Modelo individualizado de usinas hidrelétricas baseado em técnicas de programação não linear integrado com o modelo de decisão estratégica

Abstract

On the horizons of long term, the official computing model to the planning of the operation of the National Grid (NG) in Brazil is based in the technique of Stochastic Dual Dynamic Programming (SDDP) and using equivalent systems of energy. In order to produce individual estimates of the hydro plants, NG’s official model uses operative heuristics which search for the operation in parallel of the reservoirs and consider the nonlinear and operative restrictions associated to the problem. In other words, the operative decisions generated by SDDP to the energy equivalent systems are disaggregated between its own hydro plants. Both computing models are developed by CEPEL. The model NEWAVE produces cost-to- go function monthly through SDDP technique and these functions are also used by the SUISHI- O model (developed by CEPEL), responsible for the disaggregation of the decisions associated to the equivalent systems between its hydro plants. The aim of this thesis is to propose an Individualized Model of Hydro Plants (IMHP) based on nonlinear programming for the monthly planning of the operation using the cost-to-go functions produced by the NEWAVE model. IMHP uses an alternative representation of fourth degree polynomials associated to the outflow-level downstream relationship based on the logistic function in order to stabilize the convergence of the nonlinear optimization model. It has been developed a robust computing platform which is capable of using the same set of datas of the official computing models of NG, which is monthly available by the Indepen- dent System Operator (ISO). Consequently, it was possible to compare the results generated by SUISHI-O model with the ones generated by IMHP. In order to do that, the datas regarding the Monthly Program of Operation of January 2011 were used. This platform is developed in C++/Java to the operational systems Windows and LINUX, once that the model LINGO is used to solve problems of nonlinear optimization. IMHP is able to represent the majority of the details associated to the problem, such as, for example, the expansion of thermal and hydraulic parks, the market growth, and the minimum obligatory outflow restrictions, the minimum volume to be transfered in hydro plants with reservoirs, the generation of small hydro plants, the limits of exchange between equivalent systems of energy, various levels of deficit of energy, the evaporation and the act of filling the dead volume.