by Patrick Quinlan
Not long ago, wind generation was such a small contributor to overall generation that few utilities were concerned with its variability. Today, wind is becoming a significant energy source across North America. To avoid load-management migraines, independent system operators and utilities that offer wind power to their customers are focusing increased attention on wind forecasting.
According to the Global Wind Energy Council, cumulative installed global capacity has reached 76 GW, with the U.S. total at about 12.7 GW, and the Canadian total topping 1.7 GW. Over the next several years, North America is expected to add about 4 GW of new wind generation capacity annually. Wind will account for 20 percent of total generating capacity in several states in the next few years.
Wind offers many advantages to the power generation industry. After making the capital investment in a wind farm, the resource is freely available and nonpolluting. There are no fuel or transportation costs, and no wastes for disposal. From a load-management standpoint however, wind generation presents a challenge. The fuel-wind-is variable and can’t be controlled. However, interconnected wind generation is predictable and manageable using the latest forecasting technologies. Over the past few years, a half-dozen utility sponsored system integration studies, reviewed by the Utility Wind Integration Group, have concluded that wind adds a tractable (but not insignificant) cost for ancillary services to a load management portfolio. As Mike Milligan of the National Renewable Energy Laboratory recently said, “State-of-the-art wind forecasting can reduce costs. (The) majority of the value can be obtained with current state-of-the-art forecasting, and additional incremental returns [are available] from increasingly accurate forecasts.”
Wind anemometers, sonic and vane, in Golden, Colo. Photo courtesy of NREL. Click here to enlarge image
Utilities have been relying on temperature forecasts to predict daily peak demand. Now, wind forecasting is being added to the portfolio. The hardware elements of a wind forecasting system can include anemometers and wind vanes affixed to meteorological masts, weather satellite systems and locally based remote sensing systems such as lidar or sodar.
Forecasting software interprets the data from these sources to predict upcoming power production. Major wind forecasting providers in the U.S. include 3Tier, AWS TrueWind, Garrad Hassan and Windlogics. Several large system operators and utilities, such as Hydro-Quàƒ©bec, are also developing in-house programs.
Forecasting providers typically offer hour-ahead, day-ahead, and longer (two- to three-day) estimates. Each has its own proprietary approach based on computer modeling of available data from a variety of sources. Some providers rely more strongly on satellite data providing wide-ranging (mesoscale) estimates; some rely more on regional wind farm production data and local meteorological resources. The computer modeling can include artificial intelligence algorithms, such as neural-nets. With these numerical approaches, it’s common for the forecasting models to require several weeks or months to “train” using available local data.
Several large wind forecasting studies are now underway. 3Tier is conducting a major simulation of the effects of forecasting wind-based ramp rates for Bonneville Power Administration, modeling the potential effectiveness of an integrated forecasting system in the Columbia River region. In Canada, the Alberta Electric System Operator and the Alberta Department of Energy are managing a regional forecasting study involving the use of actual meteorological and wind farm data.
Conclusions are emerging from these studies to support the notion that, as Alain Forcione of Hydro-Quàƒ©bec said at the Canadian Wind Energy Association Conference in October 2007,”The development of a good forecasting system totally depends on the availability of a maximum number of observation sources. The final value of a forecasting system also depends on its compatibility with the electric system management tools and processes.”
One of the biggest recent developments has come on the hardware side of the business with new remote sensing technologies. For multiple reasons, most meteorological towers in the U.S have a maximum 60 meter (200 ft.) height. One of the most vexing problems for forecasters has been the lack of data at the height of modern wind turbines, most of which are 80 meters or taller. So new, small radar, lidar (laser-based), and sodar (sound-based) systems are filling the gap. Within that group, sodar systems are emerging as important additions to the forecasting toolbox.
Armed with these new hardware tools and ever-improving software, utilities will be able to continue adding wind power while minimizing the challenges of load management.
Patrick Quinlan, P.E., is manager of business development at Second Wind Inc., the leading provider of wind farm monitoring systems in the United States. Second Wind works with wind forecasters as providers of meteorological instrumentation and as integrators of forecasting data into major wind farm data systems. Contact Quinlan at email@example.com.