A paradigm shift around the world was caused by increased concern about the use of fossil fuels used as the main source of electricity generation, the corresponding climate change and increasing environmental damage. In recent years, wind energy has shown steady growth as a sustainable alternative for electricity production, which can be seen from the growth of its installed capacity worldwide. Brazil is among the ten countries that have the largest installed capacities, and presented 9.42 percent of electricity generation from the wind source in the last year. However, wind randomness and intermittency are the biggest challenges in integrating this source into the energy system. In this context, this research proposes the application of the Singular Spectrum Analysis (SSA) technique as a forecast method for a series of wind speed in Brazil, making a comparative analysis of SSA models considering different forecast horizons and training set for different days forecast, with different time series sizes. In this way, the time series of the whole year is compared with only the last month of this series to forecast the last seven days of the month of December. The results of this application show that for most days the use of the whole year as a training set obtained better performance, indicating that the use of the SSA technique can be an alternative for time series with a large amount of data.