Large-scale benchmarking of wake models for offshore wind farms

We present a systematic framework for the validation and uncertainty quantification of wind farm wake models. The methodology is based on a new definition of the freestream wind speed. We apply the framework on data from 19 offshore wind farms. Our results show that the new wake model TurbOPark is overall unbiased and that the wake model error at each specific site does not depend on the mean turbine spacing. The Park model underestimates the wake loss unless a slower wake expansion than typically used is assumed. In either case, the Park model tends to underestimate the wake loss more for increasing distances between the turbines. We estimate the wake model uncertainty as less than 10% of the wake loss for the considered models.