Optimization of wind power forecasts for wind farms, network nodes, control zones and Germany. The project used methods such as machine learning, combining deterministic and probabilistic weather forecasts.
1st September 2014 to 31st December 2016
The aim of this research project was to significantly improve the operational wind power forecasting system, jointly operated by the project partners in several areas, thereby reducing the forecast error to a level not previously achieved.
A new combination of weather models and machine learning methods improved the forecasting quality of wind power forecasts for different aggregation levels (wind farms, geographical areas, Germany). Both deterministic and probabilistic prediction systems were developed and tested within the framework of the project.
The RMSE standard for installed capacity day-ahead forecast for Germany for 2014 was reduced from 4.1% to less than 3.1%, zone and park forecasts were also improved.
Center for Solar Energy and Hydrogen Research Baden-Württemberg (ZSW)