Development of a Supervisory Control System as a Transition Technology Towards Autonomous Reactor Plant Operations

Abstract

The economic viability of small and microreactors depends on reducing energy generation costs. The implementation of autonomous reactor control systems provides an avenue for reducing operations and maintenance expenses. Advanced reactor designs with enhanced passive safety features, reduced source terms, and digital instrumentation and control systems, directly support autonomous controllers. In these plants, where the need for human operators is already reduced, the introduction of supervisory control systems (SCS) for dynamic operations further lessens operator dependence while building trust in these systems, laying a solid foundation for the transition to fully autonomous reactor control.

Finite state automata (FSA) provide a framework for engineering fully verifiable and validatable supervisory controllers, and thereby facilitate the transformation to autonomous operations in nuclear power plant operations. FSA serve as a foundational mathematical tool for modeling discrete event systems (DES). Properties such as nonblocking and controllability can be formally demonstrated and verified by leveraging the extensive set of mathematical proofs within the scope of regular languages. Furthermore, a DES can be directly linked to reactor plant systems and operational procedures within a hierarchical architecture by using a graded functionalization approach analogous to that of complex dynamic systems, such as self-driving vehicles. In this scheme, feedback controllers can regulate low-level actuation functions while a supervisory controller can govern high-level plant state transitions.

A generic supervisory controller was developed as a transition technology toward autonomous reactor operations. This controller was then tailored for application on a limited feedback model, for initial proof-of-concept testing, and then was scaled for use on light water reactor (LWR) simulators. In the absence of advanced reactor simulators for operational testing, LWR simulators were used because they provide realistic feedback and controls within a more conservative operating margin than advanced reactors. These supervisory controllers successfully executed operational procedures within a fully verifiable framework, establishing the foundation of this modeling approach and laying the groundwork for its implementation in advanced reactor designs. This scalable model thus facilitates a smooth transition from functioning as an operator aid to fully autonomous operation as a comprehensive plant controller, increasing the economic viability of nuclear power.