Evaluation of low-frequency wind variations on wind turbine fatigue damage estimations
M. Garcia Bravo (TU Delft - Aerospace Engineering)
W. Yu – Mentor (TU Delft - Aerospace Engineering)
Taeseong Kim – Mentor (Technical University of Denmark (DTU))
S.J. Watson – Graduation committee member (TU Delft - Aerospace Engineering)
Alan Wai Hou Lio – Graduation committee member (Technical University of Denmark (DTU))
N.S. Dangi – Mentor (TU Delft - Aerospace Engineering)
More Info
expand_more
Other than for strictly personal use, it is not permitted to download, forward or distribute the text or part of it, without the consent of the author(s) and/or copyright holder(s), unless the work is under an open content license such as Creative Commons.
Abstract
Wind turbine fatigue assessments commonly represent load histories as independent 10-minute records, although slow atmospheric and operational changes can generate cycles that cross these boundaries. The resulting fatigue error and its dependence on structural response and material sensitivity remain insufficiently established.
This thesis determined how independent 10-minute processing affects fatigue estimates compared with longer continuous histories and whether the missing cycles can be recovered efficiently. Six months of load and operational measurements from a 7 MW offshore turbine were analysed for blade root flapwise and edgewise bending, and for two tower-bottom bending directions. Continuous rainflow counting, conventional independent-block processing and residue-aware low-frequency fatigue dynamics (LFFD) processing were compared across observation periods and Wöhler exponents. BHawC aeroelastic simulations compared 600 and 3600 s histories, continuous and segmented operating-state sequences, and concatenated short simulations.
LFFD reproduced the continuous reference while enabling analysis of the full six-month record. Independent 10-minute processing underestimated the flapwise damage equivalent load (DEL) by approximately 16% to 20%, and the tower bottom DELs by 3% to 20%, depending on the Wöhler exponent, while the edgewise difference remained below 0.2%. Daily residue aggregation recovered approximately 95%–96% of the full-record flapwise DEL and weekly aggregation about 99%. Cycles connecting different operating regimes contributed approximately 61% of the residue-aware flapwise fatigue contribution.
Extending stationary simulations from 600 to 3600 s increased the combined DEL by 3.62% for flapwise and 1.76% for fore-aft tower bending, whereas continuously processing an idling–startup–production–shutdown sequence increased them by 19.19% and 9.74%, respectively. Concatenated short simulations did not consistently reproduce directly simulated one-hour histories.
The results show that 10-minute storage remains practical, but processing those blocks independently can produce non-conservative estimates for aerodynamically driven responses. Residue-aware processing and targeted continuous transition simulations therefore provide more reliable fatigue estimates.