Virtual models enable real-time decision making for next-generation nuclear reactors

Digital twins are a virtual copy of a real-world system. They are a transformative tool that can assist scientists across numerous disciplines. Researchers at the U.S. Department of Energy’s (DOE) Argonne National Laboratory have created a digital twin technology that could make nuclear reactors more efficient, reliable and safe. This technology uses advanced computer models and artificial intelligence (AI) to predict how reactors will behave, helping operators make decisions in real time.

“Our digital twin technology introduces a significant step toward understanding and managing advanced nuclear reactors, enabling us to predict and respond to changes with the required speed and accuracy,” said Rui Hu, Argonne principal nuclear engineer and co-author of a recent paper published in Nuclear Technology about the accomplishment.

A virtual model that thinks in relationships

Digital twins allow scientists to monitor and predict how small modular reactors and microreactors will act under different conditions. The Argonne team developed a new methodology and applied it to generate digital twins for two types of nuclear reactors: the Experimental Breeder Reactor II (EBR-II) and a new type of reactor, the generic Fluoride-salt-cooled High-temperature Reactor (gFHR). While the EBR-II is no longer in operation, a digital twin was developed for it as a test case, helping validate the simulation models.

The key to this digital twin technology is graph neural networks (GNNs), a type of AI. These are advanced computer models that process data structured as graphs—a collection of nodes and edges representing interconnected components.

Nodes represent entities and edges show relationships. GNNs excel at recognizing complex patterns and connections. By combining the pattern-recognition abilities of neural networks with the relationship-focused structure of graphs, GNNs offer powerful insights into systems where connections are crucial.

“GNN-based digital twins help scientists understand complex systems by looking at them as networks of connected parts, facilitating a comprehensive understanding of the system’s dynamic behavior,” said Hu.

By preserving the layout of the reactor systems and embedding fundamental laws of physics into the digital twin, he added, the approach ensures a robust and accurate replica of the real system.

The researchers used the Argonne Leadership Computing Facility (ALCF), a DOE Office of Science user facility, to train the GNN and for uncertainty quantification, which is the process of identifying, measuring and reducing uncertainty in models.

Quickly predicting reactor behavior

GNN-based digital twins are significantly faster than real-time or traditional system code simulations. They can rapidly predict how the reactor will behave during different scenarios, such as changes in power output or cooling system performance. They can do this by training on simulation data from the Argonne-developed System Analysis Module, a modern tool for analyzing advanced nuclear reactors.

The trained model is able to make accurate predictions based on limited real-time sensor data. This ability to deliver fast, authentic insights supports better planning for how reactors will respond to changes and better decision-making about their design and operation. It can help reduce maintenance and operating costs.

A digital twin could also be used to continuously monitor the reactor to detect any unusual behavior, called an anomaly. If something seems out of the ordinary, the system can suggest changes to keep the reactor safe or run smoothly.

A step forward for advanced reactors

Argonne’s new digital twin technology offers many benefits over traditional methods. By understanding how all parts of a reactor work together, the digital twin may provide more reliable predictions. It can be used to plan for emergencies, make informed decisions and even operate reactors autonomously in the future.

The research team’s innovation is a big step forward in the development and deployment of advanced nuclear reactors. By simulating diverse scenarios, digital twin technology helps ensure that reactors run safely, reliably and efficiently, reducing costs and extending the life of reactor components.