Posters

Abstract

Optimal exploitation of wind farms requires having good estimates of the expected long-term wind speed (WS) average at hub-height. Firstly, prospective sites are identified on a regional level by analyzing wind atlases; Secondly, in-situ measurements are carried out to provide a finer analysis of the wind potential and to determine the most cost-effective wind farm size (Sempreviva et al., 2008). In contrast to sites located on land, sea WS measurements at hub-height are scarce since installing and operating an offshore mast or a floating lidar is expensive. Moreover, they cannot provide any information on horizontal gradients. Badger et al. (2016) proposed a new methodology for assessing offshore WS. It combines Synthetic Aperture Radar (SAR) and the Weather and Research forecast (WRF) numerical model using a long-term averaged Monin Obukhov (MO) extrapolation formula which is valid only for the surface layer. From SAR 10m WS estimates were derived the friction velocity and roughness length that aim to better account for air sea interaction. Results found showed that WS estimates of the combination has similar accuracy than the WS estimates of WRF; yet, the combination has the same horizontal resolution as the SAR (1 km) and the same vertical resolution as WRF. Nevertheless, Badger et al. (2016) method did not preserve individual SAR samples and consequently some hypothesis is needed to build the WS probability density function. We propose a new method inspired from Badger and al (2016), with two main novelties allowing preserving the individual SAR sample. The first novelty consists in using a vertical extrapolation formula valid for the entire atmospheric boundary layer (ABL) (Peña et al. 2008). This formulation requires an assessment of the friction velocity and roughness length both derived from SAR 10m WS as well as ABL height, MO length and stability function derived from WRF. Our second innovation consists in improving the accuracy in the determination of WRF ABL height, MO length and stability function by combining an ensemble of 14 forecasts. We validate our methodology by comparison with 2016 wind observations from one undisclosed offshore site for 5 heights varying between 34m and 104m. We compared the raw WRF WS corresponding to the mean of the 14 ensemble members, hereafter WRFRA, to the wind estimate (hereafter WRFAX2) obtained by applying the extrapolation formula to the SAR WS, using the ensemble mean of ABL height, MO length and stability function. Results found show that: the MBE improvement of WRFXA2 over WRFRA ranges between 0.5 m/s and 0.2 m/s. We also found that a MAE improvement of WRFXA2 over WRFRA is found at 53m and 70m AMSL; it ranges between 0.10 and 0.14 m/s. For higher height it may be considered negligible. The major implication of this study is demonstrating the ability of SAR observations at 10 m to improve wind mapping from coarser NWP models estimates at turbine hub height. We downscaled successfully WS from WRFRA at 3km grid resolution to WRFXA2 (1 km grid resolution) without downgrading the accuracy of WS estimates.