Recent advances in accurate neutronic calculation of GEN-3 LWR reactors

Abstract New French GEN-III reactors are characterized by an improved safety (core catcher for corium, stainless-steel neutron reflector to minimize vessel embrittlement, permanent in-core instrumentation thanks to SPND, …). Moreover, the assembly central guide-tube can be replaced by a fuel pin and the First core is controlled through numerous 8% Gd burnable poison fuel pins. These new features harden the neutron spectrum and increase the neutronic calculation challenge. Therefore, a new calculation package APOLLO2.8 was developed, based on an efficient Method of Characteristics (MOC) and the new SHEM 281-group energy mesh that enables accurate resonant absorption prediction. The JEFF3.1.1 recent nuclear data library, which involves the feedback from Critical Experiments and LWR Post Irradiation Experiments, was used. The experimental validation of the APOLLO2.8/JEFF3.1.1 package allowed the Uncertainty Quantification associated with PWR design parameters. In order to meet target-accuracy required in specific GEN-III parameters, targeted integral experiments where achieved. To validate heavy neutron reflector the mock-up experiment PERLE was carried out in EOLE critical facility. PERLE allowed the calculation check of SS reflector-saving δ refl and the improvement of Fe cross-sections. Thanks to the large French experimental database, we determined reliable covariance matrices associated to JEFF3.1.1 evaluations, in order to perform the propagation of nuclear data uncertainties. Sensitivity/uncertainty analysis has shown that the predominant component on radial power map comes from the 238 U inelastic scattering cross-section. To reduce this 3% uncertainty component, we performed a 238 U( n,n ′) re-estimation, using relevant selected experiment benchmarks. This theoretical and experimental extensive work performed since 2003 allows a reliable prediction of the local flux and K eff (±200 pcm in 1 σ ) in large GEN-III reactors. The challenging pin-by-pin power prediction is today overcome, and 3D-stochastic TRIPOLI4 or 2D-deterministic APOLLO2.8 calculation uncertainty is decreased from prior ±8% to current ±2.6% value.