Integração da decomposição de Benders em dois estágios e programação dinâmica dual para o planejamento hidrotérmico-eólico-solar com funções de custo intra-estágio

Abstract

The generation of renewable energy from wind and solar in Brazil has grown significantly in recent years. Although this represents an essential step toward a more sustainable and diversified electricity matrix, it also poses new and relevant challenges for power system operation planning. This is because the intrinsic variability of these sources, subject to hourly fluctuations, has progressively rendered the simplified medium- and long-term representations, based on low temporal granularity, inadequate. To address this challenge, Intra-Stage Cost Functions (FCIE) have been proposed in the literature as a means to incorporate short-term effects into medium- and long-term planning without explicitly discretizing stages into hourly intervals. However, these functions are traditionally constructed a priori, before solving the Dual Dynamic Programming (DDP), resulting in a prohibitive number of subproblems for large-scale systems, such as the Brazilian system. In this context, this work proposes an integrated formulation that combines two-stage Benders decomposition and DDP, enabling the iterative and on-demand construction of FCIE throughout the DDP solution process, while accounting for the hydropower production function and the representation of multiple interconnected subsystems. This integration allows selective solution of only the necessary subproblems at each stage, considering hourly aspects such as load profiles and the intermittency of wind and solar generation. As a result, FCIEs that are more consistent and tailored to the DDP problem being solved are obtained, thereby overcoming the limitations of previous preprocessing-based approaches. The proposed methodology is validated through illustrative cases, designed to isolate and highlight the effect of the approach, through a large-scale case based on data from the Brazilian power system and a short-term simulation coupled with the generated Future Cost Function, demonstrating gains in realism and efficiency in the formulation of medium and long-term operation policies.