Ramp Skill of Reforecast Runs from HRRR and HRRRNEST Models during the Second Wind Forecast Improvement Project (WFIP2). Part II

The second Wind Forecast Improvement Project (WFIP 2) is a DOE and NOAA led multi-agency project being conducted in Columbia River Gorge and Basin in the Oregon and Washington states. The main goal of this project is to better understand the meteorology and to improve the forecast skill of NWP models in complex terrain. Since the WFIP2 study region is well-known for its excellent wind resource, many wind farms are installed there. One of the biggest challenges for wind power production is the accurate forecast of wind ramp events, i. e. large changes of generated power over short periods of time. Certain weather events that are scheduled to supply energy for the electric grid may contain wind ramps that are not forecast accurately (in terms of their directionality, amplitude or timing) and this may require large and sudden changes in conventional generation to compensate in order to balance the electrical load. These forecast issues can ultimately increase the costs of power production due to the need to rely on maintaining back-up sources of energy.

A Ramp Tool and Metric (RT&M) was developed as part of the first WFIP experiment, held in the U.S. Great Plains (Sep 2011 – Aug 2012). The RT&M was designed to measure the skill of NWP models explicitly at forecasting ramp events. In this study we apply the RT&M to 80-m wind speeds (turbine hub-height) measured during WFIP2 by 19 sodars and several lidars, as well as control and experimental reforecast runs for four 6-week periods spanning each season, using both the 3 km resolution HRRR and 750 m resolution HRRRNEST models.

While in the previous presentation (Part I) we show the bulk statistics results in terms of MAE and BIAS, in this presentation the seasonal and diurnal distribution of ramp events in both the observations and models are analyzed, as well as model skill at forecasting ramp events, including a comparison between the control and experimental runs.