Scale covariance libraries for sodium-cooled fast reactor systems

As basis for systematic uncertainty and sensitivity analyses of fast spectrum systems with the GRS code XSUSA, new multi-group covariance libraries are generated considering a typical neutron flux spectrum of a sodium-cooled fast system as weighting function and a sufficient number of energy groups in the fast energy region. A covariance library with only a few energy groups is desired to allow informative sensitivity analyses when performing uncertainty analyses based on the random-sampling approach. Several libraries with a reduced number of energy groups are therefore generated. The performance of these few-group covariance libraries in comparison with a very fine 2082-group library is investigated in uncertainty analyses using the SCALE code TSUNAMI with a metal and an oxide fuel assembly model, respectively, in terms of the eigenvalue, the Doppler reactivity and the sodium void reactivity. The total output uncertainties obtained with the few-group covariance libraries show relative differences to reference calculations of less than 5% for all investigated cases. The comparison of the individual contributions of the most important covariance matrices also leads to satisfying results. A comparably coarse energy group structure of the covariance data with only 8 energy groups is found to be adequate for a reasonable assessment of the uncertainties of important quantities in the analysis of the fast systems under consideration.