This paper presents the results of the analysis of persistence of wind speed from a statistical point of view applying the concepts of the autocorrelation function, conditional probability and wind duration curves for two weather stations. The name of the stations are Puerto Bolívar and Riohacha in the Department of La Guajira, Colombia. Initial data for the analysis is from historical hourly data provided by the IDEAM during the period 2006-2012. Additionally, monthly averages wind power density was calculated to estimate the power generated by future wind farms that could be installed in the Department of La Guajira. The obtained results for the three evaluated concepts indicate that the Puerto Bolívar, located on the northern coast of La Guajira is the most suitable place to generate energy with the wind. Even when the autocorrelation function, conditional probability and wind duration have same result under high wind potential at the location studied, a noticeable difference is observed between them, because of the correlation function and the wind duration curve are based on meteorological conditions, and the conditional probability depends on the truncated speed established in order to determinate the "high wind" and "calm wind" two stages.