Genetic Algorithm in Identifying Non Linear Auto Regressive with Exogenous Input Models for Non Linear Systems

A Genetic Algorithm (GA) based system identification method is developed to automatically construct linear-in-the-parameters non-linear difference equation models of nonlinear systems of unknown structure from observations of inputs and outputs. Three simulated examples, NARX system corrupted with noise, the Mackey-Glass equation and the Van der Pol oscillator, are identified to demonstrate the effectiveness of the algorithm and to compare it with the orthogonal parameter estimation algorithm.