Integrated Real-Coded Genetic Algorithm and Particle Swarm Optimization for Solving Constrained Global Optimization Problems

Abstract

AbstractThis work presents an integrated real-coded genetic algorithm and particle swarm optimization with a penalty function method (RGA-PSO) to solve constrained global optimization (CGO) problems. An outer RGA is used to optimize the parameter settings of an inner PSO algorithm, which includes the constriction coefficient, cognitive parameter, social parameter, penalty parameter and mutation probability. To increase the diversity of particles, the inner PSO method uses a multi-non-uniform operator. Performance of the proposed RGA-PSO algorithm is evaluated using a set of benchmark CGO problems. Additionally, this work compares the numerical results obtained using the proposed RGA-PSO algorithm with those of GAs and artificial immune algorithms (AIAs). Experimental results indicate that the proposed RGA-PSO algorithm converges to a global optimum solution to a CGO problem, and that the performance of the proposed RGA-PSO algorithm is competitive with those of GAs and AIAs. Although a hybrid method increases the complexity of an algorithm, the proposed RGA-PSO algorithm can obtain the optimum parameter settings for the inner PSO algorithm by using the outer RGA and can find a global optimum solution to a CGO problem by using the inner PSO algorithm, simultaneously. Hence, the proposed RGA-PSO algorithm can be considered as an alternative stochastic global optimization method for solving CGO problems.Keywordsreal-coded genetic algorithmparticle swarm optimizationnonlinear programmingconstrained global optimization