Sensitivity of Wind Power Production to Changes in Surface Roughness: Impacts on Wind Turbine Siting and Energy Production

Accurately determining the exact placement of wind turbines can be a challenge, especially given complex terrain or intricate boundary layer structures.. Improper placement of wind turbines can occur when using limited wind speed and direction information from tall tower observations because local influences such as terrain and surface heating may not be reflected by point measurements. To compensate for the under representativeness of available measurements, mesoscale numerical weather prediction models are often used to model the spatial heterogeneity of the near surface wind field because they can model complex terrain and, to a limited extent, temporal and spatial changes in boundary layer structure. A key aspect of this modeling approach is an accurate depiction of the local terrain and vegetation, as reflected by the surface roughness. However, executing a mesoscale model simulation at the horizontal resolution given typical wind turbine spacings (e.g. 200 – 400 m) comes at a considerable computational cost. Additionally, mesoscale models utilize ensemble averaging with turbulence closures which are at times inconsistent with reality and the ensemble average that the equations predict. Coupling mesoscale models with computational fluid dynamics models (CFD) to alleviate this problem has shown promising results; however, many challenges remain. A common approach utilized in wind resource assessment is to combine output from a mesoscale model with a microscale model (e.g. WAsP). In lieu of trying to resolve all small-scale terrain features, the mesoscale modeling stops at a resolution of approximately 1 – 2 km. Local wind speed analyses are then generated with a microscale model using output from the mesoscale model. Since the mesoscale model simulation provides initial conditions to the microscale flow model, it's output needs to be as accurate as possible. A current operational issue with mesoscale model is the parameterization of surface roughness values. Surface roughness is often specified using a lookup table value associated with the dominant land cover class for a particular model grid cell. These roughness values typically don't change on a seasonal basis and more importantly don't vary within a specified land cover class. This study presents results from a modeling study examining the sensitivity of model performance to surface roughness for wind resource assessment applications. New roughness values are estimated from site photography and aerial imagery (Figure 1). Results from several mesoscale model simulations over selected regions will be presented and the impact of surface roughness changes on wind energy production estimates will also be discussed.