Multi-criteria Decision And Multi-objective Optimization For Constructing And Selecting Models For Systems Identification
Free (open access)
31 - 42
J. Hernández-Riveros & A. Arboleda-Gómez
An alternative form for the identification of dynamic systems with the application of multi-objective optimization concepts, through the evolutionary algorithm MAGO is presented. A computational tool using operational data of a SISO system has been designed to automatically perform the construction and selection of the best model representing it. After a data acquisition, strategies for the system identification by parametric modelling are developed. The application on the fitness function of appropriate criteria to choose models representing the system is also studied. Different models (ARX, ARMAX, and OE) are built and compared. The models obtained, by evolution, provide better fit and final prediction error regarding that chosen by an expert. The computational effort is low considering that the proposed method is more effective on identification of dynamic systems. Applying this evolutionary method to more complex systems such as MISO, MIMO, and non-linear is proposed as future work. Keywords: computational and experimental methods, system identification, evolutionary computation, multi-criteria decision, multi-objective optimization.
Keywords: computational and experimental methods, system identification, evolutionary computation, multi-criteria decision, multi-objective optimization.