Geração de cenários sintéticos mensais de vazões condicionadas ao el niño / oscilação sul :aplicação ao planejamento da operação de sistemas hidrotérmicos

Abstract

The energy operation planning of the Brazilian National Interconnected System (NIS) is a complex problem of resources optimization, due to its continental proportions and its hydrothermal energy generation system. Currently, a periodic auto-regressive (PAR(p)) model is used to generate synthetic scenarios of monthly streamflow and energy for the mid-term and long-term optimization models. It is known that the climate phenomenon El Ni˜no - Southern Oscillation (ENSO) is capable of changing the Brazilian hydrological regimes through the reduction or increase of rainfall in several regions. This dissertation aims to propose a methodological improvement on the generation of synthetic scenarios of streamflow, to include information from the ENSO. A new stochastic model called Markov-switching periodic auto-regressive (MS-PAR(p)) is proposed, in which the model parameters are segmented and estimated between different states in the same month. For the scenario generation of the ENSO states, monthly Markov chains are adjusted. A procedure for inclusion of probabilistic forecasts of the climatic phenomenon is developed. The original and proposed models are compared through several simulations performed with the configuration of the NIS. The MS-PAR(p) model presents better performance in the south and north regions, where there is a more significant influence of the ENSO. It is demonstrated that the proposed methodology is able to generate synthetic scenarios which distribution better resembles the observed one. This fact occurs, mainly, during periods when the streamflows are below-average, which may represent an improvement for the planning of the Brazilian energy operation.