PG&E to Use AI at Diablo Canyon Nuclear Plant
Pacific Gas and Electric (PG&E) announced that it will implement Atomic Canyon’s Neutron Enterprise solution at its Diablo Canyon nuclear facility. This marks the first use of generative AI on-site at a U.S. nuclear power plant.
Utilizing advanced NVIDIA AI technology, the Neutron Enterprise system aims to enhance the way the plant manages its extensive document library, thus significantly improving operational efficiency and reducing costs. This new tool is expected to streamline how plant employees search through Diablo Canyon’s vast records, ultimately aiding in construction and operations at the facility.
The initial rollout of this AI solution is planned for early 2025, but its potential applications extend beyond just document retrieval. Diablo Canyon’s Vice President of Business and Technical Services, Maureen Zawalick, mentioned that generative AI could also be employed for data analysis, maintenance, and engineering tasks.
Complexity of Record Retrieval
Diablo Canyon has accumulated millions of pages of documentation over its 40 years of operation, including many records that have been converted from older formats such as microfiche. This complexity requires significant time and effort to navigate, often taking hours or even days for employees to find pertinent information during operational challenges. A well-designed AI could drastically cut down this time to mere seconds.
“Maximizing human capability for decision-making and data analysis is essential,” Zawalick stated. Although PG&E will need time to gather precise data on efficiency gains, the AI tool is expected to greatly enhance the plant’s maintenance strategies and operational performance.
Tailored Solutions for the Nuclear Sector
Neutron Enterprise is based on Atomic Canyon’s unique FERMI AI models, created specifically for the nuclear industry in collaboration with the U.S. Department of Energy’s Oak Ridge National Laboratory. Previous attempts to apply more general AI models to the extensive records governed by the U.S. Nuclear Regulatory Commission faced difficulties due to the specialized language used in nuclear documentation. However, the FERMI models now achieve results that are significantly more accurate and efficient than previous solutions.
Atomic Canyon has invested nearly a year into understanding Diablo Canyon’s specific operational requirements and training the AI models. They are currently in the process of installing the necessary NVIDIA hardware at the plant to facilitate local operation of Neutron Enterprise.
Looking Ahead
This initiative is reminiscent of previous ventures by Lauderdale in regulated industries, highlighting the initial resistance that often accompanies the introduction of AI technologies. However, he believes that record retrieval poses low risks and can pave the way for further integration of AI in maintenance and operations.
The adaptability of Neutron Enterprise may eventually lead to its expansion into design and engineering support, acting as an essential “copilot” for day-to-day tasks at Diablo Canyon. Generative AI could also simplify tedious regulatory processes associated with extending the plant’s operational timeline.
Atomic Canyon’s vision goes beyond current operations; they aim to leverage advancements in AI for future nuclear power plant constructions as well. This marks the beginning of their journey to lead the way in safely integrating AI within the nuclear sector.