A Promising Outlook for Solar Energy Forecasting

Solar energy continues to grow in the United States, but its relative unpredictability remains a hurdle in deploying it on the grid. Now a research team is working to create detailed 36-hour forecasts of incoming energy from the sun.

The three-year effort, led by the National Center for Atmospheric Research (NCAR), is funded by a $4.1 million grant from the U.S. Department of Energy. NCAR is working with universities, utilities and other energy companies, as well as commercial forecast providers, to predict with far more accuracy and specificity when cloud cover could reduce the amount of energy coming from the sun.

More than half of all states in the U.S. have required that utilities increase their use of renewable energy, but renewables are inherently variable. The hope is that solar can follow the example of wind, which now has far more reliable forecasts from a previous NCAR effort. (Related post: “Focusing on Facts: Can We Get All of Our Energy from Renewables?“)

The team is designing a prototype system that would forecast sunlight and the resulting power every 15 minutes over specific solar facilities.

One of the biggest challenges energy companies face with solar power is the ability to anticipate how much of it will be available — and when — so that they can reliably work it into the grid. (See related: “The Big Energy Question: What to Develop Next?“)

If an incorrect forecast shows that there will be more solar energy available than there is, a utility has to buy more on the wholesale market to make up for it — likely at a higher price than they would pay if they could plan ahead for it.

“What happens when a cloud comes over and cuts the production in half, and we as an ISO [independent system operator] have to go out and procure that energy? Then when you go to buy that energy it’s like buying an airline ticket” at the last minute, so it’s more expensive, explained Jim Blatchford, who helps integrate renewable energy into the smart grid for the California ISO, one of NCAR’s partners in the project.

“If we can predict what’s going on and we can line up that generation and buy it in the future instead of in real time,” the company can save money, he said.

Likewise, if more sun than is expected produces excess solar power, that extra energy can go to waste because currently there is no cost-effective way to store it.

“It’s critical for utility managers to know how much sunlight will be reaching solar energy plants in order to have confidence that they can supply sufficient power when their customers need it,” said Sue Ellen Haupt, director of NCAR’s Weather Systems and Assessment Program and the lead researcher on the solar energy project, in a statement. “These detailed cloud and irradiance forecasts are a vital step in using more energy from the sun.”

Nick Depmer, one of the managers on the trade floor at Xcel Energy in Colorado, knows that firsthand.

“You have to be able to unload or load up other assets to fill in that void,” he said. “If you can anticipate that issue, then you can react to it. The more accurate your forecast, the better.”

NCAR worked with Xcel to create a detailed wind energy forecast that saved Xcel ratepayers an estimated $6 million in a year. But determining cloud cover accurately and specifically always has been a challenge for meteorologists, because there are different types of clouds, and they’re affected by so many factors, including wind, humidity, surface heat, atmospheric gases, and more.

Russ Bigley, a meteorologist with Xcel, said the NCAR-led research team will start with the same atmospheric model that was used for wind, and tweak it to work for solar. He said solar forecasts that look further out might be easier than those for wind, but “solar on a five-minute basis is probably going to be a lot more difficult than the wind.”

Both Bigley and Depmer said that wind forecasting had come a long way with the NCAR project, in large part because information that companies might otherwise have kept to themselves was released. They’re hopeful the same will be true for solar, but they’re not convinced better forecasting will be the ultimate game changer because of solar’s cost.

Utilities do currently use solar forecasting, but they are looking for more detail.

“We do use computer models and we push that into a solar forecast based on cloud cover,” Bigley said. “I think right now … the state of the forecasting is probably in its infancy, and that’s partly because the penetration level of solar is not that great compared with other generation assets.”

The California ISO uses some solar forecasts in the two-hour range, but “we want to get it in closer to real time,” Blatchford said.

The research team will put in place a range of observing instruments, including lidars (laser-based technology that takes measurements in the atmosphere); specialized computer models; and mathematical and artificial intelligence techniques, according to a press release. A key part of the system will be placing groups of three sky imagers in each of several locations. They will observe the whole sky, triangulate the height and depth of clouds, and trace their paths across the sky.

Researchers plan to test the system in several geographic areas and during different weather patterns throughout the year.

The forecasts would then be able to predict when, where and what type of clouds would form over a specific area, as all of those factors have a varying impact on the amount of sunlight that gets through.

The utilities and ISOs can then look at the forecast and determine, “Where’s the sun in relation to those clouds? How’s it going to hit my solar farm?” said Blatchford. “This is all really just in its infancy. We’re [becoming] a little bit smarter, a little bit more advanced.”