Hier finden Sie wissenschaftliche Publikationen aus den Fraunhofer-Instituten.

How to construct a reliable ensemble forecast?

Presentation held at 10th German Wind Energy Conference (DEWEK) 2010, Bremen, 17/18 November 2010
: Dobschinski, J.; Lange, B.; Wessel, A.; Bremen, L. von

Fulltext urn:nbn:de:0011-n-4067107 (247 KByte PDF)
MD5 Fingerprint: bb4c802c9815d2d53c02733b664cad52
Created on: 6.8.2016

2010, 4 pp.
German Wind Energy Conference (DEWEK) <10, 2010, Bremen>
Presentation, Electronic Publication
Fraunhofer IWES ()

Regarding the development of a wind power management system that is utilizable for the prediction of wind power feed-in under offshore weather conditions the predictability of offshore wind speeds as the main impact parameter are investigated. A multi-model ensemble prediction system (EPS) is used to generate deterministic and probabilistic wind speed forecasts which are compared with measurements from the offshore research platform FINO 1. A multi-model EPS offers the possibility to provide a high-quality forecast with a reliable estimation of the potential uncertainty on top. In the first part of this study a method is presented that upgrade the EPS in a way that the final ensemble average forecast is not only BIAS-corrected but also optimized in terms of the RMSE. The resultant EPS is used to generate probabilistic wind speed forecasts (quantiles and prediction intervals). Due to the fact that the first probabilistic results are not reliable two calibration methods are applied that lead to an improvement of the EPS consistency and to reliable quantile forecasts and prediction intervals.