A team of experts from the US Department of Energy’s Argonne National Laboratory has demonstrated the transformative potential of machine learning (ML) in the operations of sodium-cooled fast reactors (SFR), a cutting-edge nuclear reactor. The application of ML in this specialized reactor aims to improve security and efficiency, showcasing advancements that could revolutionize power generation and contribute to nuclear waste reduction.
SFRs utilize liquid sodium as a core coolant, allowing for efficient electricity generation without carbon emissions from heavy atom splitting. While not currently employed for commercial purposes in the US, these reactors are seen as a promising avenue for cleaner and more sustainable energy in the future.
Overcoming Operational Challenges
The technology faces a significant challenge related to maintaining the purity of high-temperature liquid sodium coolant, crucial for preventing corrosion and system clogs. In response, Argonne scientists developed a groundbreaking ML system to continuously monitor and detect anomalies, advancing the state of instrumentation control in nuclear energy systems.
Alexander Heifetz, principal nuclear engineer at Argonne, emphasized, “By harnessing the power of machine learning to continuously monitor and detect anomalies advances the state of the art in instrumentation control. This will create a breakthrough in the efficiency and cost-effectiveness of nuclear energy systems.”
Key Capabilities of the ML Model
The ML model, created by the team, is designed to continually monitor the cooling system, analyzing data from 31 sensors at Argonne’s Mechanisms Engineering Test Loop (METL) facility. METL serves as an experimental setup to assess materials and components for these reactors safely and precisely, also acting as a training ground for engineers, technicians, and machine learning models.
The ML model demonstrated its ability to quickly and accurately detect operational irregularities. In a simulated loss-of-coolant anomaly, marked by a sudden spike in temperature and flow rate, the model detected the anomaly within approximately three minutes of initiation.
Challenges and Future Improvements
While the research offers significant improvements for future models, limitations exist, such as the potential for false alarms due to random spikes or sensor inadequacies. The current model issues an alert when a spike surpasses a predetermined threshold, prompting the team to refine the model to distinguish between genuine process anomalies and random measurement noise.
Heifetz concluded, “Although we’re using the unique capabilities of METL to develop and test our algorithms in a liquid metal experimental research facility, there is potential to see applications in advanced reactors that can provide more carbon-free energy in the future.”
By Impact Lab
