Wind turbine fatigue damage evaluation based on a linear model and a spectral method. Issue 7 (8th September 2015)

• Record Type: Journal Article
• Authors: Tibaldi, C.; Henriksen, L. C.; Hansen, M. H.; Bak, C.
• Is Part Of: Wind energy. Volume 19:Issue 7(2016)
• Page Start: 1289
• Page End: 1306
• Publication Date: 2015-09-08
• Subjects: wind turbine -- aeroelasticity -- fatigue damage -- spectral method -- wind turbine control -- multidisciplinary design optimization -- linear model; Wind power -- Periodicals; 621.312136
• DOI: 10.1002/we.1898
• Languages: English
• ISSNs: 1095-4244

Abstract

Abstract: Wind turbine multidisciplinary design optimization is currently the focus of several investigations because it has showed potential in reducing the cost of energy. This design approach requires fast methods to evaluate wind turbine loads with a sufficiently high level of fidelity. This paper presents a method to estimate wind turbine fatigue damage suited for optimization design applications. The method utilizes a high‐order linear wind turbine model. The model comprehends a detailed description of the wind turbine and the controller. The fatigue is computed with a spectral method applied to power spectral densities of wind turbine sensor responses to turbulent wind. In this paper, the model is validated both in time domain and frequency domain with a nonlinear aeroservoelastic model. The approach is compared quantitatively against fatigue damage obtained from the power spectra of time series evaluated with nonlinear aeroservoelastic simulations and qualitatively against rainflow counting. Results are presented for three cases: load evaluation at normal operation in the full wind speed range, load change evaluation due to two different controller tunings at normal operation at three different wind speeds above rated and load dependency on the number of turbulence seeds used for their evaluation. For the full‐range normal operation, the maximum difference between the two frequency domain‐based estimates of the tower base lateral fatigue moments is 36%, whereas the differences for the other sensors are less than 15%. For the load variation evaluation, the maximum difference of the tower base longitudinal bending moment variation is 22%. Such large difference occurs only when the change in controller tuning has a low effect on the loads. Furthermore, results show that loads evaluated with the presented method are less dependent on the turbulent wind realization; therefore, less turbulence seeds are required compared with time‐domain simulations to remove the dependency on the wind realization used to estimate loads. Copyright © 2015 John Wiley & Sons, Ltd.