Improved wind direction measurement through blade loads

A previously described observer of wind misalignment is validated using field test data collected on NREL's CART3 wind turbine. The observer uses blade root bending moment 1P harmonics, computed using the transformation of Coleman and Feingold, to infer the rotor-equivalent relative wind direction. The observation model parameters are determined by a least squares fitting using recorded blade loads and met-mast measured wind direction and speed; a random sample consensus (RANSAC) algorithm is used to robustify the parameter estimation procedure while detecting outliers in the experimental samples. The observer is validated using an independent verification data set: recorded blade bending loads are fed to the observer and the estimated wind misalignment is compared to both the one provided by the met-mast, assumed as the ground truth, and by an on-board nacelle-mounted wind vane. The validation is limited in this work to the full load regime, because of very noisy data and high turbulence in the available time histories in the partial load region. Results show that the rotor-equivalent wind misalignment estimates provided by the proposed observer are well correlated in the low frequency spectrum with the met-mast reference, and in general in much better accordance with it than the on-board wind vane measurements.