Nuclide number density prediction in the lattice physics calculation based on Dynamic mode decomposition

Abstract

Burnup analysis in nuclear reactors requires iterative computation of neutron transport and fuel depletion, which is computationally intensive, particularly for large-scale scenarios. This study introduces an innovative approach leveraging the Dynamic Mode Decomposition (DMD) algorithm to predict the temporal evolution of nuclide densities. By identifying and utilizing the DMD modes and eigenvalues from snapshots of nuclide density, this method aims to alleviate the computational demands of the coupled transport and burnup calculations. Firstly, the methodology selects the key reactivity-contributing nuclides to evaluate the correlation between the complexity of the reduced-order model and the precision of predictions. Subsequently, an optimized reduced-order model is employed for forecasting nuclide densities in a pin-cell. In most cases, DMD predicts more accurately than traditional quadratic extrapolation methods. Moreover, the DMD algorithm demonstrates commendable accuracy in predicting the nuclide density distribution within a PWR fuel assembly, suggesting its promising potential for reactor burnup analysis applications.