Nonlinear System Identification Of Hammerstein Model Based On Firefly Algorithm

The development of control theory is based on the mathematical model of the system. The demand for the precision of control theory is higher, as the demand for the precision of mathematical model is higher. There are two ways of building a mathematical model, respectively named mechanism analysis method and identification. Comparing the two methods identification has a certain advantage, so it is widely used in industrial control. Now the theory of linear system identification has developed more mature and more perfect. However, there are many nonlinear systems in the real world, so the identification of nonlinear system is still an active branch of control field.There is not a consensus theory about the nonlinear system, all of the studies is only for a class of nonlinear model. In this paper, we mainly study on the nonlinear model of Hammerstein. This model is a kind of modularized model, it is composed of linear part and nonlinear part. The structure of Hammerstein model is simple and flexible, so it is widely applied to various fields. In this paper, we first introduce the traditional identification method of Hammerstein model, then carry on the simulation and analysis the experimental results.In this paper, we put forward an identification method of Hammerstein model based on the firefly algorithm. First, the basic principle and process of the firefly algorithm is introduced in detail. Aiming at the defects of the basic firefly algorithm, we propose an improved firefly algorithm based on nonlinear inertia weight. Secondly, we expound the specific steps of the Hammerstein model identification, this method imply the key term separation principle so that the system output can be expressed a regression equation. Then we build an auxiliary model, the auxiliary output and real output structure the error criterion, we use the firefly algorithm to minimize the error so as to get a set of estimated parameter values. Finally the simulation is carried on, the experimental results shows that the method in this paper is feasible and effective, and the improved firefly algorithm gets the identification values is more close to the real values.