Presentations

Abstract

This contribution describes a methodology and use case to identify heterogenous flow patterns in a  wind farm in highly complex terrain. The investigated onshore site in Turkey consists of ca. 80  turbines with three different turbine models. The turbines are located on the ridge of a  peninsula. The orography steeply climbs to ca. 500 m above sea level, with inter turbine  elevation differences of up to 300 m. Two years of SCADA (supervisory control and data  acquisition) data are available in 10 min format. The data was filtered and the yaw error of the  turbines was corrected with the FLASC toolbox [1]. The analysis of the turbine production  shows that the farm experiences strong wake effects, highly heterogeneous wind conditions as  well as diurnal effects.  The modelling approach is based on the “wind farm as sensor” method, a combination of data-driven and explicit modelling [2]. The resulting, site-specific model can be applied e.g. for AEP  estimation, a model for project repowering or forecasting applications [3]. The baseline flow is  provided by an engineering wake model, which is augmented with an unknown background  flow correction field. The whole farm is then used as a distributed sensor, through the turbines  operational SCADA data. Synthesized as long-term observations, the data is used to  simultaneously learn the parameters that describe the correction field and to tune the ones of  the engineering wake model. The method showed good agreement to CFD simulations in  previous studies, however in more moderate terrain complexities [2]. As the different turbine  models operate on different hub heights, estimating the correct wind shear is a further source  of model error.  The identified flow fields reveal the presence of significant terrain-induced effects from all  investigated wind directions. To validate the corrections, high fidelity CFD RANS simulations  of the site are available [4]. They resolve the peninsula with the domain inlet over the open  sea and feature a range of atmospheric conditions for surface heat flux. The results show that  for the investigated most frequent northern and southwest wind directions, both methods  agree on the local speed up effects.  The applied SVD method further allows to visualize the corrective measures in terms of  eigenflowfields. The higher order flow fields correct smaller, local orographic effects. As for  previous cases, it showed that the first eigenflowfields correct large trends in the wind farm  background flow, e.g. a north-south wind speed increase.   [1] FLASC. Version 2.0.1 (2024). Available at NREL/flasc.  [2] Braunbehrens, Robert, Andreas Vad, and Carlo L. Bottasso. "The wind farm as a sensor:  learning and explaining orographic and plant-induced flow heterogeneities from operational  data." Wind Energy Science 8.5 (2023): 691-723.  [3] Braunbehrens, R., et al. "Site-specific production forecast through data-driven and  engineering models." Journal of Physics: Conference Series. Vol. 2767. No. 9. IOP Publishing,  2024.  [4] Alletto, M., et al. "E-Wind: Steady state CFD approach for stratified flows used for site  assessment at Enercon." Journal of Physics: Conference Series. Vol. 1037. No. 7. IOP Publishing,  2018.