Comparison of measured and simulated fatigue loads on a multi-megawatt wind turbine

Abstract

The Mann turbulence model is widely used in the design and certification of multi-megawatt wind turbines. However, these turbines operate in a region of the atmosphere where the model's assumptions are violated. One of the most significant assumptions is that of neutral stability conditions, which raises concerns about the model's accuracy for load simulations. To investigate this, we compare fatigue loads measured on a 15 MW wind turbine to simulations performed using an aeroelastic solver. The inflow was characterized using data from a meteorological mast equipped with sonic and cup anemometers. The turbulence model was fitted to measurements of auto-spectra under varying wind speeds and stability conditions, while the vertical profile of wind speed was represented by a power law. The resulting wind fields were then used as input to the aeroelastic simulations. We first present a comparison of measured fatigue loads on the tower and blades across different atmospheric stability regimes. On average, the fore-aft loads at the bottom of the tower were 98 % higher under unstable atmospheric conditions as compared to stable conditions. On the other hand, the difference in flapwise blade loads between unstable and stable conditions was 20 % below rated wind speed and -2% at higher wind speeds. These results underscore the importance of accounting for atmospheric stability in wind turbine siting and load verification campaigns. A subsequent comparison of measurements and simulations revealed that measured and simulated loads tend to fall within 3 standard deviations of each other, even under non-neutral conditions. However, the simulated fatigue loads on the tower were overestimated by a margin of 5 standard deviations under some stable conditions, likely due to incorrect predictions of the spectral coherence made by the turbulence model. Shear extrapolation based on the power law might also lead to overestimation of blade loads in the simulations. These results indicate that, despite its simplifying assumptions, the Mann model, when fitted to measurements of turbulence auto-spectra, does not introduce significant errors in fatigue load simulations for solitary multi-megawatt wind turbines.