Predictive-Protective Control of Nuclear Reactor Excursions: A Point Kinetics Approach using the AMDCF Kernel

The Chernobyl disaster (1986) remains a stark example of the catastrophic consequences of uncontrolled positive reactivity feedback and human error in nuclear systems. In this study, we introduce a mathematically rigorous successor to our previous Agent-Based Model (ABM) by applying the Adaptive Memory-Driven Coupling Framework (AMDCF) to a Reduced-Order Point Kinetics model of the RBMK-1000 reactor. Unlike phenomenological approaches, this simulation incorporates one-group delayed neutrons, thermal-hydraulic feedback (positive void coefficient and negative Doppler effect), Xenon-135 poisoning dynamics, and the infamous positive scram effect (graphite-tipped control rods). The results demonstrate that while an unprotected reactor undergoes a catastrophic prompt criticality excursion (numerical divergence), the AMDCF kernel—acting as an active safety controller—successfully detects the power surge. By utilizing a dynamic sigmoid safety valve, the framework instantaneously overrides the flawed human command and inserts neutron-absorbing control rods (Boron), stabilizing the reactor power and temperature. This work establishes the AMDCF kernel as a robust, mathematically sound, predictive-protective algorithm for critical infrastructure, offering a new paradigm for nuclear control and safety systems. This work serves as a mathematically rigorous successor to the Agent-Based Model (ABM) presented in our previous preprint: 'Universal Safety Kernel v4' (SAiM Mohammed Tahar, 2026).