基于人工智能的核安全级DCS验证与确认新范式

An AI-Based New Paradigm for Verification and Validation of Nuclear-Safety-Class DCS

摘要:

数字化控制系统（DCS）是保障核电站安全稳定运行的“神经中枢”，直接关系到核安全，其可靠性和安全性要求极为严格。围绕AI技术在DCS验证与确认（V&V）中的创新应用，在三个关键领域取得重要突破。基于大语言模型的自然语言分析技术实现了需求文档的智能解析与比对，结合核电知识图谱提升了文档验证效率与准确性；在结构化语言分析方面，提出了基于IO自动化分配验证方法和CRNN驱动的仪控功能图智能比对算法，显著提高了IO分配验证和设计变更审查的效率；针对核工业遗留系统代码分析难题，基于大模型的代码语义理解与强大重构能力，实现了老旧代码的智能化分析。研究结果表明，AI技术的引入使核安全级DCS验证与确认过程的效率提升显著，同时降低了人工操作风险，为核领域V&V工作提供了智能化的全新解决方案。

Abstract:

The Digital Control System serves as the "nerve center" ensuring the safe and stable operation of nuclear power plants, directly impacting nuclear safety with extremely stringent reliability and security requirements. Significant breakthroughs have been achieved in three key areas regarding the innovative application of AI technologies in the Verification and Validation of nuclear-safety-class DCS. Large model-based natural language analysis technology enables intelligent parsing and comparison of requirement documents, significantly improving the efficiency and accuracy of document verification when combined with nuclear power knowledge graphs. In structured language analysis, an IO-automated allocation verification method and a CRNN-driven intelligent comparison algorithm for instrumentation and control function diagrams have been proposed, which significantly improved the efficiency of IO allocation verification and design change review. Additionally, to address the challenges of legacy code analysis in the nuclear industry, large model-based code semantic understanding and powerful refactoring capabilities have been employed to achieve intelligent code analysis of outdated code. The research findings demonstrate that the integration of AI technologies significantly improves the efficiency of nuclear-safety-class DCS V&V processes while reducing manual operation risks, providing an intelligent and novel solution for V&V in the nuclear industry.