Meta-heurísticas aplicadas à identificação de sistemas

Abstract

System identification has the goal to determine mathematical models to describe dynamic characteristics of systems from observations. The identification process is generally divided into the following steps: i) experimental data collection, ii) determination of model structure, iii) parameter estimation and iv) model validation. In this work, the problem determining of structures is investigated. An algorithm was developed to determine the structure of polynomial NARX models using optimization techniques known as meta-heuristics. Unlike traditional methods, metaheuristics use a set of possible solutions and strategies, usually based on nature, to find the solution of the case applied. Among the techniques studied are the genetic algorithm, the particle swarm optimization, and the bat algorithm. The methodology proposed in this work was applied to identify three experimental examples: an electric heater, a buck converter and a pneumatic valve. The results demonstrate that metaheuristics can be applied to the problem of the selection of polynomial NARX model structures.