Identificação paramétrica de sistemas a partir da resposta em frequência empregando evolução diferencial

Abstract

The Differential Evolution (DE) algorithm can be used to solve mono and multiobjective optimization problems. However, stagnation and premature convergence to a local optimum can occur. To avoid this, the introduction of two new operators: Diversity control and chaotic search to original algorithm are proposed. The first operator aims to maintain diversity in the population. The second aims to expand the search space. The DE algorithm with these new operators was called Differential Evolution associated with Chaotic Search Models and Diversity (DE-CSMD). Both algorithms (DE and DE-CSMD) were used to identify transfer functions with uncertain delay from frequency response data. It should be noted that the presence of an uncertain delay makes the optimization problem involved in the identification non-convex, which justifies the use of DE. The results showed that, in systems with well-defined global minimum, the DE-CSMD algorithm was able to identify transfer functions with greater representation. However, new operators can hinder the convergence of the algorithm in systems with multiple local minima. This occurs due to restarts in the population, which generates the dependence of a greater number of generations. The results also showed that DE-CSMD can present better results than other optimization techniques, such as genetic algorithms and the least squares method.