|
Title:
|
|
Extreme loads using measured and calculated responses for offshore applications
|
|
|
|
Author(s):
|
Peeringa, J.M.; Bierbooms, W.A.A.M.
|
|
|
|
Published by:
|
Publication date:
|
|
ECN
Wind Energy
|
26-10-2009
|
|
|
|
ECN report number:
|
Document type:
|
|
ECN-E--09-053
|
ECN publication
|
|
|
|
Number of pages:
|
Full text:
|
|
56
|
Download PDF
|
Abstract:
An extreme load extrapolation is performed using calculated and measured loads from one of the research turbines at the ECN Wind turbine Test site Wieringermeer (EWTW).
The extreme loads are extrapolated using two different approaches:
- The method based on unconstrained 10-minute time series
- The constrained wind simulation method
For the unconstrained method a comparison is done between measurements and simulations for free stream and wake conditions. Attention is paid to the selection of the extreme value distribution and the amount of data needed. Based on the study the following can be concluded:
- For the global data approach, using one maxima from every time series, fifty maxima are sufficient to estimate the extreme load distribution conditional on the wind speed,
- There is quite a difference between the calculated and measured loads for the free stream, partial wake and full wake condition. The calculated flat blade root bending moment is smaller compared with the measurements, while the calculated fore-aft tower bottom bending moment is larger than the measured values.
An alternative method to derive at extreme loads of wind turbines is by application of a so-called constrained stochastic simulation. Simulations with a linear model are considered and just done for demonstrating the probabilistic method. A side effect is that a theoretical expression of the
distribution function is available so this may serve as a reference case of fit procedures to an extreme value distribution.
The estimated 50-year response is 0.960.02 according to the analysis with unconstrained simulations and 0.990.01 based on constrained simulations. The theoretical value equals 1. So, constrained stochastic simulation outperforms normal, unconstrained simulation: the 50-year estimate is closer to the theoretical one, the uncertainty margin is less and the required computational effort is just half.
Back to List