Extended Abstract (72K)5B.3
Utility Sector Wind Power Forecasting: Status and Measurement Needs
Marc Schwartz, National Renewable Energy Laboratory, Golden, CO; and E. Ela
The installed capacity of wind power plants in the United States is expected to reach 30 Gigawatts by the end of 2009. This level of penetration into the electricity grid will emphasize the importance of wind forecasting to utilities and Independent System Operators for its application in electricity markets and system operations. Wind power forecasts use a two-step process. First wind speed forecasts at the wind plant site are derived using output from a numerical weather prediction model and/or a statistical model. The forecasted wind speed(s) are then converted to power output for the wind plant in megawatts using conversion methods that account for factors such as the power production curve(s) of the turbines, wind plant layouts, and wind direction. Typical power forecasts produced for clients include day-ahead hourly averages, and hourly or more frequent same day average forecasts used for unit commitment and dispatch of non-wind resources. Inputs for these targeted forecasts usually include historical power output from the aggregate wind plant, and basic meteorological data from at least one location on site in addition to the data from the global and regional numerical models. Often, ensemble model packages are used to produce confidence intervals for the wind forecasts.
Organizations that operate electricity markets in California, Texas, New York, and the upper Midwest, to name a few, have selected wind forecasting vendors to produce wind power forecasts for wind plants in their region. Also, individual utilities such as Xcel Energy in Colorado, and the Bonneville Power Administration in the Pacific Northwest faced with rapidly increasing penetration of wind onto their systems have been especially aggressive in implementing wind forecasting programs. Some of the mentioned organizations are quite interested in a specialized type of forecast that predicts wind power “ramps”. These are rapid increases or decreases in wind power output over a short period of time (30 to 120 minutes) that operators need to know to reliably operate the system. A significant question facing utilities is whether setting up a network of offsite real-time boundary layer measurements from anemometers and remote sensing equipments such as SODAR and LIDAR will improve the accuracy and value of one-to-six hour wind power forecasts plus characterizing local weather events. Anecdotal evidence in the Pacific Northwest demonstrated that additional offsite observations have increased the accuracy of the short-term wind speed and wind power forecasts but no public data is yet available that can quantify any increase in forecast accuracy. Xcel Energy in Colorado is working with the National Center for Atmospheric Research on a large field campaign to evaluate how high resolution atmospheric data can be used to develop a robust and cost-effective wind characterization program including more accurate forecasts. The National Renewable Energy Laboratory is aiding the Xcel project by conducting research on converting the wind speed forecast into an accurate wind plant production forecast. This 2-year project started in late 2008 and the initial results are expected to be released in about one year.
Session 5B, Societal Impacts of Weather Forecasts
Tuesday, 2 June 2009, 1:30 PM-3:00 PM, Grand Ballroom West
Previous paper Next paper
Recorded presentation