Nuclear power plants can provide large amounts of low-carbon energy. However, their construction and fuel do generate a lot of carbon dioxide. And the disposal of spent fuel and decommissioning of nuclear power plants also produce carbon dioxide. The levelized cost of nuclear power is greater than hydro, wind or solar power. Analysts report that choosing to support massive expansion of nuclear power would preclude similar expansion of renewable power sources. In addition, nuclear power plants are becoming more expensive to construct, operate and maintain while the cost of renewable energy sources keeps dropping. A great deal of effort is being put into making nuclear power plants cheaper to construct and operate.
Researchers at the U.S. Department of Energy (DoE) Argonne National Laboratory are developing systems that could make nuclear energy more competitive by employing artificial intelligence (AI). Argonne is midway through a one million, three-year project to explore how smart, computerized systems could change the economics of nuclear power.
The Argonne AI project is funded by the DoE office of Nuclear Energy’s Nuclear Energy Enabling Technologies program. The project’s goal is to create a computer architecture that could detect problems early and recommend appropriate actions to human operators at a nuclear power plant. Roberto Ponciroli is a principal nuclear engineer at Argonne. He and his colleagues estimate that the Argonne project could save the nuclear industry more than five hundred million dollars per year.
A typical nuclear power plant can contain hundreds of sensors which monitor different parts to ensure that they are working properly. Currently, the job of inspecting each sensor and checking the performance of system components such as valves, pumps and heat exchanges is performed by the plant’s staff. They have to walk around the plant to check each component. With the new Argonne AI system, algorithms could verify data by learning how normal sensors function and looking for anomalies.
After validating the operation of the power plant’s sensors, the AI system will then interpret the signals from them and suggest specific actions where necessary. At a nuclear power plant, computers could detect problems and flag them to plant operators as early as possible. This would help optimize control and avert more expensive repairs in the future. In addition, computers could prevent unnecessary maintenance on equipment that does not need it.
Richard Vilim is an Argonne senior nuclear engineer. He said, “The lower-level tasks that people do now can be handed off to algorithms. We’re trying to elevate humans to a higher degree of situational awareness so that they are observers making decisions.”
Collaborating with the nuclear industry to develop testing scenarios, Argonne engineers have built a computer simulation of an advance nuclear reactor. While the system is designed to serve new reactor technologies, it is also flexible enough to be applied at existing nuclear power plants.
Currently, researchers are validating their AI concept on the simulated reactor. They have completed systems to control and diagnose the vital parts of the reactor. The remainder of the Argonne project will focus on the system’s decision-making ability and what it does with the diagnostic data. An autonomous nuclear power plant requires varied functions. The endproduct of the Argonne project is a system architecture that combines multiple algorithms.