Measuring the Skill of Numerical Weather Prediction Models at Forecasting Wind Ramp Events during WFIP

The first Wind Forecast Improvement Project (WFIP) was a DOE and NOAA-funded one year long observational, data assimilation, and modeling study aimed at demonstrating improvements in the accuracy of wind forecasts for wind energy generated by the assimilation of additional observations. The model we focus on in the analysis presented here is the 13-km-Rapid Refresh (RAP). To measure the impact generated by the additional observations controlled data-denial RAP simulations were run for six separate 7-12 day long periods (for a total of 55 days) over different seasons.

We present the results of applying a Ramp Tool and Metric (RT&M), developed during WFIP, for measuring the skill of the RAP model at forecasting wind ramp events.  This tool identifies ramp events in the time series of power, matches in time each forecast ramp event with the most appropriate observed ramp event, computes the skill score of the forecast model incorporating both phase (timing) and amplitude errors. An important feature of the RT&M is that it computes the ramp skill integrating over a range of changes of power (Δp) and periods of time (Δt) values.  Results include the impact of assimilation of the special WFIP data, as well as variations in model skill between seasons, daytime versus nighttime, and up-ramps versus down-ramps.