High-resolution simulations of flow over terrain of varying complexity for wind energy applications

This study examines the performance of WRF, a mesoscale atmospheric model, using high horizontal resolution for wind power forecasting applications. Current operational wind energy forecasting models are run at horizontal resolutions of approximately 3 km. The objective here is to investigate whether there are improved results at higher resolution, and whether these improvements depend on the complexity of the site topography. This could allow computational resources to be saved if coarser resolution gives adequate results under certain conditions. Simulations are performed at two wind farm sites in Western North America, one with relatively simple terrain and one with very complex terrain. Two 48 hour test cases are chosen for each site, one a synoptically forced event and the other locally driven. The domains are refined from mesoscale to finer scales using grid nesting to adequately resolve topography and turbulence in the atmospheric boundary layer. The results show that the complexity of terrain and type of forcing may greatly influence results at various spatial resolutions. Results over simpler terrain show that the use of increased horizontal resolution, vertical resolution or using one- vs. two-way nesting yields minimal differences in comparisons of wind speed, shear, direction, and turbulence for both synoptically or locally forced event cases. It appears that for this particular domain the key topographic features are adequately resolved at even the coarser resolutions, so that only changes to the physical parameterizations in the model such as soil moisture or landuse have any significant effect on the simulation results. In contrast, over complex terrain, results prove to be quite sensitive to the particular model configuration, including grid resolution and nesting choices for the locally driven event. However, this sensitivity appears to greatly diminish for the synoptic event. Given proper model configuration, comparison with field observations of hub-height wind speed and direction is quite favorable at both wind farm sites and shows promise for improvements in wind forecasting over complex terrain sites. This study illustrates the need for care when simulating flow over complex terrain, and also shows that computational resources can be saved when working over simpler terrain or even in some cases with a synoptically forced event over complex terrain.