An efficient simulation algorithm for non-Gaussian stochastic processes

Abstract

Using linear filter method and Johnson transformation, an efficient algorithm for simulating non-Gaussian wind load is developed in this study. Simulations with target first four-order marginal moments and power spectral density are produced by filtering the underlying non-Gaussian white noise that generated by Johnson transformation into non-Gaussian process using linear filter, such as autoregressive (AR) or autoregressive moving average (ARMA) model. For AR and ARMA models, the explicit relationships between the first four-order marginal moments of the input and those of the output are ascertained, thereby the moments of the underlying non-Gaussian white noise input can be calculated. Different from the traditional probability-dependent incompatibility for existing algorithms, a distinctive type of incompatibility which is characterized by spectrum-dependent for the new simulation algorithm is discovered and discussed. In addition, the incompatible ranges of the two types of incompatibilities are compared. We investigated several numerical examples and they demonstrated that the first four-order moments, PSD, and correlation functions of the generations closely matched the targets.