A cross-entropy method applied to the In-core fuel management optimization of a Pressurized Water Reactor

The In-Core Fuel Management Optimization (ICFMO) is a prominent problem in Nuclear Engineering. In the present paper, the application of the Cross Entropy (CE) method to the ICFMO is described. The CE method was initially developed for rare event simulation, and adapted to combinatorial and continuous optimization. Unlike the several metaheuristics biologically or physically inspired that have been applied to the ICFMO, the CE is an algorithm based on statistics and information theory, specifically on the minimization of the Kullback-Leibler divergence (or cross entropy). Notwithstanding the high level theory underlying the algorithm, its implementation mainly comprises the sampling of candidate solutions at each iteration with an update of parameters for generating the random sample in the next iteration given by the analysis of an elite group within the sample. Thus, it is possible to drive the algorithm to optimal or near optimal solutions iteratively. The CE algorithm was applied to the optimization of the 7th cycle of Angra 1 Nuclear Power Plant, in Brazil. The results compare favorably to other algorithms such as Genetic Algorithm, Population Based Incremental Learning, as well as Particle Swarm Optimization under the same conditions of simulation.