LAUSR

Wind power forecasting is presently one of the challenging tasks to deal with supply-demand balance in modern electric power systems. Accurate wind power predictions are needed to reduce the risk in uncertainty and enable for better dispatch, scheduling, and power system integration. This article deals with the challenge of wind power forecasting by proposing the application of the forecasting methodology using the wavelet packet decomposition principles, neuro-fuzzy systems, as well as the benefits of data preprocessing and forecast combination framework. The used data consist of the quarter-hourly observations of wind power generation in France, and the proposed method is used to ensure forecasts for a time horizon of an hour-ahead. The obtained results indicate the superior accuracy of the proposed model with an average mean absolute percentage error of 3.408%, which means there is possibility to construct a high-precision method using only the historical wind power data.