Large language model agent for nuclear reactor operation assistance
This study proposes an artificial intelligence (AI) driven approach to nuclear reactor operation, focusing on a novel large language model (LLM) agent system designed to assist operators with various tasks within a nuclear reactor simulator. Previous studies have demonstrated the potential of deep l...
Saved in:
| Main Authors: | , , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
Elsevier
2025-12-01
|
| Series: | Nuclear Engineering and Technology |
| Subjects: | |
| Online Access: | http://www.sciencedirect.com/science/article/pii/S1738573325004103 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Summary: | This study proposes an artificial intelligence (AI) driven approach to nuclear reactor operation, focusing on a novel large language model (LLM) agent system designed to assist operators with various tasks within a nuclear reactor simulator. Previous studies have demonstrated the potential of deep learning for tasks such as anomaly detection and heat-up mode automation. Building on these efforts, studies have been conducted to employ LLMs for nuclear reactor diagnostics and the automation of system engineering tasks. The emergence of AI agents that utilize external tools, capable of natural language reasoning and retrieval-augmented generation, offers broader opportunities for decision-making and operation. We developed an AI agent architecture that integrated documentation, functions, and other modules. Two experiments were conducted to demonstrate the usefulness and potential of the proposed system. The first experiment evaluated whether the AI agent could address a boron concentration anomaly using a predictive module and relevant documentation. The second experiment evaluated whether the AI agent could generate a new procedure that parallels the provided manual without additional model training. The results indicate that the AI agent can execute commands and adapt the operational parameters, thereby demonstrating a comprehensive approach to automated reactor operations for safer, cost-effective performance. |
|---|---|
| ISSN: | 1738-5733 |