Articles | Volume 20
17 Jul 2023
 | 17 Jul 2023

Decentralized forecasting of wind energy generation with an adaptive machine learning approach to support ancillary grid services

Lukas Holicki, Manuel Dröse, Gregor Schürmann, and Marcus Letzel

Cited articles

Abels, A., Fleßner, T., and Holicki, L.: A field testing of ancillary services for grid restoration with wind power plants, ETG Kongress, Kassel, Germany, 25–26 May 2023, ISBN 978-3-8007-6108-1, 2023. 
Becker, H., Schütt, J., Schürmann, G., and Spanel, U.: The SysAnDUk-project: Ancillary services provided by distributed generators to support network operators in critical situations and during system restoration, Wind Integration Workshop,, 11–12 November 2020 (Virtual event), 2020. 
Becker, H., Valois-Rodriguez, M. F., Holicki, L., Malekian, K., and Gartmann, P.: Evaluation of wind power plants' control capabilities to provide primary frequency support during system restoration, International Conference on Smart Energy Systems and Technologies (SEST), Vaasa, Finland, 6–8 September 2021,, 2021. 
Gomes, V., Wang, Y., Breton, A., Mourier, M., Holicki, L., and Letzel, M.: Provision of FCR reserve by wind power plants: capability and performance assessment based on experimental results, Wind Integration Workshop, 11–12 November 2020, ISBN 978-3-9820080-8-0, 2020. 
Holicki, L., Abels, A., Nikolai, S., Schürmann, G., Schauerte, U., Schmidt, T., and Flessner, T.: Employing wind power plants in grid restoration processes- a field testing, Wind & Solar Integration Workshop, Den Haag, Netherlands,, 2022. 
Short summary
We present a wind power forecasting procedure, that consists of a physics-based component generated at a central server, and a data-based component generated on-site the wind power plant (WPP). It provides blackout-robust data transmission to grid operators and high forecast reliability, especially in the very-short term horizon. This endeavor aims at employing WPPs for support in exceptional or critical grid situations, where short term decision making is most relevant.