Wind speed deficits downstream offshore wind parks – A new automised estimation technique based on satellite synthetic aperture radar data

Abstract

Wind speed deficits behind offshore wind parks in the German Bight are estimated from satellite synthetic aperture radar (SAR) data using a new filter technique. The deficit computation requires knowledge about the undisturbed wind field, which is derived by a two-dimensional (2D) convolution filter tailored to the geometry of the wake. Both the wind direction and the size of the wind farm are taken into account. The most relevant scale for the wind speed deficit estimator (WISDEM) is the width ξ$\xi$ of the wake. Unlike approaches used so far, the proposed technique is suitable for a full automisation of the estimation process. Furthermore, the rigorous definition of the method and the reproducibility of the results can help in the consistent analysis of big data sets and the meaningful intercomparison of different geographic study areas. The filter is applied to Sentinel‑1 SAR data demonstrating the ability of the method to quantify and visualise wind speed deficits in a very efficient way. The method also allows the study of the 2D structure of wakes, in particular curved shapes, which are found frequently. A statistical wake analysis is performed for one year of data showing the most frequent occurrence of wakes during the spring and summer seasons. According to mast measurements taken at the FINO‑1 platform, this period is characterised by relatively strong atmospheric stability. Error estimates are derived for WISDEM wind speed deficit estimates based on a 2D spectral analysis of a Sentinel‑1 SAR data set acquired over one year. The impact of the wake filter on the background wind spectrum is quantified by application of the convolution theorem. The deficit estimation error is shown to increase with decreasing deficit values and with increasing wake width. The error is most sensitive to spectral components with wavelength in the across wake direction near 2ξ$2\xi$. The slope of the derived wind spectra is very close to the Kolmogorov k-5∕3$k^{-5/3}$ law, at least down to wave length of about 3 km. A significant dependence of the spectra on the atmospheric stability was found with energy levels increasing with instability. This relationship is beneficial for wake estimations, because wakes are more likely to occur in stable conditions, where relatively homogeneous background wind fields lead to reduced deficit estimation errors.