Parameterized cross-section library generation with the SERPENT 2 code using novel parameterization methods and testing in VVER-1200 core geometry with the KIKO3DMG nodal solver

In this paper, a two-step neutronics calculation methodology using the Serpent 2 Monte Carlo code as the first step and the diffusion solver of the KIKO3DMG nodal core simulator as the second step is presented and tested for a core model with VVER-1200-like geometry. For the group constant parameterization, a novel parameterization package was used with hyperparameter optimization and automated model selection. In order to verify this pipeline, a single-state test calculation was used to assess the effects of each methodological step on the effective multiplication factor and the assembly-wise normalized power distribution. It was found that the multiplication factors are in good agreement, and the difference in power distribution primarily comes from the difference between nodal and Monte Carlo codes. The handling of the group constants was improved with two steps, which stabilized the nodal output values. To verify the method across the state space, the coupled KIKO3DMG results were compared to reference Serpent 2 calculations in 72 different states, with effective multiplication factors and reactivity coefficients in good agreement. Moreover, the robustness of the methodology with respect to Monte Carlo uncertainty and parameterization bias was demonstrated by repeating all calculations 100 times and quantifying uncertainties in the final deterministic results. The calculations were also performed with linear interpolation as a reference fitting method, to which our results were compared.