The identification of the probability distribution model that provides the best fit to the wind speed databases is necessary for defining investment and developing projects about the wind potential of different locations. For this, the estimation of the parameters of the models is essential in this process. The aim of this study is to investigate among the distribution models and methods for estimating their respective parameters with better modeling in the literature which of them provides better fit to the wind speed data of Petrolina-PE. Through the case study, of quali-quanti nature, the adjustment of the Moment Method, the Estimation of Maximum Likelihood and the Particle Swarm Optimization (PSO) algorithm with Weibull were evaluated in this work, as well as the PSO with the Lognormal-Weibull and Weibull-Weibull distributions to the historical series of information. The results, investigated with the RMSE, R^2 and X^2 error measures and by verifying the percentage of correctness between the theoretical and sample quantiles, demonstrated a better modeling of the Lognormal-Weibull distribution model with the PSO algorithm to the historical speed series of the wind. Thus, from the determination of the best distribution model that fits the data in the region, it may be possible to generate estimated wind speed series for areas where these historical series do not exist.