Representação de cenários de demanda e da função de produção hidrelétrica no planejamento da operação de sistemas hidrotérmicos a médio prazo

Abstract

The operation schedulling of Brazilian System constitutes a task of high complexity, either due to the large number of hydro and thermal plants distributed in the four interconnected submarkets, or due to the outstanding characteristics of the hydro plants, with respect to the uncertainties in the hydrological scenarios future and due to spatial and temporal couplings. The objective of this planning is the determination of an optimal dispatch policy of plants that minimizes the expected value of the operating costs in the considered horizon. An important constraint is the supply of demand, which is obtained by the difference between the effective load of each submarket and the generation of non-simulated plants, composed of renewable energies (wind, solar, biomass), small hydropower plants, and others. However, the growing penetration of renewable generation, coupled with the uncertainties of natural sources such as wind and solar (because they depend on climatic factors and are therefore unpredictable), contributes to the need to represent demand in different scenarios for a correct dispatch of the plants and a more reliable system operation. Knowing that the current official models deal only with the uncertainties arising from the hydrological scenarios, due to the difficulty of implementation and memory allocation in the consideration of other uncertainties in Stochastic Dual Dynamic Programming, this work proposes methodologies that include the several scenarios demand, besides representing in greater detail the generation of hydroelectric plants. In addition, algorithms are proposed for analytical representation of the Immediate Cost Function in isolated systems with multiple areas in the search for a reduction of the dimensions of the problem. Finally, tests are performed on tutorial systems and the results are analyzed to evaluate the computational performances of the proposed methodologies.