Evaluation of Time Series Patterns for Wind Speed Volatilities in Anzali Meteorological Station

Abstract. One of the major problems in using wind energy is that wind-generated electricity is more unstable than electricity generated by other sources, and therefore integrating wind energy use with traditional power generation systems can be a challenge. This problem can be effectively reduced by having accurate information about the mean and wind speed volatilities. Therefore, in this paper, we model the average and wind speed volatilities using time series models. For this purpose, we examine different types of GARCH models. The data studied is the weekly average wind speed from the beginning of 2002 to the end of 2016 in Bandar Anzali meteorological station, which includes 783 observations. For modeling, first a moving average autoregressive model is fitted to the data of this station for the average wind speed, then for the residuals obtained from the model fit, we use a variety of GARCH models including GARCH models, GARCH in mean and asymmetric GARCH models. Finally, using the Bayesian information criterion among all fitted models, we select the best model for modeling wind speed volatilities at this station and conclude that the asymmetric ARMA-EGARCH model is suitable for wind speed modeling.