Intelligent optimization of power distribution for fast reactor NCLFR-Oil based on SPN method

Abstract

A custom-developed neutron transport module based on the COMSOL finite element solver was created to enable efficient optimization and parameter evaluation in core design, and it can be integrated with other built-in modules for enhanced capabilities. This work began by establishing a practical foundation for a multi-dimensional SP_N method using the PDE solver, capable of simulating both steady-state (k-eigenvalue) and time-dependent transport problems. The steady-state solver showed good agreement with 3D TAKEDA and 2D C5G7 benchmarks, while the transient solver was well-validated with TWIGL and LMW benchmarks. For modeling the self-designed fast reactor NCLFR-Oil, OpenMC was used to generate few-group constants, which were then imported into COMSOL’s SP₃ neutron transport module as equation coefficients. The SP₃ model’s capability to simulate the core’s physical field was validated by testing eigenvalues, control rod worth, the power and neutron flux distribution. Sensitivity analysis was performed using COMSOL’s uncertainty quantification module to assess the impact of control rod positions on core eigenvalues and power distribution, refining the parameter space for optimization and enhancing efficiency. To further improve optimization efficiency, a surrogate model based on “Polynomial Chaos Expansion” was employed to approximate the core’s physical model, predicting relationships between input parameters and optimization objectives. This model proved more efficient than the gradient-free “Coordinate Search” method, reducing computational resource consumption. The optimization results showed a significant reduction in the custom power flattening factor, bringing more power factors closer to the target value of 1.