In some countries, as France, plutonium is used in industrial Pressurized Water Reactors (PWRs), using Mixed Oxides (MOX) fuels. To avoid a reactivity increase during a loss of coolant accident, the plutonium content in a PWR is limited at 12 wt%. This limit, that have been determined in the early 1990 s using a very conservative approach, is a strong constraint on the possibility to stabilize the plutonium inventory. Thus, further studies are now conducted to refine the plutonium limit content in the fuel. One interesting case is the local moderator voiding, which could for instance occurs at the beginning of a core depressurization accident. To study this phenomenon, a 2-steps neutronic deterministic scheme is proposed in this paper, using the APOLLO3® code. It is shown that the choice of the selfshielding method to use at the lattice step has a strong influence on the precision of the calculation. It is shown that, as expected, the subgroup formalism is the more precise to cope with both nominal and voided configurations. However, methods relying on this formalism are numerically costly. The use of a method relying on the finestructure formalism allow to conserve a satisfying estimation of the void effect on the reactivity or the spatial distribution of energy-integrated reaction rates. Unfortunately, we show that those good results are due to compensation of large energetic biases in the nominal case. A combination of Tone’s and finestructure selfshielding methods is presented. This method permit to strongly improve per-energetic-groups precision in the voided configuration compared to the method solely relying on the finestructure formalism, while drastically decreasing the numerical cost and duration compared to the subgroup method. The necessity to perform environed lattice calculation is demonstrated in the case of the voiding of a single assembly. A Local Flux Volume normalization of SPH coefficients is proposed to run equivalence calculation on this heterogeneous geometry. A method of characteristic solver is used for the flux calculation at the lattice step, while a S8 transport calculation is performed at the core level, using pin-by-pin discretization and a 26-groups energy mesh. Finally, the precision of the calculation scheme is assessed by comparison to Monte Carlo calculation ran with TRIPOLI–4® on a two-dimensional clusters of assemblies.
Read full abstract