On improving wind-turbine hub-height wind-speed forecasts

To improve the forecasts of wind speed at the height of wind turbines, I used numerical weather prediction (NWP) models for two types of experiments. One experiment tested the value of better horizontal and vertical resolution in the model, so as to directly forecast the winds at grid points within the height range swept out by the turbine blades instead of interpolating from coarser grid points. The other experiment tested whether forecasts for a cluster of horizontal locations near the wind turbine could yield a better cluster-ensemble average wind forecast than using winds from the one grid point closest to the wind turbine. Three case-study days of poorly forecast fast winds at the Dokie wind farm in northeastern British Columbia were chosen for this study. Dokie is located in the eastern foothills of the Rocky Mountains, where the prevailing westerly winds cause strong mountain waves and downslope windstorms at the wind farm. It was found that better wind speed forecasts are possible for some of the cases when both better vertical and horizontal NWP grid resolutions are used, because these resolutions better capture the dynamics of the mountain weather. It was also found that the cluster-ensemble average wind speeds improve accuracy by compensating for overly-smoothed terrain that NWP models use to maintain numerical stability.Science, Faculty ofEarth, Ocean and Atmospheric Sciences, Department ofGraduat