making in the use and management of water resources. However, hydrological modeling and its predictions are subject to several types of uncertainties; thus, estimates of these uncertainties must be associated with hydrological forecasts to support decision making, in order to quantify the reliability of these forecasts and allow the decision maker to have complete information for risk and control management. One way of explaining these uncertainties is through the use of probabilistic forecasts, instead of deterministic ones, allowing to signal possible future scenarios in view of the various uncertainties associated with these forecasts. This type of approach is already widespread in Meteorology and has evolved in hydrological forecasts, but it can still be considered incipient in Brazil and all South America, in general. In this context, this work addresses the use of long-term (or seasonal) probabilistic hydrological forecasts, with the incorporation of uncertainties, for operation planning of the Itaipu Hydroelectric Power Plant reservoir. For this purpose, an operating system for seasonal forecasting of its reservoir inflows was developed, composed of the hydrological model SMAP, in its monthly version, associated with the climate forecasts by ensemble prepared by ECMWF. Besides the uncertainties associated with precipitation forecasts, represented by the spread of the ensemble, this systems also incorporates those related to model parameters and model states. For this purpose, the GLUE method was applied, in order to estimate the uncertainties associated with the hydrological model. The developed system proved to be a potentially useful tool for probabilistic forecasting of Itaipu’s inflows, and the forecasts’ sample analyzed, although limited and insufficient for generalizing the results, indicated the benefits of using probabilistic forecasts in favor of strictly deterministic ones, represented by the median of the forecasts ensemble.