Based On Particle Swarm Alorithm System Identification Methods Research And Simulation

Based on the modern industrial process, scholars have proposed many advanced control technologies, but most of the modern technologies are based on exact objects mathematical model. System identification is a kind of effective method to establish a mathematical model of the process object. At present, we already have perfect and mature traditional identification methods. At present, the most popular identification algorithm including:least-square algorithm, genetic algorithm, differential evolution algorithm, particle swarm algorithm, etc, which can overcome shortcomings of the traditional estimation method in a certain extent. But, there are some improvements measures can be put forward to those methods.In this thesis, based on the standard particle swarm algorithm, we put forward some improvement measures, and the improved algorithm is called improved second-order particle swarm algorithm. Then, the improved algorithm is used in a class of industrial process model structure which can be described as block in the nonlinear system model. The simulation results show that the application of the improved algorithm in the nonlinear system is very effective.Then, aimed at a multiple inputs and a single output static system, we put forward a new method of system identification. This new method can achieve identification of structure and parameters in a multivariate system. The basic theory of this method is:the conbination of classic mathematical models constitute a system model, so the problem of system structure identification is transformed into a problem of combinatorial optimization. Then using hybrid particle swarm optimization algorithm and the sequential quadratic programming algorithm realizes both structure and parameter identification of the system at the same time. The simulation results show that the proposed new method is availably, and the given hybrid algorithm is effective, high precision.