Multi-group SP₃ approximation for simulation of a three-dimensional PWR rod ejection accident

Highlights: • The multi-group SP₃ method developed and implemented in PARCS for the MOX analysis. • The verifications were performed in 2D and 3D, 2G and MG, diffusion and transport, with and without feedback. • All results show consistency with the reference results obtained from the ANL P_N transport code VARIANT for steady-state and transport calculations. • It was found that the SP₃ angular approximation captures sufficient transport effects for both steady-state and transient, and provides essentially the same results as the VARIANT P₅ method. • From the transient results of the full-core problem, it was noted that MG is more conservative than 2G, and P₁ is more conservative than SP₃. - Abstract: Previous researchers have shown that the simplified P₃ (SP₃) approximation is capable of providing sufficiently high accuracy for both static and transient simulations for reactor core analysis with considerably less computational expense than higher order transport methods such as the discrete ordinate or the full spherical harmonics methods. The objective of this paper is to provide a consistent comparison of two-group (2G) and multi-group (MG) diffusion and SP₃ transport for rod ejection accident (REA) in a practical light water reactor (LWR) problem. The analysis is performed on two numerical benchmarks, a 3 × 3 assembly mini-core and a full pressurized water reactor (PWR) core. The calculations were performed using pin homogenized and assembly homogenized cross sections for a series of benchmarks of increasing difficulty, in two-dimensional (2D) and three-dimensional (3D), 2G and MG, diffusion and transport, as well as with and without feedback. All results show consistency with the reference results obtained from higher-order methods. It is demonstrated that the analyzed problems show small group-homogenization effects, but relatively significant transport effects which are satisfactorily addressed by the SP₃ transport method. The sensitivity tests also show that, for the REA simulation, the MG is more conservative than 2G, P₁ is more conservative than SP₃ for a 1/3 MOX loaded full-core problem