Otimização implícita estocástica para operação de reservatórios

Abstract

Analysis of monthly streamflows was performed. A deterministic periodic component in the mean and the variance was identified, as well as the dependence (first order linear autoregressive model) in the stochastic component. Then a probability distribution was adjusted to the independent stochastic component. The structural model was used to generate several series of monthly streamflows in which critical periods were identified. These critical periods were used to determine the optimal operation of a multi-purpose reservoir with the objective of maximizing produced energy, making use of backward dynamic programming. Multiple linear regression analysis was used to obtain monthly operation rules for the reservoir, using the results of the optimization model. Tres Marias reservoir was used as an example of application of the techniques developed, and the monthly operation rules developed were applied to the historic series.