Build I.T. and They Will Come

When Adam Wierman joined Caltech’s faculty in 2007, he set out to find a new challenge. “I wanted to do something about a problem of fundamental importance,” he says. “Climate is the problem.”

Wierman was then purely a computer scientist, an expert in how work is scheduled and resources are allocated in the cloud. Through his quest, he would add economist and power-system engineer to his résumé.

Back then, Wierman was one of the first to recognize that data centers devour energy. Today, the I.T. sector accounts for approximately 7 percent of global electricity use. It has as big a carbon footprint as the airline industry. Streaming your favorite hour-long show can lead to as much power usage as running a refrigerator for a week.

To help clean up computing, Wierman decided to design new algorithms for the management of data centers, communication networks, and our power grid. He hoped to find ways to improve the energy efficiency of I.T. infrastructure.

Enter Jevons

But these efforts lead to Jevons paradox—a variation of “If you build it, they will come.” Economist William Jevons wrote in 1865, “It is wholly a confusion of ideas to suppose that the economical use of fuel is equivalent to a diminished consumption.” In other words, as people like Wierman make computing and the grid more efficient, we use more, out-spending the savings.

So, Wierman began thinking up ways to go beyond efficiency. He had found his challenge: Outwit Jevons paradox.

“The big picture here is smart cities, a smart grid, and control systems that allow you to incorporate intermittent renewable energy, like wind and solar,” he says. “It involves hard-core electrical engineering and hard-core algorithms. But these aren’t enough. You can’t design the control systems without thinking about the markets. And you can’t design the markets without thinking about the control systems.”

From Wires to Markets

On his second Friday at Caltech, Wierman stopped by the Athenaeum for a drink. He joined a table of economists. One of them was John Ledyard, now Caltech’s Allen and Lenabelle Davis Professor of Economics and Social Science, Emeritus.

“John and I got talking about so many things,” Wierman says. “We taught a course together four months later. He had worked on electricity markets and led me to other Caltech collaborators.”

In Wierman’s view, Caltech is the only place thinking from wires to markets.

He credits Caltech’s tradition of seed-funding exploratory, interdisciplinary projects. In his case, financial backing from two Caltech centers, both funded by philanthropists, helped him connect computer science and economics.

“Seed funding like this is a distinctive thing about Caltech,” he says. “It lets you and your students make progress without having everything already mapped out.”

Wierman also credits Caltech’s emphasis on bringing together people with diverse viewpoints, which he is working to enhance. Now executive officer of the Department of Computing and Mathematical Sciences, he is advancing Caltech Computes, an initiative that nurtures projects that integrate computer science with other fields. More broadly, he and colleagues are successfully promoting gender equality in Caltech computer science, for instance by raising philanthropic support that helped 40 Techers travel to the world’s largest gathering for women in technology. And by teaching science nights, hosting summer programs, and advising the school district’s pioneering Math Academy, the researchers are actively helping schoolchildren in the diverse city of Pasadena prepare for lives in science and engineering.

Roll-Out Time

Now, Wierman and his students are putting their Jevons-outsmarting ideas to the test.

The group’s theorems informed the design of the first net-zero-energy data center. They showed that machine-learning algorithms could manage workloads in the cloud in a way that cut reliance on the grid by 80 percent.

They also introduced “follow-the-renewables routing,” which directs work to data centers where the sun is shining and the wind is blowing—new algorithms instead of new power plants.

Then, Wierman thought, if data centers can flex their energy demands, what else can? What if warehouse store chains could save money by honing the timing of their HVAC systems? What if homeowners and HOAs could save with smart pool pumps that run when energy demand is low? Wierman’s group is quantifying the surprising effectiveness of changes like these. The team also is testing new market designs on a network of 200 web-connected devices such as wall batteries, dishwashers, and car chargers.

If the biggest energy users can balance their workloads to tap into renewables, exploit lulls, and use each other as batteries, Wierman sees potential far beyond what he first expected—renewable sources can and should be used to provide almost all the power we need.

Giving Priorities