Reduced order modeling of linear MIMO systems using soft computing techniques

Abstract

A method is proposed for model order reduction for a linear multivariable system by using the combined advantages of dominant pole reduction method and Particle Swarm Optimization (PSO). The PSO reduction algorithm is based on minimization of Integral Square Error (ISE) pertaining to a unit step input. Unlike the conventional method, ISE is circumvented by equality constraints after expressing it in frequency domain using Parseval's theorem. In addition to this, many existing methods for MIMO model order reduction are also considered. The proposed method is applied to the transfer function matrix of a 10 th order two-input two-output linear time invariant model of a power system. The performance of the algorithm is tested by comparing it with the other soft computing technique called Genetic Algorithm and also with the other existing techniques. © 2011 Springer-Verlag.