Uncertainty propagation through an aeroelastic wind turbine model using polynomial surrogates

Juan Pablo Murcia Leon*, Pierre-Elouan Réthoré, Nikolay Krasimirov Dimitrov, Anand Natarajan, John Dalsgaard Sørensen, Peter Graf, Taeseong Kim

*Corresponding author for this work

    Research output: Contribution to journalJournal articleResearchpeer-review

    475 Downloads (Pure)


    Polynomial surrogates are used to characterize the energy production and lifetime equivalent fatigue loads for different components of the DTU 10 MW reference wind turbine under realistic atmospheric conditions. The variability caused by different turbulent inflow fields are captured by creating independent surrogates for the mean and standard deviation of each output with respect to the inflow realizations. A global sensitivity analysis shows that the turbulent inflow realization has a bigger impact on the total distribution of equivalent fatigue loads than the shear coefficient or yaw miss-alignment. The methodology presented extends the deterministic power and thrust coefficient curves to uncertainty models and adds new variables like damage equivalent fatigue loads in different components of the turbine. These surrogate models can then be implemented inside other work-flows such as: estimation of the uncertainty in annual energy production due to wind resource variability and/or robust wind power plant layout optimization. It can be concluded that it is possible to capture the global behavior of a modern wind turbine and its uncertainty under realistic inflow conditions using polynomial response surfaces. The surrogates are a way to obtain power and load estimation under site specific characteristics without sharing the proprietary aeroelastic design.
    Original languageEnglish
    JournalRenewable Energy
    Pages (from-to)910-922
    Publication statusPublished - 2018


    Dive into the research topics of 'Uncertainty propagation through an aeroelastic wind turbine model using polynomial surrogates'. Together they form a unique fingerprint.

    Cite this