Ensemble-based estimation of wind power forecast uncertainty

Wind power prediction tools promote secure energy supply with large wind energy integration, which is fundamental for achieving a low-carbon economy. While deterministic forecasts provide a time-series, probabilistic forecasts provide interval predictions, thus enabling the perusal of forecast uncertainty for decision-making. This work presents a new nonparametric statistical approach to model the uncertainty in wind power forecasts. The methodology leverages an ensemble of deterministic forecasts using order statistics and linear quantile regression, thus providing an efficient distribution-free probabilistic power forecast system, with sample size and coverage error controlled by the required coverage rate. Experiments include daily data from three months and forecast horizons complying with market applications. The performance is evaluated using the reliability/calibration and the sharpness metrics. The results show competitive reliability, with coverage and sharpness characteristics that compare favorably with reference methods.