How a Car Racing the Outback Can Boost Solar in the US

University of Michigan students are using IBM's new weather forecasting tool, which could help them win a race across Australia and eventually make solar power a bigger part of America's energy grid.
Image may contain Building Office Building Electrical Device Solar Panels City Town High Rise and Urban
The IBM researchers are happy to help out a crew of young engineers chase the sun through Australia, but they’ve got bigger goals: increasing the amount of solar power going into our electrical grid.Jon Simon/Feature Photo Service for IBM

Later this month, 30 experimental race cars powered only by the sun and built entirely by college kids will take off on the World Solar Challenge, an epic 1,800-mile race across Australia that draws some of the best engineering students in the world.

Seventeen of those students are from the University of Michigan, and they've spent about a month Down Under preparing their car, a sleek, futuristic machine called Aurum. Michigan is a top-tier solar racing team and a veteran of this race, but with five third-place finishes, it's tired of being an also-ran. It wants the title.

And it's looking to IBM for help winning it.

The team tapped Big Blue and its new weather forecasting tool, which ramps up forecast precision by combining on-the-ground measurements with a deep archive of past data. The idea is to determine exactly when and where the team can optimize its performance, using technology that solar power plants can use to predict and manage production, making our aging grid more flexible and adaptable to renewable energy.

The Car

The Michigan crew is one of 30 descending on Australia from around the world. They'll set off from Darwin, on Australia's northern coast, on October 18 an spend one week racing south to Adelaide, on the opposite side of the continent.

The route will take them through a vast, punishing desert on a journey of four to seven days, depending on the weather and how Aurum performs. Like everyone else, the team will camp each night, and though they're bringing a portable grill, they're ready to eat a lot of PB&J along the way.

The World Solar Challenge is a biennial event, so the team's had some time to improve on its last racer, Generation. This year's car is more aerodynamic, the result of hundreds of iterations tested in computer simulations and a wind tunnel. Aurum resembles a catamaran on wheels, its "hulls" covering the wheels and its "deck" covered with 65 square feet of solar panels. The 5 kilowatt-hour lithium-ion battery and DC brushless motor are quite capable of propelling the carbon fiber car—which weighs 550 pounds with the driver—to highway speeds, but don't expect to see it in the fast lane.

Being light and sleek trump comfort, so the drivers—who have to be small and light to fit inside—will be only slightly more comfortable than Mad Max strapped to the front of a war machine. "They have enough room," says Pavan Naik, a rising senior from Herndon, Virginia and the team's project manager.

Forecasting

Weather prediction models aren't new to the race, and Michigan uses high tech (checking online forecasts) and low tech (looking up to scan for clouds) tools when crafting its strategy. IBM brings a bit more sophistication, and some big data, to the process.

The company's two-pronged approach, under development for more than two years, includes new hardware and a whole lot of data.

The first prong is the sky camera, which is exactly what it sounds like. It uses a wide-field camera and a pyrometer to measure solar irradiance, which you probably call sunshine. It's a measurement of how much of the sun's power is reaching a given point, depending on things like cloud cover and haze, and directly translates to how much energy a solar panel can produce.

It also packs all the computing power needed to analyze photos of the sky and determine how that will change over time. Taken together, this data indicates how much sunshine the team is getting at that exact moment, and how that will change in the next 10 to 15 minutes. It's not so different from sticking your head out the window, but it's a more quantitative, and thus reliable, approach.

That hardware fills a canister about three feet tall, which will be strapped to the roof of a support vehicle driven by Ted Van Kessel, Sr., the IBM researcher who did most of the development work on the unit. That included making the mounting system, which he tested by slapping it on his Prius and doing what he calls "stand on the brakes" tests.

The second prong is Big Data. IBM gathers past weather predictions from various agencies around the world—up to 10 years' worth in some spots—and compares them with the actual temperature, cloud cover, and so on, for that time and place. That gives it error rates, which in turn let the system interpret new predictions, and tweak them accordingly. For example, the researchers could look at a forecast of 80 degrees for northeast Utah. If they know that over the past six years, the temperature there at that time of year has been 5 degrees hotter than predicted, they increase their own estimate. Based on testing in the US, the system has outperformed predictions by the National Oceanic and Atmospheric Administration by 10 percent, says Lu.

Together, these approaches cover short-, immediate, and long-term weather forecasting, making each more accurate. All that info's neatly translated into clear predictions of how much power the car will generate, and handed over the to the race team.

"This is a concrete application of big data," says researcher Siyuan Lu.

The system's not perfect, of course. It yields more nuanced predictions, provides more data. For the UM students, the result is, effectively, a far more accurate fuel gauge.

Race Strategy

Since arriving in Australia last month, the Michigan team has assembled Aurum and spent lots of time shaking everything down. Driving the car is but a small element of the operation. The students also must manage a caravan of support vehicles carrying gear and transmitting data, not to mention the logistics of keeping everyone fed and rested.

The cars are powered by batteries, constantly being recharged by the sun, via solar panels. The bulk of the race is through the heart of Australia, where sunshine's a near guarantee, so the strategy is straightforward—go as fast as you can, as long as you can, because solar power's plentiful. For Leda Daehler, a senior from Portsmouth, Ohio, the fun really begins on the last day, as teams approach the coast and cloud cover becomes a bigger issue. She's the head strategist, and her job, she says, is "figuring out how fast you can drive."

"That day is the really big gamble," Daehler says. If you set a course expecting sunshine, then find clouds, you must quickly decelerate to make sure you've got enough battery power to cross the line. That can cost you up to 80 minutes, Daehler says, in a race that's usually decided by an hour or two. Advance notice that it's going to be cloudy means you can slow down earlier, conserve energy, and avoid a dramatic drop in speed later on.

That matters, because moving at a constant speed is the most efficient way of using precious battery power. And the more accurate your forecast of oncoming weather—and thus how much power you can generate—the better you can plan to make sure you're using everything you've got, and not leaving anything on the table.

Better data means the UM crew can make decisions other teams can't, Naik says. "It could give us the race."

Beyond the Outback

The IBM researchers are happy to help out a crew of young engineers chase the sun through Australia, but they've got bigger goals: increasing the amount of solar power going into our electrical grid.

An increasing portion of our power is coming from the sun. In 2005, solar photovoltaic setups produced just 16,000 megawatt hours of electricity each month. This year, that number will hit 15.9 million megawatt hours. That's still just 0.4 percent of what Americans consume in a year, but the Energy Information Administration expects that share to double within two years.

That's good news for the planet, but it presents a problem for utilities. In the US, the generation of electricity is inextricably linked to its consumption. Utilities don't really store electricity—which is a whole other problem—so they produce exactly as much as they need at any given moment.

As major solar installations take on a great role in electricity production, the threat of unexpected clouds becomes serious. "One good cloud bank can really hit you hard," Van Kessel says. "If you're managing a power grid and you suddenly have the equivalent of either a 1,000-megawatt nuclear plant, or a 1,000-megawatt coal plant drop offline, that's a major loss." That power must be replaced almost immediately to meet consumer demand, but you can't just turn on a nuclear or coal plant at a moment's notice. The more accurately you can predict what a solar plant will produce on any given day, the more carefully you can prepare backups.

In July, IBM announced it would provide its solar forecasting data to government agencies and other organizations for a limited time, to show off the value of the technology. But come October 18, the team's attention will shift to Australia, as the Michigan students race into the outback.

The students can't predict whether the forecasts will give them an edge, let alone the win. But they'll be happy to have the additional information . And once that information spreads to help use sunshine to power our homes as well as experimental race cars, lots of other people should be grateful as well.