The accurate fitting of wind speed probability distribution plays an important role in estimating the wind energy in potential wind farms.In this study, the two parameters Nakagami distribution is introduced to fit the wind speed data of eight stations covering coastal, island and inland in the western United States, and the Rayleigh, Gamma, Weibull, lognormal, generalized extreme value and Johnson SB distribution are compared by using coefficients of determination, root mean square error, sum of squares of errors and Kolmogorov-Smirnov, and compared with fewer parameters Rayleigh, Gamma distribution, lognormal and Weibull distributions, the goodness-of-fit of Nakagami obtained the optimal fitting precision among the mean values of four fitting indices at eight stations.Compared with three parameters generalized extreme value and four parameters Johnson SB distribution, Nakagami obtained the optimal accuracy at four stations. According to the statistical results of wind speed, it was found that when the average wind speed was lower than 3m/s, Nakagami obtained the highest fitting accuracy.Considering calculation and precision, Nakagami is more suitable for fitting wind speed probability distribution.