Evaluating the Ability of the HRRRv3 and HRRRv4 to Forecast Wind Ramp Events in the Great Plains

Wind ramp events (WREs) are generally defined as rapid changes in wind speed over a short duration of time. Accurate WRE forecasts are becoming increasingly critical as countries shift towards renewable energy sources, including wind power. Energy providers with wind power resources rely on day-ahead wind forecasts from high-resolution numerical models to determine how much of the energy demand can be provided by wind power versus other sources (coal, natural gas, solar, etc) to meet the anticipated energy demand. An unexpected WRE can result in either an over or underprediction of wind-generated power, which have costly economic impacts on the energy provider. In the United States, the High-Resolution Rapid Refresh version 4 (HRRRv4) has been praised for its ability to predict WREs.

This study attempts to compare how well the HRRRv4 (which became operational in December 2020) predicted WREs versus the HRRRv3 (operational from mid-2018 to December 2020) across the Great Plains in the United States from 1 January 2020 to 31 December 2022. Ideally, this analysis would use observations at 80 meters above ground level (i.e., wind turbine hub height) along with HRRR forecasts at that level. However, due to a lack of observations at this altitude, 10 meter wind speed observations from the METAR network and 10 meter wind HRRR forecasts were used instead. It is assumed that the ability of the model to capture a WRE at 10 meters is highly correlated with the model’s ability to capture a WRE at 80 meters. A variety of statistics over several spatial, temporal, and WRE characteristics like up versus down ramps, intensity, and timing will be analyzed. Spatial patterns include comparing the different subregions of the Great Plains; temporal patterns include grouping results by forecast lead time, model initialization time, season, and time of day.