Installation and evaluation of a nuclear power plant Operator Advisor based on artificial intelligence technology. Final report

摘要

The Artificial Intelligence Group in the Nuclear Engineering Program has designed and built an Operator Advisor (OA), an AI system to monitor nuclear power plant parameters, detect component and system malfunctions, dispose their causes, and provide the plant operators with the correct procedures for mitigating the consequences of the malfunctions. It then monitors performance of the procedures, and provides backup steps when specific operator actions fail. The OA has been implemented on Sun 4 workstations in Common Lisp, and has been interfaced to run in real time on the Perry Nuclear Power Plant full-function simulator in the plant training department. The eventual goal for a fully functioning Operator Advisor would be to have reactor operators receive direction for all plant operations. Such a goal requires considerable testing of the system within limited malfunction boundaries, an extensive Verification & Validation (V&V) effort, a large knowledge base development effort, and development of tools as part of the system to automate its maintenance. Clearly, these efforts are beyond the scope of the feasibility effort expended during this project period. However, as a result of this project, we have an AI based platform upon which a complete system can be built.