Forecasting the wind power distribution for wind parks

Methods to forecast electricity production from wind farms in Austria for forecast horizons of 6 hours to 10 days are compared and evaluated for their potential operational installation. State-of-the-art methods documented in the literature and new methods will be implemented. They use best-guess and ensemble predictions of a numerical weather prediction (NWP) model and an archive of historical wind farm and NWP data spanning several years. Probabilistic methods will maximize the information content of the forecast.

Wind is a volatile source for producing electricity. The project developed new methods which combine output from computer models forecasting weather with novel statistical methods to reach much better results than by using only of them. Since most wind farms have been put into place, we strived to develop methods that work well with short data sets of electricity produced by the farm. As a crucial piece of additional information, the uncertainty of the forecasts is also given.A freely available package for the open source R software allows the energy trading sector to implement our results into their operational forecasting without needing to have strong expertise in atmospheric science or statistics.