Based on the Detrended Fluctuation Analysis (DFA) method, this paper focuses on the long-range correlation characteristics of wind speed time series observed by 103 wind measuring towers. The results show that: (1) The wind speed time series at different heights have almost consistent scale behaviour, regardless of the resolution. (2) For the 6-hour average wind speed series, the DFA index α has a numerical range basically of 0.55 to 0.91 for all 103 wind towers, showing strong long-range persistence. (3) For the 10-minute average wind speed series, taking the 24-hour scale as the boundary, the DFA curve shows two obvious independent scaling intervals: on the larger time scale, the numerical size of scaling index α is 0.80, while α is about 1.38 on the smaller time scale. This study quantifies the long-range persistence characteristics of wind speed using the nonlinear time series analysis method. It reveals the physical mechanism behind it, which is of great significance for building wind speed models and accurately predicting wind speed.