Bringing the wind to the grid

NCAR helps Xcel Energy generate renewable energy and slash costs

In February 2011, the Public Service Company of Colorado (PSCo) did something unprecedented. It turned off several coal-fired power plants for a long weekend, based on the high levels of wind generation forecasted for that weekend.

wind farm
A wind farm near Grover, Colorado, that is among those that provide power to Xcel Energy. (©UCAR. Photo by Carlye Calvin. This image is freely available for media & nonprofit use.)

"The concept of turning off a reliable, coal-fired plant for a long weekend in favor of a variable and difficult-to-predict source of energy such as wind energy would have been unthinkable just a couple of years ago," says Drake Bartlett, a renewable energy analyst at Xcel Energy, PSCo's parent company. "But now we have a new capability."

That new capability is a partnership between Xcel Energy and NCAR that has produced a wind forecasting system that is widely considered, nationally and internationally, to be one of the most advanced of its kind. The highly accurate forecasts it generates mean that Xcel can anticipate when it will have enough wind energy to meet customer needs, enabling it to power down traditional coal- and gas-fired plants and thereby save money. The system, which came online in September 2009, has already reduced wind energy prediction error by 40% and, in 2010 alone, saved Xcel Energy’s ratepayers over $6 million.

chart
Sample output from NCAR's wind energy prediction system for Xcel Energy. The time moves ahead from left to right, with hours (in UTC) shown along the horizontal axis at bottom and power output (in megawatts) shown along the vertical axis at left. The green dots at left represent actual power generated in megawatts up to the initial forecast point of 2130 UTC (2:00 pm Mountain Time) on February 1, 2010. The solid line shows the predicted power in megawatts for the next 36 hours. Orange shading around the solid line represents the typical forecast error from the preceding seven days. The actual wind should fall within the orange prediction zone about 75% of the time. Operators can choose to display data by turbine, farm, or larger region, and by time up to 120 hours in the future. (©UCAR. Image courtesy NCAR and Xcel.This image is freely available for media & nonprofit use.)

"The system has dramatically improved our wind generation forecasting accuracy," Bartlett says. "This allows Xcel to confidently manage our electric portfolios more efficiently—as in the case of the PSCo turning off the base-load coal plants—and it also allows Xcel to aggressively install more wind generation."

Wind may blow freely, but harnessing its energy is neither cheap nor easy. Utilities, which often balance a portfolio of different power sources, need accurate wind forecasts in order to know when to power on and off the traditional power plants that back up wind energy. If a forecast is wrong, a utility may need to purchase energy from the spot market, which is very costly.

In late 2008, NCAR scientists and software engineers teamed up with Xcel, the nation’s lead utility in wind energy, to tackle the wind forecasting challenge. A technique called numerical weather prediction, used by meteorologists for weather forecasts and research, underlies their approach. The system they created for Xcel works by pulling data from the company's nearly 3,000 wind turbines across the United States and injecting it into a nested grid of the atmosphere to make high-resolution forecasts of details as small as how a hill or valley within a single wind farm may affect energy production at a particular turbine. The system then generates wind energy forecasts every 15 minutes out to three hours and on an hourly basis out to 168 hours.

William Mahoney
William Mahoney. (©UCAR. Photo by Carlye Calvin. This image is freely available for media & nonprofit use.)

Underlying the system is a wealth of basic scientific research carried out by NCAR scientists on atmospheric dynamics, numerical weather prediction, computer modeling, data assimilation, statistical techniques, and more. This expertise is critical to forecasting wind, since landscape features such as hills and trees can affect wind speed and direction, as well as cause wind shear and turbulence in ways that may greatly influence the amount of energy that is produced. In addition, few weather forecasting models are designed to predict wind at about 200 feet, where Xcel’s turbine hubs are typically located.

 “Xcel Energy has been very proactive in adding wind energy to its system, but one of the major obstacles is the difficulty in predicting when and how strongly winds will blow at the locations of turbines," says William Mahoney, the NCAR program director on the project. "Every fraction that we can improve the forecasts results in real savings.”

City at nightIn their own words

“Never has so much real-time data been leveraged in an operational wind energy forecasting system. This is truly groundbreaking work.”

—Eric Pierce, Xcel Energy

“NCAR has addressed a critical gap in the development of applied atmospheric science expertise for wind energy and transfer of that expertise to the wind energy industry.”

—Michael Knotek, Renewable and Sustainable Energy Institute, University of Colorado Boulder and National Renewable Energy Laboratory

“This project demonstrates that improved forecasting of wind energy is feasible. It improves system reliability, and it makes wind energy more economically competitive with traditional energy sources.”

—Melinda Marquis, NOAA Earth Systems Research Laboratory