Development of a production MCNP CANDU 3D full-core model with practical remedies to the issues in deriving reliable tally results

This paper documents the development of a production MCNP model that is equivalent to the production WIMS-AECL/DRAGON/RFSP model for a CANDU full-core problem. The MCNP simulations confirm that the issues such as huge computing resources and large memory requirement, slow fission-source convergence, and fission-source correlation remain challenging in the Monte-Carlo calculation of the CANDU 3D full-core problem with higher dominance ratio. These issues are more challenging for the prediction of tally results (such as channel-power and bundle-power distributions) than the prediction of the k-effective values. In order to obtain a reliable MCNP full-core results, especially the power distributions, new approaches are proposed to remedy the above-mentioned issues in the MCNP modelling and simulation of the CANDU 3D full-core problem. Numerical results demonstrate that the new approaches proposed in this paper are practical in deriving reliable power distributions for the realistic CANDU 3D full-core problem from MCNP with limited computing resources. With the enhanced batch method proposed in this paper, CANDU full-core channel-power distributions are obtained from the MCNP simulations with the relative standard deviation less than 0.3% in the central high-power region and less than 1.2% in the peripheral low-power region.