Scientists creating advanced computer simulation of nuclear reactor
Oak Ridge National Lab leads effort using world's fastest supercomputers to create virtual reactor
A team of four Energy Department national laboratories, three universities and three energy companies will receive up to $122 million over the next five years to develop a highly accurate computer simulation of a light water nuclear reactor.
The Consortium for Advanced Simulation of Light Water Reactors (CASL), led by the Oak Ridge National Laboratory, is the first of three Energy Innovation Hubs expected to be announced this year by DOE as part of a plan to increase federal research and development spending on clean energy. The consortium’s goal is to create a virtual reactor that can be used to improve the performance and safety of existing reactors and to help design the next generation.
“Before you go testing a new reactor design, you need to have that design validated against existing reactors,” said Douglas Kothe, director of science at Oak Ridge’s National Center for Computational Sciences and director of the consortium. “What we are building is a modern, high-fidelity simulation tool, rigorously validated, to push the state of the art.”
To simulate the complex physics and chemistry of the reactor core, the consortium will have access to three of the most powerful supercomputers in the world: Jaguar, a 2,331 trillion-operations-per-second Cray computer and Kraken, a 1,029 trillion-operations-per-second Cray computer, both at Oak Ridge, and Roadrunner, a 1,375 trillion-operations-per-second IBM computer at Los Alamos National Laboratory.
At the time CASL made its funding proposals, these were the three fastest supercomputers in the world, as determined in the TOP 500 biannual list. But in the most recent list released last week, China’s Nebulae-Dawning supercomputer entered the rankings in the number two spot, just behind Jaguar.
Core members of CASL are Oak Ridge, Los Alamos, Idaho and Sandia national naboratories; the Massachusetts Institute of Technology, the University of Michigan, and North Carolina State University; and the Electric Power Research Institute, the Tennessee Valley Authority, and the Westinghouse Electric Co. A dozen other universities, companies and organizations in this country and in Europe also are contributing to the program.
CASL is DOE’s Nuclear Energy Innovation Hub, an effort to use computer modeling and simulation to advance reactor design and engineering. It will receive up to $22 million this fiscal year and up to $25 million each of the next four years of the program. The program could be extended for another five years. Other innovation hubs expected to be announced this year will focus on solar energy and on retrofitting existing structures to improve energy efficiency.
Most of the country’s 140 active reactors are second-generation facilities, some of which have been in operation for 40 years or more. A number of proposals have been made for third-generation designs. Accurate modeling and simulation could help not only in development of those designs, but also could improve operation of current facilities, which have to operate within conservative parameters because of the lack of detailed understanding of core conditions in a reactor. Better understanding could enable current reactors to safely remain in operation longer, make more efficient use of nuclear fuel and produce less waste, and increase power output.
The current fleet of reactors has been able to increase output by about six gigawatts by improving efficiencies, Kothe said. “We think that there is five or six, or even 10 or 20 additional gigawatts out there in the existing fleet,” which would be the equivalent of adding another five or 10 reactors.
DOE has a long history of modeling nuclear reactions. The need to simulate nuclear explosions to take the place of live weapons testing is one of the reasons the department’s national labs have developed the world’s fastest supercomputers. There is a large existing body of software that will be used in creating the virtual reactor, Kothe said.
“By no means does it make sense to start from scratch,” he said. “There is too much that has been done in the past 40 or 50 years. We have good software to start with.” The consortium’s first task will be to take existing code that models a handful of phenomena and begin integrating it to enable more complex simulation. “The key is to simulate multiple simultaneous phenomena that are non-linear and interactive,” he said.
The consortium hopes to release the first version of the virtual reactor this year and to improve it over the next four years.