ANALYSIS OF PROBABILISTIC METHODS FOR WIND POWER FORECASTINGCurrently, wind energy is showing prominence in the Brazilian scenario because it is an energy source that has high availability in the territory. This work considers the problem of lack of assertiveness when estimating wind power from wind speed by the conventional model of the power curve, in which there is an emission of real values around the theoretical curve. The study has two main objectives, the first of which, through parametric models such as linear, quadratic, cubic and weibull, is to understand how to more faithfully approximate the forecast to reality. For this, a comparison of the root mean squared error (RMSE), between the theoretical and real value obtained from data from 16 wind turbines collected in the field, generating a participation with 52,428 measurements, was carried out. Subsequently, the study turns to verifying the behavior of the same methods, but based on temporal, seasonal and moon phase groupings, in order to find the one that provides the greatest reduction in the error when compared to the generated power. Finally, it can be concluded that the linear and weibull models presented the best modeling results for the database, with a difference in error of more than 90 kW compared to the power curve. Clustering by seasons of the year, followed by a second clustering of moon phases contributed to error reduction in most cases, as in summer with a crescent moon the difference in error compared with no clustering dropped by more than 200 kW.