Wind Energy Prediction System

Accurate wind forecasts are crucial for power-grid integration and load balancing. Current wind-forecasting methods, which are primarily based on statistical algorithms that use wind-farm observations, have insufficient accuracy beyond a few hours into the future. Numerical weather prediction (NWP) models have historically only played a secondary role in providing 0–12 h wind-power forecasts because the model products have had relatively low horizontal resolution and the models have not been customized for wind-power applications. 

The Real-Time Four Dimensional Data Assimilation (RTFDDA) and forecasting system, that has been developed by RAL to satisfy the meteorological needs of Army test ranges, has been adapted for wind-energy prediction. In 2009, in collaboration with Xcel Energy, NCAR implemented an operational RTFDDA system over the western and central states for supporting wind-power forecasting. This system contains three modeling domains with grid sizes of 30, 10 and 3.3 km. The 3.3 km domain covers the Rocky Mountains from New Mexico to Montana, the High Plains states, and most areas of the Central Plains. The system runs with a 3-hour cycle. In each cycle it produces 27-hour forecasts for the innermost domain and 72-hour forecasts for the two coarser domains. The real-time weather forecast maps and power-production forecasts for the ~30 wind farms in Colorado, Minnesota, New Mexico and Texas are provided to Xcel operational centers.

The boundary-layer wind is one of the most difficult variables for NWP models to predict. This is especially true for complex terrain regions. RAL’s ongoing research for improving such NWP model forecasts includes the optimization of physical-process parameterizations, the development of more-effective methods for the assimilation of meteorological measurements from wind farms, and the post-processing of model forecasts to remove errors. Furthermore, to quantify model-forecast uncertainties, RAL is implementing for Xcel a high-resolution (10 km grid increment) ensemble RTFDDA prediction system. Beginning in early 2010, this ensemble system will run for the same geographic region, and in parallel with, the 3.3 km deterministic system.

Cheng W. Y.Y., Y. Liu and T. Warner, 2009: Sensitivity of a Simulated Winter Storm to WRF Model Physics over Complex Terrain. 10th WRF users’ workshop, Jun 20-23, 2009. Boulder, CO.

Liu, Y., W. Wu, G. Roux, T. Warner, F. Chen, S. Swerdlin and J. Boehnert, 2008: A Successive Downscaling LES Simulation of Local-scale Weather Circulations. Wind Energy Meteorology, AGU Fall Meetings. Dec. 20 – 22, 2008. San Francisco.

Liu, Y., Yuewei Liu, W. Wu,W. Cheng, T. Warner, and K. Parks, 2009: Simulating intra-farm wind variation using the NCAR WRF-RTFDDA-LES model. 10th WRF users’ workshop, Jun 20-23, 2009. Boulder, CO.

Liu, Y., T. Warner, S. Swerdlin, W. Mahoney and the RTFDDA Team, 2009: Operational Wind Forecasting Using the NCAR Real-Time Four-dimensional Data Assimilation (RTFDDA) System. Preprint of Meeting on Short-term Wind Forecasting, International Energy Association. 11-12 September 2008. Madrid, Spain.

Liu, Y., T. Warner, S. Swerdlin, W. Mahoney and the RTFDDA Team, 2009: Operational Wind Forecasting Using the NCAR Real-Time Four-dimensional Data Assimilation (RTFDDA) System. Preprint of Meeting on Short-term Wind Forecasting, International Energy Association. 11-12 September 2008. Madrid, Spain.

Roux G., Y. Liu, L. Delle Monache, R.-S. Sheu and T. T. Warner, 2009: Verification of High-resolution WRF-RTFDDA Surface Forecasts over Mountains and Plains. 10th WRF users’ workshop, Jun 20-23, 2009. Boulder, CO.