Multiobjective loading pattern optimization by simulated annealing employing discontinuous penalty function and screening technique

Abstract

The problem of multiobjective fuel loading pattern (LP) optimization employing high-fidelity three-dimensional (3-D) models is resolved by introducing the concepts of discontinuous penalty function, dominance, and two-dimensional (2-D)–based screening into the simulated annealing (SA) algorithm. Each constraint and objective imposed on a reload LP design is transformed into a discontinuous penalty function that involves a jump to a quadratic variation at the point of the limiting value of the corresponding core characteristics parameter. It is shown that with this discontinuous form the sensitivity of the penalty coefficients is quite weak compared to the stochastic effect of SA. The feasible LPs found during SA update the set of candidate LPs through a dominance check that is done by examining multiple objectives altogether. The 2-D–based screening technique uses a precalculated database of the 2-D solution errors and is shown to be very effective in saving the SA computation time by avoiding 3-D evaluations for the unfavorable LPs that are frequently encountered in SA. Realistic applications of the proposed method to a pressurized water reactor reload LP optimization with the dual objectives of maximizing the cycle length and minimizing the radial peaking factor demonstrate that the method works quite well in practice.