How artificial intelligence can improve advanced nuclear reactors

Researchers are still looking for the ideal properties of molten salt, which can serve as a coolant and fuel in advanced nuclear reactors. Credit: Argonne National Laboratory

Technology developed at Argonne could help narrow the search for molten salt, a new study suggests.

Scientists are looking for new materials to develop the next generation of nuclear power plants. In a recent study, researchers at the United States Department of Energy’s (DOE) Argonne National Laboratory have shown how artificial intelligence can help find the right types of dissolved salts, a key component for chemicals. advanced nuclear weapons.

The ability to absorb and store heat makes molten salt important to clean energy and the country’s climate goals. Molten salts can serve as a coolant and fuel in nuclear power plants that generate electricity without emitting greenhouse gases. They can also store large amounts of energy, which is increasingly needed on grids with variable sources such as wind and solar.

If you heat salt on your kitchen table to 801 C (1,474 F), it would melt and you would have molten salt. However, to make and save energy, not just any salt will do. Scientists have been testing different salt compounds to find the properties needed to properly cool and fuel a nuclear reactor for decades. These properties include low melting point, good compatibility, and high heat absorption capacity, among others.

Which molten salt structures will provide the desired properties for a nuclear reactor? The possible changes are almost endless. The study aims to determine how computer simulations powered by machine learning can guide and improve real-world experiments at the Advanced Photon Source (APS), a user facility for the DOE Office of Science at Argonne. The results were recently published in the journal Physical Examination B.

“We used the experimental results from APS to validate our simulation. At the same time, the simulation results gave us more information about which salts we should study more. They work together,” said Jicheng Guo, an Argonne chemical engineer and lead author of the paper. “This allows us to learn many songs at the same time.”

Researchers use powerful X-rays at the APS to better understand certain salt compounds by looking closely at their structures. But the time and cost associated with real-world testing make it desirable to limit the pool of candidates taking the test.

“The potential for molten salts is huge,” said Nathan Hoyt, an Argonne researcher and co-author of the paper. “Therefore, it is impossible to try to take test data for every possible design.”

At the 6-ID-D site, a technique called high-energy X-ray diffraction captures the patterns produced when X-rays scatter a sample of dissolved salt.

“APS is unique for these types of measurements,” said Chris Benmore, senior physicist at APS and co-author of the paper. “The powerful X-rays it produces are very useful for looking at the structure of molten liquids, glasses and amorphous materials in general.”

Machine learning involves training a computer to analyze a situation based on available data. But in this case, the researchers did not have many confirmed examples to show the pattern. Building on previous modeling that examined heat-resistant materials, the researchers used what is known as active learning to create a model that could be transferred to analyzing molten salt.

Instead of being equipped with one or two special components of the molten salt mixture, a transferable model can be used to mix throughout the compound area. The model makes assumptions based on principles; in other words, rather than a set of predetermined responses. Machine learning simulations were run using high-performance computing resources at the Argonne Leadership Computing Center (ALCF), a user center of the DOE Office of Science, and using the Bebop suite at the Center for Materials Science. of Argonne’s Laboratory Computing.

“We didn’t train the model with examples of the shape of the sweet spot, where you get the exact melting point,” said Ganesh Sivaraman, an Argonne computational scientist and corresponding author of the paper. “Our model was successful in predicting that sweet spot, even without the same training input.”

Now that researchers have shown this method can work, the next step is to work with more complex data.

“The molten salt environment is a very dynamic environment. Conditions change over time, and sometimes waste can enter the salt,” Guo said. “We want to generate a small amount of these contaminants to see if the model can predict how that affects the overall composition of the dissolved salts and their properties.”

Guo, Hoyt, Sivaraman and Benmore’s co-authors are Logan Ward, Yadu Babuji, Mark Williamson and Ian Foster of Argonne and Nicholas Jackson of the University of Illinois Urbana-Champaign.

Additional information:
Jicheng Guo et al, Potential mechanical transferability for molten LiCl-KCl salts confirmed by high energy x-ray spectroscopy, Physical Examination B (2022). DOI: 10.1103/PhysRevB.106.014209

Provided by Argonne National Laboratory

Excerpt: Hot salt, clean energy: How artificial intelligence can improve advanced nuclear reactors (2022, December 15) retrieved on December 15, 2022 from /2022-12-hot-salt-energy-artificial-intelligence. html

This document is subject to copyright. Except for any legitimate activity for the purpose of private study or research, no part may be reproduced without written permission. Content is provided for informational purposes only.

Leave a Comment