Using Artificial Intelligence to Lower Nuclear Energy Costs

Nuclear power reactors generate around 10% of the world's electricity. Advanced reactor designs and recent progress in materials, computational methods, and diagnostics have significantly improved reactors' safety and reliability. However, to be competitive with modern energy systems, nuclear power plants must be economical and efficient.

 

Image used courtesy of Unsplash

 

Cost of Nuclear Power Plant Operation

A nuclear plant's costs consist of capital and operation costs. Capital cost includes the cost of engineering, commissioning, manufacturing, construction and financing. Operation costs include fuel costs, maintenance, decommissioning, and waste disposal.

New developments and research focus on reducing engineering, manufacturing, and construction costs. What could also help in lowering maintenance costs is automizing tedious tasks using artificial intelligence (AI). Skilled workers and special processes currently do these tasks.

Artificial Intelligence and Nuclear Energy

AI has the potential to make nuclear energy technology more reliable and flexible by predicting accidents, material interactions and leakages. It could also help cut operating costs associated with fuel, decommissioning and waste disposal. 

Researchers at Argonne National Laboratory (ANL) are developing an AI system that could streamline the operation and maintenance of nuclear reactors. They report that these power plants are expensive because they need constant monitoring and servicing to ensure safety. The researchers are exploring how AI-based algorithms could replace these activities.

"Operation and maintenance costs are quite relevant for nuclear units, which currently require large site crews and extensive upkeep," said Roberto Ponciroli, a principal nuclear engineer at Argonne. ​"We think that autonomous operation can help to improve their profitability and also benefit the deployment of advanced reactor concepts."

A nuclear power plant includes hundreds of sensors to monitor different parts of the plant to ensure they are working reliably. The reliability of data from these sensors is critical for decision-making. However, sensors can degrade. 

Currently, inspecting officers monitor each sensor and the performance of system components. ANL engineers are trying to make AI that can verify sensor data and learn how the normal operation of sensors differs from degraded ones.

AI could also evaluate the information from sensor data, make decisions, and take necessary actions. The engineers can achieve this by an AI method called reinforcement learning that replicates the human brain's logic by learning how to make decisions based on their possible outcomes. The system could alert the plant operators quickly if there are any problems and help optimize controls to prevent damage to expensive parts. 

As an autonomous nuclear plant requires diverse functionalities, the researchers plan to merge multiple algorithms such as System Analysis Module (SAM), an analysis tool for advanced reactors.

System Analysis Module for Advanced Nuclear Reactors

SAM is a system analysis tool developed by ANL, in collaboration with Kairos Power, for advanced safety analysis. It uses an object-oriented application framework called MOOSE and its meshing and finite element library to support advanced software environments and numerical methods.

SAM is a computationally efficient, high-fidelity modeling and simulation tool. It can model and simulate heat transfer and fluid dynamic responses in reactors. ANL engineers demonstrated the simulation capabilities of typical reactor accidents of various advanced reactor types with SAM. It also has advanced and efficient thermal mixing and stratification modeling capacity to improve safety analysis and reduce uncertainties.

 

Temperature distributions of an advanced reactor type with SAM. Image used courtesy of Argonne National Laboratory

 

SAM also embeds flexible core modeling capabilities. A core in the nuclear power plant is the critical part containing nuclear fuel. The incapacity of the power plant to perform the desired operation on its core can result in high operating costs and reduced system security and reliability. ANL engineers developed a multi-channel rod bundle to account for the temperature difference between the center region and the edges of the coolant channel in the fuel assembly.

The tool can integrate with other advanced multi-scale and multi-physics modeling simulation tools.

Future Direction of AI and Nuclear

The ANL engineers have developed a simulated reactor to test their AI system. The team is currently validating their algorithms on the simulated reactor, and so far, they have achieved systems to control and diagnose its virtual part. The next step will be to focus on the decision-making ability of the system.

 

Feature image used courtesy of Unsplash