Application de la méthode de modélisation de réflecteurs Baff-Refl dans Cocagne

In neutronic simulation, accurately modeling the reflector, surrounding the reactor core, is a critical challenge for ensuring high-quality results. Industrial calculation schemes, which are designed to be faster than reference models, often rely on the infinite medium hypothesis. While this assumption is valid for the center of the core, it becomes a major source of error for fuel assemblies located at the periphery. The resulting flux distortions at the core-reflector interface lead to significant power calculation inaccuracies that must be addressed. To mitigate these effects, EDF currently utilizes the 1D Lefèbvre-Lebigot method within its proprietary CARABAS and COCAGNE codes. However, alternative approaches exist, such as the Baff-Refl method, which is implemented in the Framatome’s platform SCIENCE and which is available in the DONJON neutronic environment developed at Polytechnique Montréal. This study aims to benchmark these two methodologies by comparing power distribution results against a CARABAS reference. The evaluation was conducted across several configurations, including homogeneous cores and nominal parameters, with a specific focus on hot standby conditions. The research also investigated the influence of the D1 diffusion coefficient and compared the specific Golfier-Vergain and Todorova models within the Baff-Refl framework. The results indicate that both Lefèbvre-Lebigot and Baff-Refl deliver comparable performance levels, although the former remains slightly more accurate in hot standby scenarios. In both cases, radial power shifts were observed, with overestimations and underestimations reaching their peak at the core-reflector boundary—particularly in the corners. These localized errors are primarily attributed to the inherent limitations of 1D modeling and the persistent bias of the infinite medium hypothesis. Despite these similarities, Baff-Refl has the advantage of being adaptable to 2D geometries and allows for the direct calculation of the diffusion coefficients in the reflector, inlike Lefèbvre-Lebigot which miposes these values for each type of reactor. It can also be based on different angular (SPN beyond order 1) or spatial discretization formalisms (nodal expansion method, nodal analytical method, Raviart-Thomas finite element). These advantages could lead to further research about Baff-Refl at EDF.