Forecasting wind speed in peninsular Malaysia: An application of Arima and Arima-Garch models

• Main Authors: Hussin, Nor Hafizah, Yusof, Fadhilah, Jamaludin, 'Aaishah Radziah, Siti Mariam Norrulashikin, Siti Mariam Norrulashikin
• Format: Article
• Published: Universiti Putra Malaysia Press 2021
• Subjects: QA Mathematics
• Online Access: http://eprints.utm.my/id/eprint/94965/; http://dx.doi.org/10.47836/pjst.29.1.02

In the global energy context, renewable energy sources such as wind is considered as a credible candidate for meeting new energy demands and partly substituting fossil fuels. Modelling and forecasting wind speed are noteworthy to predict the potential location for wind power generation. An accurate forecasting of wind speed will improve the value of renewable energy by enhancing the reliability of this natural resource. In this paper, the wind speed data from year 1990 to 2014 in 18 meteorological stations throughout Peninsular Malaysia were modelled using the Autoregressive Integrated Moving Average (ARIMA) to forecast future wind speed series. The Ljung-Box test was used to determine the presence of serial autocorrelation, while the Engle’s Lagrange Multiplier (LM) test was used to investigate the presence of Autoregressive Conditional Heteroscedasticity (ARCH) effect in the residual of the ARIMA model. In this study, three stations showed good fit using the ARIMA modelling since no serial correlation and ARCH effect were present in the residuals of the ARIMA model, while the ARIMA-GARCH had proven to precisely capture the nonlinear characteristic of the wind speed daily series for the remaining stations. The forecasting accuracy measure used was based on the value of root mean square error (RMSE) and mean absolute percentage error (MAPE). Both ARIMA and ARIMA-GARCH model proposed provided good forecast accuracy measure of wind speed series in Peninsular Malaysia. These results will help in providing a quantitative measure of wind energy available in the potential location for renewable energy conversion.