Short-term forecasting of wind speed based on recursive least squares

Abstract

To properly manage the variability of wind generation, this paper presents an adaptive procedure for short-term forecasting of wind speed based on recursive least squares. Firstly, hourly wind speed data are transformed to make their distribution approximately Gaussian and standardized to remove the diurnal nonstationarity. Then, the procedure fits an AR model to the standardized transformed hourly wind speed data. Finally, the parametric AR model is regularly updated during online operation by a recursive least squares algorithm. The hourly wind speed data from a wind power site located in Hong Kong validate that the adaptive AR model can effectively forecast wind speed for horizons up to a few hours ahead.