Estudo do algorítimo evolução diferencial aplicado a otimização global

Abstract

The differential evolution is a population based stochastic search algorithm applying biological operators. It is an algorithm for solving global optimization, and can handle nondifferentiable, non-linear and multimodal. It’s performance is highly correlated with the configuration of it’s control parameters, the choice selection of the set fundamental for the success of the algorithm. In this work, a study was carried out on the differential evolution algorithm, as well as a study on it’s control parameters and it’s strategies and a focus on an analysis of the behavior of the algorithms. For validation, were the ability to test the parameters implemented, there was no problem of global optimization. For this, the five functions tested were chosen, being two functions unimodal and three multimodal. In order to analyze the data of the underlying procedure of the algorithm, a graphical interface was developed to aid the understanding of the differential evolution algorithm process. Finally, the results that the algorithm presents a very efficient performance for global optimization solution and it is quite sensible in some of the other control parameters values, on the other hand, the inexpressive influence of the strategies on the performance of the differential evolution algorithm.