Multivariable Closed-Loop System Identification Based On Particle Swarm Optimization

In recent years, closed-loop system identification has always been the research focus in automatic control field. For the closed-loop system identification,on one hand,the effective information is less than the open-loop system;on the other hand,because of the presence of feedback,the system output noise is always relevant to the input siginals.Based on the above reasons,it leads many classical identification methods unable to be used for the closed-loop control system directly.In this paper,a new method CPSO applying to any input siginal is presented to solve the problems of the identification for multivariable system.First of all,we introduce the background and the purpose of my research as well as the basic konwledge of least squares method,the instrumental variables and an improved method is brought in to estimate the parameters of Box-Jenkins model. We compared the noise model for each identification method and do the simulation for each method.Then a particle swarm optimization algorithm is proposed to identify the structural characteristics of the closed-loop delay system model.Through the simulation,we know the feasibility of this method. Although the above method can be used to identify the closed-loop delay system,it is easy to fall into local optimum which leads to the identification result is not accurate enough.For the above reasons,a novel method CPSO is proposed to identify the closed-loop system.This method is mainly achieved through the unique ergodicity characteristics of the chaotic system.It can prevent the search process into local minimum evolutionary algorithm optimization mechanisms.We used the method to complete the simulation for the single and multiple system.For the single variable system identification,we use the method that a frequency response estimation based on CPSO.For the multivariable system identification,the MIMO system is decomposed into many SISOsystem,then we use the above method to identify every SISO system.Finally,we get all the parameters of the multivariable system.