Characterizing Nonstationary Wind Speed Using the ARMA-GARCH Model

Abstract

This paper aims to accurately calculate the time-varying standard deviation of nonstationary wind speed by modeling wind speed as a time-varying mean wind speed plus a uniformly modulated nonstationary process and applying the autoregressive moving average–generalized autoregressive conditional heteroscedasticity (ARMA-GARCH) model to analyze the time-varying standard deviation of the wind speed. This paper also proposes a convenient and practical first-order difference GARCH method for calculating the time-varying standard deviation of uniformly modulated nonstationary processes that can decrease the computation time. The applicability of the ARMA–GARCH model, first-order difference GARCH method, and existing common methods for calculating the time-varying standard deviation of nonstationary wind speed are verified and analyzed using numerical simulations. The results show that the ARMA-GARCH model and first-order difference GARCH method are superior to the existing common methods. Finally, with the combination of the ARMA-GARCH model and first-order difference GARCH method, the nonstationary wind characteristics (considering the nonstationarities of mean wind speed and standard deviation) of Typhoon Chan-hom are investigated and compared with the results only taking into account time-varying mean wind speed.