The efficient integration of wind in the energy market is limited by its natural variability and predictability.This limitation can be tackled by using the probabilistic predictions that provide accurate deterministicforecasts along with a quantification of their uncertainty. We propose as a novelty the application of ananalog ensemble (AnEn) method to generate probabilistic wind power forecasts (WPF). The AnEnprediction of a given variable is constituted by a set of measurements of the past, concurrent to the pastforecasts most similar to the current one.

The AnEn performance for WPF is compared with three state-ofthe-science methods for probabilistic predictions over a wind farm and a 505-day long period: a wind powerprediction based on the ensemble wind forecasts from the European Centre for Medium-Range WeatherForecasts (ECMWF) Ensemble Prediction System (ECMWF-EPS), the Limited area Ensemble PredictionSystem (LEPS) developed within the COnsortium for Small-scale MOdelling (COSMO-LEPS) and aquantile regression (QR) technique.

The AnEn performs as well as ECMWF-EPS, COSMO-LEPS and QRfor common events while it exhibits more skill for rare events. A comparison with the performances obtainedwith a deterministic forecasting method based on a Neural Network is also carried out showing the benefitsof using AnEn.