Sensor Placement Optimization and Online Power Reconstruction in Nuclear Reactors via Hyper-Reduction

This paper presents a comprehensive study on sensor placement optimization and online power reconstruction in nuclear reactors utilizing hyper-reduction methods. An effective reduced-order model (ROM) for reactor neutronics is constructed using proper orthogonal decomposition (POD), incorporating typical reactor operational parameters such as burnup, fuel/coolant temperature, and control rod insertion. The Q-DEIM and Gappy+E algorithms are employed to optimize sensor placement, enabling the efficient integration of measurements into the ROM. Furthermore, an indirect reconstruction method is introduced for POD-based ROMs, allowing the reconstruction of unmeasurable high-energy neutron flux fields and directly contributing to accurate power reconstruction. Numerical results demonstrate that the optimized sensor placements exhibit quasi-optimal reconstruction performance. The hyper-reduction method achieves accurate fission power reconstruction, with errors falling within a 5% error range under sensor noise levels of 0.01. This study contributes to improving the traditional design process of neutron flux monitoring systems (NFMS) and has the potential to drive technological innovations in NFMS design, thereby promoting safer and more efficient reactor operations.