Modelos estatísticos de séries temporais para previsão do comportamento de instrumentos da barragem de Itaipu

Abstract

The dissertation addresses the use of statistical time series models to predict the behavior of measuring instruments installed at the Itaipu Binational Hydroelectric Dam. The motivation for the research lies in the importance of structural safety of dams, especially in the Brazilian context, where failures have caused significant environmental, financial, and social impacts. The research adopts a quantitative and descriptive approach, defining and comparing three statistical methods adjusted for the particularities of the time series of piezometers embedded in the dam's foundation. Highlighted were the adjusted exponential smoothing methods in a context where the blocks containing the instruments showed greater variability in the center and better accuracy in predictability at the extremes.