Identification and Analysis of Forecast Model Large Error Events During WFIP2

The second Wind Forecast Improvement Project (WFIP2) took place in the Columbia River Basin of Oregon and Washington states from Fall 2015 until Spring 2017. The project had an extensive field campaign with over 100 instruments and a model development component, aimed at improving short-term wind energy forecasts. Included in the instrumentation were 18 sodars located throughout the region, measuring vertical profiles of wind speed from 10 to 200 m for the duration of the project. Using all 18 sodars and the operational High-Resolution Rapid Refresh (HRRR) model, the 80-m (wind turbine hub height) winds were converted to power using a standard turbine power curve. The 40 events with the largest aggregate forecast power errors at forecast hour 12 were identified, and a meteorological analysis was performed to identify the conditions under which these events occurred. The WFIP2 observations, as well as model analysis fields, are used to categorize the conditions for the large error events. The meteorological regimes that were identified as contributing to large model errors included mountain waves and topographic wakes, stable layers and cold pools, and easterly flows.