KIWi -Hybrid Parameter Identification with Invertible Networks

KIWi Project Logo

Researcher: Hannes Albers, Tobias Kluth (PI)
Project funding: Bundesministerium für Bildung und Forschung (BMBF)
Project sponsor: Deutsches Elektronen-Synchrotron DESY
Partners: Prof. Dr. Kathrin Flaßkamp (Universität des Saarlandes), Dr.-Ing. Tobias Meyer (Fraunhofer-Institut für Windenergiesysteme IWES), Mariam Paktiani (P.E. Concepts GmbH), Dr. Martin Piechnick (marpitec GmbH)

In the near future, many wind turbines (WTs) will reach the end of their design life, typically 20 years. An extended operating life would increase the energy yield of each turbine and thus significantly reduce the energy-related greenhouse potential. An analysis of continued operation can be performed based on model-based lifetime estimation, but this requires adapting a generic WT model to enable good model performance while maintaining low complexity to enable efficient simulations. In the joint project KIWi, data- and model-based methods for model correction and for parameter identification in inexact models are used to enable a more accurate determination of the estimation of the load in wind turbines and thus to extend their possible lifetime.

The University of Bremen participates in KIWi with the subproject Hybrid Parameter Identification with Invertible Networks, in which the connection of Deep-Operator-Networks and invertible networks for the purpose of parameter identification are investigated.