Modular System Identification Based On Improved Particle Swarm Optimization

Nonlinear systems are widely used on many fields such as biomedicine,industrial engineering,social economy and so on.The identification method of nonlinear system has been a hot research direction for worldwide scholars in the identification field.At present,we have a complete classic traditional identification method,which can overcome the shortcomings of traditional estimation methods to a certain extent,but there are still some areas for improvement in these algorithms.This paper conducts deeply research on the parameter identification method of the single-input and single-output modular Hammerstein nonlinear system.Different improved particle swarm optimization(PSO)algorithms are designed for the three common Hammerstein models.The main work is described as follows:(1)According to the different linear modules,the classification and mathematical description of Hammerstein model are explained.Three classical identification methods are discussed.In addition,the defects and shortcomings of the classic PSO algorithm is proposed to solve the system identification of Hammerstein model.(2)For Hammerstein CARMA model,the particle swarm optimization algorithm based on orthogonal matching pursuit(OMP-PSO)is proposed.The orthogonal matching pursuit mechanism is applied to the PSO algorithm.The position updating method of the PSO algorithm is reconstructed according to the principle of orthogonal intersection between the iterative residual and the particle position.On the one hand,the order and parameters of the model are identified at the same time,which reduces the probability of accurate signal reconstruction.On the other hand,it accelerates the convergence speed of the identification system,and improves the global development capability of the algorithm.The numerical simulation results verify that the identification result of OMP-PSO algorithm is better than that of the standard PSO algorithm.(3)For Hammerstein OEMA model,the particle swarm optimization algorithm based on mixed conjugate gradient(HCG-PSO)is proposed.The idea of conjugate gradient is applied to the process of particle updating,which makes the randomness of PSO algorithm relatively controllable and strengthens the algorithm's local exploration ability in the solution space.The HCG-PSO algorithm overcomes the disadvantages of premature convergence.Finally,the optimal estimation value of the system parameters is obtained.The numerical simulation results verify that the HCG-PSO algorithm can be effectively identified.(4)For Hammerstein CAR model,the particle swarm optimization algorithm based on Levy distribution(LD-PSO)is proposed.In the iterative process of the PSO algorithm,the Levy flight method with jumping characteristics is applied to enhance the flexibility and step of particle position change.The LD-PSO algorithm also expands the traversal range of particles,so that the global optimization ability of the algorithm has been significantly improved,and the loss of population diversity is avoided.The numerical simulation results verify that the LD-PSO algorithm has a better parameter identification effect in the model.