Computation of prediction intervals of wind energy based on the EWMA and BOA techniques

Abstract

The integration of renewable energy sources, such as photovoltaic or wind power, into existing electric power systems is critical for the advancement of sustainable energy development. To mitigate the challenges of the inherent variability of wind power production, accurate probabilistic forecasting methods must be developed. In this paper, a novel method for computing probabilistic forecasts is proposed employing three techniques: the Exponentially Weighted Moving Average algorithm, Berstein Online Aggregation method and the Johnson SU-distribution. After applying these mathematical techniques, the obtained combined probabilistic forecasts prove to be both accurate and reliable. To assess the performance of the developed methodology, wind power generation databases from the Spanish electricity market are employed. Exhaustive analysis indicates that the proposed methodology outperforms both the individual experts and a uniform combination of these experts: the average error level is reduced 28 %, and the interval width of probabilistic forecasts is reduced 30 % on average.