You’re reading an outdated article. Please go to the recent issues to find up-to-date content.
To highlight the variability of the New England climate, Mark Twain once wrote, “If you don’t like the weather here, just wait a few minutes,” and the phrase has since been applied to regions across the United States. For most, such unpredictability presents little more than periodic wardrobe challenges, but it can mean expensive system disruptions for electric utilities that are more dependent on renewable solar and wind resources. As a result, advanced weather forecasting is becoming a serious research topic.
For the last decade or so, electric utilities’ solar investments have focused on rooftops. The extent of such investments could mean 10 photovoltaic (PV) panels on a home or a much larger number covering a warehouse facility. Still, it has not been enough for any individual installation to affect overall grid conditions in a significant way. Now, though, utilities in some regions are installing centralized solar plants capable of generating hundreds of megawatts (MW), and even passing clouds can affect their performance.
In January, for example, Southern California Edison signed power-purchase contracts totaling more than 800 MW of solar-generated electricity, with 711 MW of the total coming from only three facilities—one with a planned capacity of 325 MW—to be completed by 2016.
When plants reach this size, the consistency of output becomes critical to overall grid stability, and knowing when clouds may force the need for supplemental power also can help both utilities and merchant power producers manage costs and revenues.
Like the rest of us, plant and grid managers count on different kinds of forecasts—or, rather, different forecast windows—for planning purposes. According to Manajit Sengupta, a senior scientist at the National Renewable Energy Laboratory (NREL), solar-forecast windows can be categorized as less than an hour, the next several hours and 24 hours ahead. The first two shorter time periods affect immediate operations and dispatch requirements, while the day-ahead view aids power-purchase planning. Which particular forecast is more important to a utility depends on its unique mix of generation resources and fallback options.
“One issue is how well a utility can respond. How flexible is its backup?” said Tom Stoffel, a group manager at the NREL in Boulder, Colo., and the leader of a panel discussion NREL hosted in June as part of a two-day solar-forecasting workshop. Knowing clouds will be passing over a 50-acre array in 15 minutes might be helpful if a natural-gas plant is able to come online quickly to help make up for reduced solar output. However, if a utility depends on slow-starting coal plants for backup, it may need to call on expensive spot-market suppliers.
Similarly, solar-plant owners need to know how much power they’ll be able to supply to wholesale buyers. Falling short on promised capacity can be very expensive for plant owners. For example, wind suppliers in some areas of California now face penalties of $1,000 per MW of unsupplied—yet promised—generating capacity, according to Jim Blatchford, a renewable integration specialist with the California Independent System Operator and another speaker at the NREL workshop.
Of course, when forecasting for solar power, one is really forecasting the possible appearance of clouds, as Sengupta said. The more accurate a cloud prediction, the more accurate you’re going to be in your solar forecast, and each forecast window requires different technologies for accomplishing this task.
For forecasts looking just an hour ahead or less, researchers now are studying the use of sky cameras installed around a solar plant to image the motion of clouds across the sky. The NREL is funding research at the University of California, San Diego, to better understand how this technology might be applied.
For the next block of time—say, five or six hours ahead—Sengupta said weather satellites are now favored. These eyes in the sky take images at a 1-kilometer resolution every 30 minutes, then computers use wind data to project cloud-movement patterns. The NREL is funding efforts at Colorado State University and the State University of New York, Albany, to build out predictive data models.
Beyond six hours, and into the day-ahead time period, weather models supplied by the National Oceanic and Atmospheric Administration (NOAA) are the best resource. Sengupta said NREL and the Department of Energy are working with NOAA to gain knowledge of the variables required to formulate more accurate solar and wind forecasts.
Making large-scale solar power work economically will depend on gaining greater forecasting expertise, Sengupta said, because system operators need to understand how much capacity they can depend on for meeting base-load requirements. Where questions remain, spinning reserves—the spare capacity that must be synched to the grid and available within 10 minutes’ notice—must be maintained at higher levels.
“The more renewables you have on the grid, the more spinning reserves you have to maintain,” Sengupta said. “The cost for integration goes down with a better forecast.”
ROSS is a freelance writer located in Brewster, Mass. He can be reached at [email protected].