Verification and analysis of hub-height wind forecasts from the NCAR-Xcel WRF-RTFDDA

In support of Xcel Energy wind power integration, NCAR has implemented an operational high-resolution RTFDDA (Real-Time Four Dimensional Data Assimilation and forecasting) system and an ensemble RTFDDA system over the western-central states. The high-resolution deterministic RTFDDA system has been running since May 2009 with a 3.3 km fine mesh domain covering the Rocky Mountains from New Mexico to Montana, the western Midwestern States, and also most parts of the Central/South Plains. The RTFDDA model supports continuous data assimilation and 3 hourly forecasting cycles, and in each cycle it provides 15 minutely forecasts at wind farms up to 24 hours on the fine-mesh domain and up to 72 hour forecasts on the coarse grid domains. In this study, the RTFDDA hub-height wind forecasts have been verified against the wind observations at selected wind farms located in Minnesota, Colorado and Texas. We have found that the model forecast errors can differ dramatically between these wind farms. And within a given farm, the forecast errors can vary significantly with meteorological conditions, diurnal evolution, and seasonal transitions.

In our analysis we have developed a number of algorithms for evaluating the high-resolution WRF-RTFDDA forecasts of wind and wind power. Comparisons of WRF-RTFDDA forecasts with GFS and NAM models have been conducted in conjunction with wavelet-based scale separation analysis and verification with the archived time-series observation and forecasts data. The result suggests a number of opportunities to improve the NWP forecast for different weather regimes and geographic locations.