Avaliação da importância da autocorrelação para a análise de tendências das séries históricas de vazão do Operador Nacional do Sistema Elétrico (ONS) no Brasil

• Ano de defesa: 2015
• Autor(a) principal: Medeiros, Priscilla Paiva de
• Tipo de documento: Dissertação
• Tipo de acesso: Acesso aberto
• Idioma: por
• Palavras-chave em Português: Recursos hídricos; Séries hidrológicas; Correlação temporal; Setor hidrelétrico
• Link de acesso: http://www.repositorio.ufc.br/handle/riufc/13099

Brazil has majority share of hydroelectric power in its generation matrix electricity, depending on the rainy seasons and settlement of multiannual reservoirs to ensure safety in the generation and distribution system power national power. In this context, time series analysis can provide information useful for planning and more efficient operation of the reservoirs, identifying possible trends in the flow of data or precipitation of the integrated stations hydroelectric plants. The trend analysis in hydrological series is held traditionally using non-parametric tests such as the Mann-Kendall test, and You can treat both monotonic trends or abrupt as in various ways. Such tests, although not require information about the distribution of data, require data to be independent, which rarely occurs in series hydrologic variables. Thus, new tests have arisen in the trend analysis, with objective of taking the autocorrelation of the series under consideration. This paper proposes trend analysis of historical streamflow series of major basins contribute to the National Interconnected System (SIN), obtained from data National Electric System Operator (ONS). The analysis will aim check importance of considering the autocorrelation of the data to detect trends, analyzing the first series without considering the autocorrelation - the nonparametric method Traditional Mann-Kendall with Sen estimator - and then applying filters Pre-Whitening and Trend-Free Pre-Whitening. The tests were implemented through the MatLab program to analyze the average natural flow monthly provided by the ONS. The spatial distribution of trends found It was presented through maps created the Free Software QGIS. The results found for the ONS posts corroborated the theory that the autocorrelation influences the detection of trends. Keywords: autocorrelation, trend analysis, hydropower se