Model predictive control with state estimation and reduced-order physics for nuclear reactor operation during unanticipated communication transients

Advanced nuclear reactors are increasingly being designed for remote operation and higher levels of autonomy, particularly for small and microreactor applications where on-site staffing and frequent fuel handling may be impractical. These new operational paradigms would require networked digital instrumentation and control systems, which could introduce communication uncertainties such as noise, latency, packet loss, and potential denial-of-service (DoS) events. Ensuring stable and efficient reactor operation under such conditions is therefore a critical challenge. Here, we present a quantitative evaluation of a proposed model predictive control integrated with a moving horizon state estimator (MPC-MHE) and reduced-order physics-based models for reactor power regulation under degraded communication conditions. Controller performance is benchmarked against a tuned PID controller under nominal conditions, high-intensity stochastic packet loss (95% probability of data loss), and extended sensor blackouts representative of DoS events. Across all scenarios, the MPC-MHE with reduced-order physics-based models consistently maintains tighter power regulation with substantially reduced control rod motion with the most significant improvements are observed during communication blackouts, where MPC-MHE reduces tracking error by 49–59% and control effort by 54–62% relative to PID control during both steady-state and transient power operations. These results demonstrate that model predictive control with state estimation could provide significant robustness and efficiency advantages over conventional PID control in network-degraded environments.