Model selection and parameter estimation of nonlinear system based on PSO

A new method for model selection and parameter estimation for Hammerstein model is presented using particle swarm optimization (PSO). The error rule is proposed to decrease computation and obtain the true optimal structure effectively. The modified identification algorithm is always convergence by adding a backward algorithm. Meanwhile, it can obtain a high precision for the parameter estimation. The experimental results illustrate that the residual variance is an efficient selection criterion, but Akaike´s information criterion (AIC) and minimum description length (MDL) criterions are not fit for the structure identification of the nonlinear system.