Advanced reactor models for plutonium multirecycling scenario studies using artificial neural networks

This paper presents an innovative way to identify the plutonium content in MOX fresh fuel according to its isotopic composition in fuel cycle dynamic simulations. In those calculations, the fuel fabrication is a challenge to model as the plutonium stock status (quantity and quality) is unknown and the fuels that are loaded in reactors should ensure criticality throughout the wanted cycle length. This model is based on metamodeling full core depletion simulation of PWRs with artificial neural networks to predict efficiently the cycle length and power factors as a function of the MOX fresh fuel composition. An exploration of the plutonium phase space has shown the impact of plutonium isotopic composition on power factors and fuel cycle length variability. Moreover, the sensitivity of the cycle length to the plutonium content have led us to propose a model that links the plutonium content to its isotopic composition by a constant step function, while accepting an uncertainty of the order of 1 GWd/t on the cycle length. The impact of this new model on PWR loaded with plutonium recovered by UOX and MOX spent fuel reprocessing is assessed in this paper. Pu content estimator based on this new approach leads to standard deviation of the cycle length discrepancies distribution lower than 150 MWd/t for a Pu fissile quality range size of 4%.