Use of swarm intelligence for the identification of a class of nonlinear dynamical systems

Abstract

In this work, properties of swarm intelligence are exploited to solve the identification problem of a class of nonlinear dynamical systems known as Hammerstein systems. Without any assumption on the structure of nonlinearity of the system, the nonlinearity is modeled using artificial neural network. Synaptic weights of the neural network are estimated via a particle swarm-based learning routine. Linear dynamics of the system are modeled using state space model. A recursive algorithm is developed to estimate the two components of the Hammerstein system. Numerical Monte-Carlo simulations are performed to test the reliability and repeatability of the identification technique presented. Identification is carried out with noisy data, having low signal-to-noise ratio (SNR). The presented identification technique provides encouraging estimation results. © 2012 Springer-Verlag GmbH Berlin Heidelberg.