Non-intrusive simultaneous calculation of nuclear data sensitivity and uncertainty of core simulator parameters for the full-scale Krško NPP core model

We present a framework for propagating uncertainties from nuclear data to integral parameters of the Krško PWR core design calculations and for calculating their sensitivity profiles to nuclear data. This non-intrusive method is intended for multi-step calculations using codes where nuclear data sensitivity capabilities based on adjoint calculation are not available. In our approach, incident neutron reaction cross-sections and ν̄ are randomly sampled based on covariance information in the source evaluated nuclear data files and perturbed core design calculations are run. The uncertainties of the critical boron concentration, of the axial difference in relative power and of their change in the reactor cycle are calculated as the standard deviation of the core design output runs. A linear regression framework is designed to compute sensitivity profiles from the nuclear data perturbations used for uncertainty propagation. After validation against the results of the Serpent Monte Carlo code for the TMI pincell criticality benchmark, we calculated the propagated uncertainty of ²³⁵U, ²³⁸U, ¹⁶O, ⁵⁶Fe and ²³⁹Pu to the selected reactor core parameters and the sensitivity of the Krško critical boron concentration at hot zero power to the fission cross section of ²³⁵U. The contribution of individual nuclides to the overall uncertainties was estimated with first-order Sobol indices.