AI Controlling Nuclear Reactors? It Could Happen

The federal government awarded a $3.4 million grant to a research consortium to explore the potential applications for the technology at nuclear power plants.

by Ray Gronberg, The Herald-Sun / June 7, 2018
Shutterstock

(TNS) — RALEIGH, N.C. — North Carolina State University researchers are heading up a consortium that's just received $3.4 million from the federal government to develop control systems for nuclear power plants that lean more heavily on artificial intelligence.

The grant comes from ARPA-E, the U.S. Department of Energy's Advanced Research Projects Administration-Energy. The agency is putting $23.9 million into 10 projects it hopes can find uses in a new generation of nuclear power plants. The award to the N.C. State-led group is the second-largest amount awarded.

"The nuclear industry in the last few years has started gaining a lot of momentum in advanced reactor technology," said Rachel Slaybaugh, a University of California-Berkeley nuclear engineering professor who's also the APRA-E project director overseeing the grants. "The space has become active and it was ripe for trying to bootstrap it."

At Norht Carolina State, the federal grant is going to a group headed by professor Kostadin Ivanov, the head of the university's Department of Nuclear Engineering.

But while Ivanov is the project's lead investigator, he's hardly alone. The team includes people from New Mexico State University and Ohio State University, the Oak Ridge and Idaho government nuclear laboratories, and two private-sector companies, TerraPower and Zachry Nuclear Engineering.

TerraPower is ultimately the key to the project, as it's trying to develop a new type of sodium-cooled "traveling-wave reactor" that can burn a wider range of nuclear fuels than traditional units like Duke Energy's Shearon Harris plant that use ordinary water in their cooling systems.

Advocates contend that such a design should ultimately be safer and more economical than traditional, Harris-like units that aren't all that different, conceptually, from the first reactors the U.S. developed in the 1940s and 1950s for the military.

TerraPower's leaders, who include multi-billionaire Microsoft co-founder Bill Gates, intend to build a prototype reactor in China by the middle of the coming decade.

As far as ARPA-E is concerned, the N.C. State-led project is appealing because it should eventually feed into TerraPower's work, and because the team also includes Zachry, a company with offices in Cary and Charlotte that writes software for commercial nuclear power plants and "really understands that space," Slaybaugh said.

Ivanov said the project's initial "case study" will use yet another type of reactor, and the system ultimately "should be applicable to multiple reactor types."

The basic idea is to apply AI and machine-learning techniques to go through the mountains of data that come from a reactor, spot patterns in it and call them to the attention of the unit's human operators.

Designers will combine simulation and real-world data, comparing scenarios from each to develop "confidence [in] what they can predict and what is the range of uncertainty of their prediction," Ivanov said, adding that, "at the end, the operator will make the final decisions."

Ivanov added that the security of a new system is an obvious concern, and that's one of the reasons computer-science specialists from the universities and the two national labs will be involved.

But the broader reason to develop new controls is to lower the operations and maintenance costs of a reactor, by making it easier among other things to plan maintenance and run the facility with a smaller staff, Slaybaugh said.

The N.C. State-led project should also help prospective operators like TerraPower figure out "what sensors do you need to put where" for the new control system to work, she said.

"We'd like to see how you prove some of these concepts, then a company like TerraPower can pick it up and use it in their designs," she said.

ARPA-E is the Department of Energy's analogue for the Pentagon's Defense Advanced Research Projects Agency, the fabled organization that among other things funded early work on what became the Internet.

Federal officials created and launched it in the late 2000s, late in former President George W. Bush's tenure and early in former President Barack Obama's. Like its Pentagon counterpart, its job is to fund risky but promising research and technology development.

President Donald Trump and his staff, however, called for eliminating the agency in their fiscal 2018 and 2019 budget proposals, arguing that the private sector has the "primary role in taking risks to commercialize breakthrough energy technologies with real market potential." Trump's fiscal 2019 budget would have closed ARPA-E by mid-2020.

But Congress hasn't gone along, and in fact raised allocations to ARPA-E this spring.

APRA-E exists to fund projects that, while promising, are generally in an early stage of development and still have "too much technical risk for the private sector to pick up," Slaybaugh said.

"We're trying to de-risk it," she added.

©2018 The Herald-Sun (Durham, N.C.) Distributed by Tribune Content Agency, LLC.