Resumo:
This work proposes to determine a consistent and representative quantitative model for
the volatility forecast of electric power demand, for independent consumers, in short-term
regimen, through the simulation of GARCH models for temporary univariate series, in which
the behavior of present industrial loads will be evaluated using actual data, aiming at the
forecast of the volatility of short-term (one week). This study presents a revision of the
literature on some statistical models of forecast and mainly on the GARCH model used in this
work, presenting its particularities and applicability. The accomplished simulation is based in
the construction of non-linear univariate models in order to forecast the volatility associated
with the demand, and is based on data of power demand time series. A software program was
developed, based on the toolbox GARCH of the Software MATLAB 7.0.1.