The University of Michigan developed a physics-based control algorithm that lets nuclear microreactors automatically follow changing power demand while keeping core pressure and coolant temperature stable. Tested to ramp at 20% per minute, the controller tracked targets within 0.234%, demonstrating precise response. Because the approach is grounded in physics rather than opaque AI, researchers say it could be easier to certify and help deploy ~20 MW thermal microreactors in remote or strategic locations.
Physics-Based Algorithm Could Enable Safer, Autonomous Nuclear Microreactors
The University of Michigan developed a physics-based control algorithm that lets nuclear microreactors automatically follow changing power demand while keeping core pressure and coolant temperature stable. Tested to ramp at 20% per minute, the controller tracked targets within 0.234%, demonstrating precise response. Because the approach is grounded in physics rather than opaque AI, researchers say it could be easier to certify and help deploy ~20 MW thermal microreactors in remote or strategic locations.
14:23, 05.11.2025Technology
Physics-Based Algorithm Could Enable Safer, Autonomous Nuclear Microreactors
A University of Michigan research team has developed a physics-rooted control algorithm designed to let nuclear microreactors automatically adjust power output in response to changing demand — a capability known as load following. The work, published in Progress in Nuclear Energy, aims to make small-scale nuclear generation practical and economical in remote or strategic locations.
Unlike large commercial reactors, where operators commonly manage power adjustments, manual control can be prohibitively expensive for microreactors. The Michigan team modeled reactor behavior and built a predictive controller that optimizes the rotation of control drums around the reactor core. By following a physics-based model, the controller preserves core pressure and coolant temperature while tracking changing power commands over extended periods and under multiple operational constraints.
Crucially, the method is not based on machine learning or opaque artificial intelligence techniques; it relies on explainable physics and mathematical models. The researchers say that grounded, interpretable behavior could ease regulatory review and help vendors design autonomous control systems that are both safer and more secure.
"Many startup and legacy companies in the U.S. are pushing towards near-term and broad deployment of nuclear microreactors, and our work establishes a clear avenue to achieve that in an economically viable way," said study lead Brendan Kochunas.
In simulations, the algorithm responded to commands to change power at a rate of 20% per minute while maintaining output within 0.234% of the target, demonstrating tight tracking under aggressive ramping. The team considered microreactors producing up to 20 megawatts thermal, a scale suitable for applications such as military bases, disaster-response hubs, or remote communities.
Context: In 2023, nuclear power supplied 18.6% of U.S. utility-scale electricity, making it the country’s largest source of carbon-free generation. Nuclear produces no operational carbon dioxide and—according to the U.S. Energy Information Administration—delivers more electricity than wind and solar combined while requiring less land for equivalent output.
Despite those advantages, public concern persists over radioactive waste and the rare but serious risk of accidents, exemplified by historic events such as Chernobyl. The Michigan team frames their algorithm as one contribution among many global efforts to improve reactor safety and expand the role of small-scale, carbon-free power.
What this means: A physics-based, explainable controller could shorten regulatory pathways and help deploy compact reactors where traditional grid-scale options are impractical, while maintaining tight operational control and safety margins.