Modeling Heteroscedasticity of Wind Speed Time Series in the United Arab Emirates

Abstract:

There has been a growing interest in wind resources in the Gulf region, not only for evaluating wind energy potential, but also for understanding and forecasting changes in wind, as a regional climate variable. In particular, time varying variance—the second order moment—or heteroscedasticity in wind time series is important to investigate since high variance causes turbulence, which affects wind power potential and may lead to structural changes in wind turbines. Nevertheless, the conditional variance of wind time series has been rarely explored, especially in the Gulf region. Therefore, the seasonal autoregressive integrated moving average-generalized autoregressive conditional heteroscedasticity (SARIMA-GARCH) model is applied to observed wind data in the United Arab Emirates (UAE). This model allows considering apparent seasonality which is present in wind time series and the heteroscedasticity in residuals indicated with the Engle test, to understand and forecast changes in the conditional variance of wind time series. In this study, the autocorrelation function of daily average wind speed time series obtained from seven stations within the UAE—Al Aradh, Al Mirfa, Al Wagan, East of Jebel Haffet, Madinat Zayed, Masdar City, Sir Bani Yas Island—is inspected to fit a SARIMA model. The best SARIMA model is selected according to the minimum Akaike Information Criteria (AIC) and based on residuals of the model. Then, the GARCH model is applied to the remaining residuals to capture the conditional variance of the SARIMA model. Results indicate that the SARIMA-GARCH model provides a good fir to wind data in the UAE.