Abstract: Based on the wave superposition technique and fast Fourier transformation (FFT), an improved method is presented to simulate three-dimensional relevant wind velocity time series for a mass of nodes of long span spatial structure. The wind Power Spectral Density (PSD) is divided into frequency spectral function and coherency function. The simulated spectrum converges upon the given target spectrum in all three directions. The time-frequency characteristics of the simulated wind velocity time series for long span spatial structure is studied by means of wavelet analysis technique, which possesses excellent localized characteristics in both time domain and frequency domain and is capable of focusing on any details of the objective signal series. The discrete wavelet transformation is introduced to decompose and reconstruct the discrete wind velocity time series. The maximum entropy method is adopted to reduce the information loss and improve computational efficiency. Results show that the present method has better accuracy and higher efficiency than traditional technique in the simulation of wind velocity time series. The reliability of the wavelet technique for the analysis of wind velocity time series is also proved through numerical examples.