Hammerstein and Wiener nonlinear models identification using a multimodal particle swarm optimizer

A new solution to nonlinear systems identification based on Hammerstein and Wiener models, is developed using tools from evolutionary algorithms. A particle swarm optimizer including a dynamic niching procedure is resorted to find, on one hand, the parameters of the dynamic model of the linear part and, on the other hand, the coefficients of the polynomial approximations of the nonlinear input and output elements. The performances of the proposed solution which are illustrated through a numerical example show that PSO is well suited to deal with system identification problems involving high order dynamics and strong nonlinearities