A New Nonlinear System Identification Method Using Gene Expression Programming

A new method for identifying the nonlinear system model is presented, which is based on gene expression programming (GEP) and can obtain accurate nonlinear models automatically and effectively in the huge nonlinear model space. In this identification method the number of genes of chromosomes is no more fixed and the elements in the terminal set are also variable. It overcomes insufficiencies of the initial identifying method based on genetic programming (GP), reduces parameter dependency of evolution algorithm, and can identify various NARMAX models under the same parameters set. The definition of fitness considers fully the factors of the model´s accuracy and complicacy, and makes the solution can get a trade-off between the accuracy and the complexity. The simulation results show that this method is effective in obtaining the nonlinear models.