Skip to Main Content
A methodology is presented to identify parameters of non-linear models of excitation systems (ESs). Based on the use of genetic algorithms (GAs), the proposed methodology carries out simultaneous parameter identification of linear and non-linear model components. The computational algorithm allows to adequately identify model parameters and it is not affected by the noise present in the measurements. The application of this methodology was developed to identify and validate ES models of different technologies that are used in stability studies through dynamic simulations. First, model parameters of DC1A and ST1A type ES were determined in a simulation environment. The performance of two identifiers based on a GA paradigm is analysed: GA with arithmetic and intermediate recombination operators (GA-BASE) and GA based on differential evolution (GA-DE) mutation. Then the GA-DE identifier is applied to estimate parameters of a static ES (EXE) model of a Brazilian hydro power plant utilising measurements corrupted by noise and registered during field tests. The results obtained are satisfactory and the responses of the identified models are close to real system measurements.