Fast simulation of large-scale non-stationary wind velocities based on adaptive interpolation reconstruction scheme

Spectral representation method (SRM) is the most classical one for the simulation of wind velocity. It is inefficiency when applied to large-scale non-stationary wind velocities with large simulation points. There are two reasons: numerous Cholesky decomposition and summation of Trigonometric terms. In order to improve the efficiency while ensuring accuracy, two aspects of work have been in this paper. (1) An adaptive interpolation-enhanced scheme is devised, which uses "average resolution" as the quantization index. This scheme can automatically realize the non-uniform distribution of interpolation points in two dimensions of time and frequency simultaneously, and improve the accuracy of interpolation. (2) The non-stationary wind velocities were reconstructed in time, frequency and space domain. Firstly, interpolation in time and frequency domain is directly applied to the H matrix, then proper orthogonal decomposition (POD) technology is introduced to decouple the wind velocities at spatial interpolation points, so as to obtain the time-dependent principal coordinates and space-dependent intrinsic mode function (IMF). Finally, IMF is reconstructed in the space domain to obtain the complete wind velocities. The above methodology is carried out to a super high-rise building containing 100 wind velocities simulation points and, results show that the proposed approach saves about 88% of the computational time compared with the classical SRM; saves about 47% of the computational time compared with the time-frequency interpolation based method. This paper achieves the rapid construction of large-scale non-stationary wind velocities.