State-of-the-art applications of short-term reservoir management integrate several advanced components, namely hydrological modelling and data assimilation techniques for predicting streamflow, optimization-based techniques for decision-making on the reservoir operation and the technical framework for integrating these components with data feeds from gauging networks, remote sensing data and meteorological weather predictions. In this paper, we present such a framework for the short-term management of reservoirs operated by the Companhia Energética de Minas Gerais S.A. (CEMIG) in the Brazilian state of Minas Gerais. Our focus is the Três Marias hydropower reservoir in the São Francisco River with a drainage area of approximately 55,000 km and its operation for flood mitigation. Basis for the anticipatory short-term management of the reservoir over a forecast horizon of up to 15 days are streamflow predictions of the MGB hydrological model. The semi-distributed model is well suited to represent the watershed and shows a Nash-Sutcliffe model performance in the order of 0.83-0.90 for most streamflow gauges of the data-sparse basin. A lead time performance assessment of the deterministic and probabilistic ECMWF forecasts as model forcing indicate the superiority of the probabilistic model. The novel short-term optimization approach consists of the reduction of the ensemble forecasts into scenario trees as an input of a multi-stage stochastic optimization. We show that this approach has several advantages over commonly used deterministic methods which neglect forecast uncertainty in the short-term decision-making. First, the probabilistic forecasts have longer forecast horizons that allow an earlier and therefore better anticipation of critical flood events. Second, the stochastic optimization leads to more robust decisions than deterministic procedures which consider only a single future trajectory. Third, the stochastic optimization permits to introduce advanced chance constraints for refining the system operation.

In this paper, the application of the method of affinely adjustable robust optimization to a planning model of an energy system under uncertain parameters is presented, and the total scheduling costs in comparison with the deterministic model are evaluated. First, the basics of optimization under uncertain data are recapped, and it is described how these methods can be used in different applications for energy systems. This is followed by the methodology of adjustable robust optimization by defining the affinely adjustable robust counterpart. Finally, a numerical case study is conducted to compare the adjustable robust method with a rolling deterministic scheduling method. Both are implemented on a model of an energy system and compared with each other by simulation using real-world data. By calculating the total operating costs for both methods, it can be concluded that the adjustable robust optimization provides a significantly more cost-effective solution to the scheduling problem.