Relevance of Uncertainty Modelling for Wind Turbine Lifetime Estimations
Item Type:Conference Paper
Citation:Clemens H�bler, Sarah Wosko, Raimund Rolfes, Relevance of Uncertainty Modelling for Wind Turbine Lifetime Estimations, 14th International Conference on Applications of Statistics and Probability in Civil Engineering (ICASP14), Dublin, Ireland, 2023.
submission_176.pdf (PDF) 6.162Mb
For the design of wind turbines, the determination of the fatigue lifetime is essential. Frequently, so-called lifetime "damage equivalent loads" (DELs) are used as a measure for the fatigue lifetime. Short-term DELs are determined by analysing time series calculated using aero-elastic simulations. To take uncertainties in environmental conditions ﾖ which always exist due to scattering of wind and wave conditions but also due to long-term changes in environmental conditions ﾖ into account, usually a large number of stochastic aero-elastic simulations is carried out for different combinations of wind speeds, wave heights and wave periods (and possibly other environmental conditions). Subsequently, the short-term DELs of all simulations are combined resulting in a lifetime DEL. However, this more or less deterministic procedure considers uncertainty only implicitly. Other non-deterministic approaches taking uncertainties explicitly into account, e.g., probabilistic approaches based on Monte Carlo simulations and joint statistical distributions of all relevant environmental conditions, are available but not state of the art. One reason why non-deterministic simulation approaches are rare is their high computing time. However, recently there was significant progress in using meta-models, e.g., Gaussian Process Regressions (GPRs) or Artificial Neural Networks (ANNs), for the short-term DEL approximation. Since short-term DELs can be quickly approximated using meta-models and, therefore, a huge number of short-term DELs can be taken into account when computing a lifetime DEL, the computational effort of non-deterministic approaches is manageable by now. Therefore, in this contribution, various uncertainty models, e.g., probabilistic, interval, fuzzy, p-box, are used to model the uncertainty of the environmental conditions explicitly. To keep the analysis simple, here, the wave height is the only environmental condition, which is assumed to be uncertain. All other environmental conditions are set to deterministic values depending on the wave height. For example, the wave period is defined as a deterministic function of the wave height according to current standards. Surely, in general, other environmental conditions or combinations of various environmental conditions could also be modelled as uncertain. Measurement data of several years of the research platform FINO3 (https://www.fino3.de/en/) in the North Sea are used to derive the various uncertainty models. The uncertainty is propagated though the simulation model, here a GRP, yielding to uncertain short-term DELs, and therefore, lifetime DELs. The influence of the various uncertainty models on the resulting lifetime DELs is investigated. This provides some insight in the relevance of selecting a suitable uncertainty model when estimating wind turbine fatigue lifetimes.
Other Titles:14th International Conference on Applications of Statistics and Probability in Civil Engineering(ICASP14)
Type of material:Conference Paper
Series/Report no:14th International Conference on Applications of Statistics and Probability in Civil Engineering(ICASP14)
Availability:Full text available