Non-Gaussian non-stationary wind speed simulation based on time-varying autoregressive model and maximum entropy method

Abstract

Accurate simulation of non-Gaussian nonstationary wind speeds is a prerequisite for the wind resistant design of some nonlinear structures. Due to its efficiency, the time-varying autoregressive (TVAR) model has been extensively employed for simulating non-Gaussian nonstationary processes. Nevertheless, these simulation techniques based on TVAR exhibit suboptimal performance when confronted with nonstationary and highly non-Gaussian processes. Furthermore, they are unable to replicate the bimodal characteristics of specific wind speeds. This paper presents a new method for simulating univariate non-Gaussian nonstationary wind speeds using the TVAR model and the maximum entropy method. Herein, the connection between the statistical moments of input and output processes in TVAR is firstly derived. Secondly, the maximum entropy method is utilized to reconstruct the probability density function of input process and the time-varying translation function is determined. Finally, the translation process theory is applied to generate the input process, which is then input into the TVAR model to output the non-Gaussian nonstationary wind speed. The numerical results demonstrate that the proposed method exhibits superior simulation accuracy for nonstationary and strongly non-Gaussian wind speed processes. Furthermore, it is capable of capturing the bimodal characteristics of certain hardening non-Gaussian nonstationary wind speeds and possesses a broader range of applications.