Squeezing Through the Solar Bottleneck: Predicting Direct Solar Irradiance

by Gordy Slack

Solar concentrators could produce as much as 15 percent of all the energy consumed in California in the next 15 years.

California's Great Central Valley is the ideal place to develop the large-scale concentrator-type solar technology that can make a real difference in the U.S. energy crisis. There is plenty of space here, lots of sun, and only a little cloud cover. And the peak demand for energy in the middle of the day—to run air conditioners—coincides with the peak availability of energy-producing photons from the sun. Central Valley residents consume a lot of energy themselves, but there are several large urban areas close enough to benefit from power generated here. Carlos Coimbra, associate professor in the School of Engineering at UC Merced, believes that, within 15 years, solar concentrators could produce as much as 15 percent of all the energy consumed in California.

But there is one frustrating obstacle to this promising scenario: our inability to predict reliably the amount of direct solar irradiance available to the state’s energy grid at least a day or two in advance. Without that, notes Coimbra, utility companies simply cannot risk relying on this highly productive breed of collectors to produce the energy that they need to deliver.

Even a small oscillation—just 2 percent up or down in the energy grid—can cause the whole system to shut down, Coimbra says. Today, the California grid remains stable on the supply side because it is powered by very predictable fuel and hydroelectric sources. The only fluctuation is in demand: if there is a heat wave and everyone turns on their air conditioner at the same time, then there can be problems. But those demand-side fluctuations can be addressed by the reliable, albeit extreme measure of rolling blackouts. However, if the grid depends on a lot of its energy coming from solar concentrators on a given day, but the solar rays are unexpectedly blocked by aerosol or cloud cover, then the power supply can dip and take out the whole grid.

Professor Carlos Coimbra at UC Merced is working to predict solar radiation availability for energy use.

“We can put whatever money we want into solar,” Coimbra says, “but if we cannot predict direct normal irradiance for the next 24 hours, then we cannot hook these things into the grid. This is a real bottleneck.”

Direct Normal Irradiance (DNI) is the solar engineer’s term for the radiation that comes straight from the sun onto a given area. (“If you look at the sun through a paper towel tube, the rays hitting your eye are the DNI,” says Coimbra.) Much of the light we live by is not direct, but diffused, buffeted around by the atmosphere, and reflected off of the ground, buildings, and water. Although that diffused irradiation still contains energy that can be exploited by some solar technologies, it is of little use to concentrators, which rely mainly on DNI for their high efficiency.

While meteorologists are able to predict on Monday what overall sunlight— global solar irradiance (GSI)—will be available on Tuesday, they are not able to predict DNI because of the many variables that can influence it. Cloud cover has the biggest impact on DNI, but aerosol content, water vapor, carbon dioxide, and ozone all diffuse DNI too.

Last year, CITRIS gave Coimbra $75,000 to run a one-year proof-of-concept project called the Solar Irradiance Mapping Initiative (SIMI). The project is a collaboration between professors Coimbra and Qinghua Guo at Merced and Jean-Pierre Deplanque at UC Davis. The researchers will set up two experimental solar stations, one on each campus, that are equipped with highly-sensitive instruments that can independently measure DNI, GSI, and the total combination of the two (called local global irradiance.) The solar data collected at those ground stations will be correlated with several kinds of satellite image data—on weather, aerosol and ozone content in the atmosphere, and other variables—as well as ground radar information about local atmospheric conditions.

Using an approach to complex statistical associations called genetic algorithms, Coimbra will process all of that data into an increasingly accurate and predictive model. Genetic algorithms borrow concepts from the biological processes of mutation and natural selection to “evolve” mathematical “genes” that “constantly evolve and improve and adapt to different conditions,” says Coimbra. “It is a perfect application for a highly stochastic situation like this.”

If the Davis and Merced ground observatories generate the kinds of associations Coimbra expects them to, they will serve as a model for a much wider network of ground stations that could be distributed across the state, eventually creating a kind of real-time and short-term predictive geographical information system (GIS) that maps for solar irradiance. The ground stations will be used to benchmark the models developed for analyzing geo-stationary satellite and radar images, which in turn will map DNI availability throughout the state. Guo is an expert in GIS. At first, the information provided by such a system would be a key tool for both utilities and for policy makers trying to set up rational and reliable networks of sustainable energy sources that rely on solar irradiance, including concentrator technologies and, perhaps, wind turbines too. “Wind is just another kind of solar energy,” says Coimbra.

Using an approach called genetic algorithms, Carlos Coimbra and colleagues will process a variety of data into an increasingly accurate and predictive model for how much solar radiation the panels can expect to receive.

Because even small oscillations in the supply side can take down the entire grid, the predictions needs to be very accurate, says Coimbra. Gathering the wide range of data sets that independently measure GSI, DNI, and local global irradiance and then studying it for relationships can unveil less-than-obvious but very helpful associations that will boost the utilities' confidence predicting DNI.

Armed with the ability to forecast accurately, utility companies and policy makers will be able to invest confidently in California’s abundant solar energy opportunities, says Coimbra. And as the growing economic and environmental price of fossil fuels makes prices for those clean and renewable energies more and more competitive, the time is ripe for these technologies.

“California’s Central Valley may be the best place to have solar power in the U.S. Theoretically, we could build 300-megawatt power plants all over the Valley. But no one is going to support a high percentage of the grid being solar until these short-term predictions can be made on a reliable and systematic basis,” says Coimbra.

“It is not too often a researcher has the opportunity to remove a major bottleneck to an important viable technology. Concentrator technology is available, but it is not yet completely viable…not without the piece that we are trying to provide here. That, for me, is a huge motivation,” says Coimbra.