Previsão de vazões mensais para o sistema interligado nacional utilizando informações climáticas

• Ano de defesa: 2012
• Autor(a) principal: Alexandre, Alan Michell Barros
• Tipo de documento: Tese
• Tipo de acesso: Acesso aberto
• Idioma: por
• Palavras-chave em Português: Recursos hídricos; Modelos matemáticos
• Link de acesso: http://www.repositorio.ufc.br/handle/riufc/3877

The hegemony of hydroelectricity in Brazil’s electricity matrix calls for a careful analysis of river regimes considering the significant impact such streamflow variations may have on energy supply and consequently on the entire country’s economy. Based on this fact, streamflow forecast statistical models have been important tools to support decision-making, planning and management of water resources used in the Brazilian Hydropower Network (Sistema Interligado Nacional - SIN). In that sense, this thesis proposes methodologies of simultaneous forecast and for developing monthly streamflow scenarios in SIN Base Stations (Postos Base - PBs) using statistical models; the aim is to make better use of available data by including methods that will maintain the national hydrographic network's spatial structure. It also aims to analyze the impact of incorporating climate information for monthly streamflow forecasting. Proposed streamflow forecasting models use natural streamflow data generated by Brazil’s Electric System National Operator (ONS) and statistical techniques such as Multiple Linear Regression, Principal Component Analysis, the Stepwise method for choosing explanatory variables, in addition to models of the Periodic Auto-Regressive (PAR) and Periodic Auto-Regressive Exogenous (PARX). PAR models show the best performances, according to the index Distance Multicriteria in most months and SIN PBs when compared with PARX models. Among spatial correlation methods for PAR models, the correlation between regression noises (CRD) and principal component analysis (ACP) stand out. There is no predominant method for all months and SIN PBs. The best PARX models are those that use climate indexes as exogenous variables, among which the following stand out: AMO (Atlantic Multi-Decadal Oscillation) and TNI (Trans-Niño Index). They show better performance during the dry season of basins in the North of Brazil - Amazon and AraguaiaTocantins; in the Middle-West region in Brazil - Atlantic East and in most rivers that make up the Paraná Basin.