Grigoriadis
Advanced reactors, including microreactors and other modular designs, are envisioned to operate with minimal staffing and, in many deployment scenarios, without continuous onsite operators. Realizing this self-regulating capability requires control strategies that go beyond conventional feedback control, anticipating disturbances and setpoint changes before they propagate into safety-relevant deviations. The research presentation is centered on anticipatory control, implemented through data-driven and adaptive model predictive control, which uses reduced-order or machine-learned models of reactor thermal-hydraulic behavior to predict future system states and compute corrective actions faster than real time. This approach has been demonstrated on heat-pipe microreactor testbeds, showing the ability to maintain temperature and power setpoints under normal operation as well as degraded conditions such as heat pipe failures. These control strategies are paired with digital twins that integrate physics-based and data-driven models, enabling diagnosis, prognosis, and discrepancy checking within a broader nearly autonomous management and control framework. Complementary work has proposed structured automation levels to guide the transition from operator-driven to increasingly autonomous reactor operation. Together, this body of research establishes a technical pathway toward resilient, self-regulating advanced reactor control that can reduce operating costs under both nominal and off-nominal conditions.