VARIANCE OVERESTIMATION IN MONTE CARLO MULTI-PHYSICS SIMULATION

Abstract

The variance overestimation in Monte Carlo (MC) multi-physics simulation is studied on BEAVRS benchmark by using MCS code developed at Ulsan National Institute of Science and Technology (UNIST). The variance overestimation is caused by the negative inter-cycle correlation induced by the multi-physics feedbacks including equilibrium xenon and thermalhydraulics. The overestimation phenomena are studied for radially integrated axial flux distribution. The real standard deviation, apparent standard deviation, and Auto-Correlation Coefficients (ACCs) are calculated. It is shown that the apparent standard deviation is about 2 times larger than the real standard deviation, and overestimation ratio can be estimated accurately using ACCs.