Researches On The Identification Algorithms Of Hammerstein Nonlinear System

In real life,almost all practical systems are nonlinear,and only when the nonlinear degree of the research objects is not strong,they can be regarded as a linear object models.With the rapid development of science and technology,the industrial control systems are becoming more and more complicated,the requirements of precision control are increasingly high,and systems with complex nonlinearity cannot be a linear model approximation,so the nonlinear system modeling has become an important research direction of nonlinear systems.However,one of the difficulties in the identification of nonlinear systems is the lack of a unified mathematical model that describes the various characteristics of the nonlinear system.The Hammerstein model,due to its simple structure and flexibility,is widely used in many fields.Based on Hammerstein ARMAX model,from the linear deterministic algorithm to the nondeterministic identification algorithm,then to the nonlinear identification algorithm,and finally to the parameter optimization problem of identification algorithm,the thesis carries out targeted researches,and put forward a variety of new Hammerstein ARMAX identification algorithms to improve the identification accuracy of nonlinear model.The main work and innovations of this thesis include:1.A recursive instrumental variable algorithm(RIV)is proposed for the Hammerstein ARMAX system and the mean square convergence of the algorithm is strictly proved.The proposed algorithm is applied for numerical example,and is compared with the classical RLS algorithm.The identification results show the effectiveness of the proposed algorithm.The algorithm is superior to RLS algorithm in both identification accuracy and convergence speed under coloured noise;2.A recursive maximum likelihood algorithm(RML)is presented for the Hammerstein ARMAX system.The derivation of the proposed algorithm is given in detail.The proposed algorithm not only can avoid overparameterization,but also can identify the parameters of the coloured noise.The proposed algorithm is applied for numerical example,and is compared with the classical RLS algorithm.The identification results show the effectiveness of the proposed algorithm.The algorithm is superior to RLS algorithm in identification accuracy;3.An APSO WLSSVM algorithm is proposed for the Hammerstein ARMAX system.This algorithm is a combination of an adaptive particle swarm optimization algorithm(APSO)and the robust least squares support vector machine(WLSSVM).Based on the PSO algorithm,the evolutionary state estimation(ESE)technology is applied to adjust the inertia coefficient and acceleration factor of particles so as to speed up the convergence rate.Meanwhile,the mutation opertation is applied to avoid the algorithm falling into a local optimal solution,thus enhances the search ability of the PSO algorithm.The proposed algorithm is applied for numerical example,and is compared with a variety of algorithms.The identification results show the effectiveness of the proposed algorithm.The algorithm not only has high identification accuracy,but also is superior to PSO WLSSVM in terms of convergence speed;4.An AM FOA WLSSVM algorithm is proposed for the Hammerstein ARMAX system.This algorithm is a combination of an adaptive mutation fruit fly optimization algorithm(AM_FOA)and the robust least squares support vector machine(WLSSVM).Based on the FOA algorithm,by judging the population fitness variance and the global optimal value,the mutation opertation is applied not only to avoid the algorithm into a local optimal solution,but also enhance the search ability of the FOA algorithm.The proposed algorithm is applied for numerical example,and is compared with a variety of algorithms.The identification results show the effectiveness of the proposed algorithm.The algorithm not only has high identification accuracy,but also is superior to FOA_WLSSVM in terms of convergence speed.