Posters

Abstract

In the assessment of the energy yield, the wind flow modeling at the site is key. More and more wind farms are planned in terrains with heavy or commercial forestry, making the modeling even more complex. EDFr has assessed the performance of two wind flow models, linear and non-linear models, on four forested projects. The study aims at challenging the current practice of its software for EDFr and at defining updated good practices for such sites. It also aims at assessing the performance of the two above mentioned software in forested terrains. The four selected sites are spread worldwide in France, UK, Finland and Chile. Thus, they have different diurnal and seasonal behaviors, as well as very different atmospheric stabilities. They all have at least two installed met masts with a common measurement period and common measurement heights. The distance between the masts ranges from 500m to 7km. The sites have been chosen for their relatively flat terrain and homogenous forestry cover. The performance of the models was evaluated through cross-predictions between the masts. Several cases were investigated: * Linear flow model with displacement heights. The displacement height is taken as ¾ of the tree heights of the surrounding trees. It also includes a weighting based on the wind rose and on the distance between the trees and the masts. This case is taken as the reference case. * Non-linear flow model, with atmospheric stability calibration only. This calibration is based on the comparison of average vertical profiles between measurement and model. * Non-linear flow model, with atmospheric stability and roughness calibrations. The roughness calibration is based on neutral atmospheric conditions (as suggested by the software editor) and completed prior to the atmospheric stability calibration. * Non-linear flow model, with atmospheric stability and roughness calibrations, also including displacement heights. This case is not in line with the usual recommendations, but has been included as regularly questioned internally. In each case, the discrepancies between the actual measurement and the predicted wind speeds at the common measurement heights have been analyzed. Even if carefully conducted, uncertainties remain in the analysis, about measurement, orography, roughness and tree heights. The atmospheric stability calibration has been identified to be key for improving the performance of the non-linear model. However, even though the roughness and stability calibration improved the results with respect to the stability calibration only, the effect of the roughness calibration was not as significant as expected. Finally, it has been seen that the software were performing in a comparable way, making them both suitable for forested terrains. Discrepancies remain, which highlights the need for additional uncertainties for such sites. Further investigations are still ongoing in order to identify possible additional tuning that would help improving the results. Extending the project sample with additional forested sites in simple terrain would also participate in comforting the proposed outcomes. Also, a further step would consist in considering a time series for the atmospheric stability, or at least a different stability per direction.