PWR core pattern optimization using grey wolf algorithm based on artificial neural network

Abstract

This paper provides a novel way to dissolve the problem of finding the best configuration for fuel assemblies in a PWR core. For this goal, the Grey Wolf Optimization (GWO) algorithm relying on the demeanor of grey wolves for hunting is introduced and an artificial neural network (ANN) is applied to estimate the fitness function value of GWO. Besides the GWO, the Genetic Algorithm (GA) and Gravitational Search Algorithm (GSA) have been applied and the performances of these algorithms in challenging test functions (Holder table and Levy) and loading pattern optimization (LPO) problem are compared. A neutronic fitness is defined for increasing multiplication factor (keff) and for flattening of power peaking factors (PPFs). To calculate the required neutronic parameters of the core, a nuclear computational code, PARCS, is employed. This code has been coupled with the GWO, GA, and GSA algorithms in MATLAB by proper procedures. By creating an artificial neural network with 3500 different loading patterns coupled with GWO, the speed of the optimization has been greatly improved. The results show the usefulness of the GWO and confirm that the GWO-ANN has appropriate speed and adaptability for loading pattern optimization.