Identification of Non-Linear Systems Using Hammerstein Model and Genetic Algorithm

Abstract

This paper presents an approach for the identification of non-linear dynamical systems which applies simple Hammerstein model integrated with Genetic Algorithm method as parameter estimator. The simple Hammerstein model consists of a non-linear static and a linear dynamic term separated into model blocks and if these blocks serially combined, it builds a non-linear model. The genetic algorithm method, which is known as populationbased optimization, will be applied to estimate the systems parameters, in order to find the best solution based on the fitness value of each model candidate. To validate the obtained model, several criteria i.e. residual and visual tests as well as evaluation of the loss function are performed. For residual analysis, statistical criteria using autocorrelation and crosscorrelation funtions are used to select the best model order representation. To evaluate the performance of the proposed algorithm, three models with different non-linear term have been used during the simulation studies. The results show how the non-linear dynamics of the models can be identified accurately by the proposed algorithm.