Studies Of System Identification Algorithm

To control the system needs to know the system model,but the complexity of the actual object model is not fully known,this needs to identify system,system identificatio n is necessary precondition to realize accurate control.For some of the actual object,due to its complexity,it is difficult to obtain mathematical model using the method of theoretical analysis.At this time it is needed to identify systematic mathematical model or parameters to through the input and output data of experiment test.system identification,state estimation and control theory are the three pillars of modern control theory,as the requirement of system’s complexity and precision of the model,system identification algorithm in continuous development.In order to improve the accuracy of identification,for general linear system,this paper mainly studies the series of classical recursive identification method that based on general least squares and maximum likelihood method to reduce noise interference on the system identification and to achieve the online identification.For complex nonlinear system,this paper studies the swarm intelligence optimization identification method and the neural network recognition method for identification of some modern methods.Swarm intelligence optimization identification method in the main genetic algorithm and particle swarm optimization(pso)algorithm is studied combined with reducing noise method of classical identification method in removing the influence caused by noise in nonlinear system identification and achieve the purpose of improve the identification accuracy.Neural network identification method of BP neural network combine with Deep Neural Networks for nonlinear systems is presented in this paper the problem of modeling,and used to solve nonlinear systems are studied when the noise identification method,including the BP neural network for data compression method or deep neural network to the system output data feature extraction method.Finally,simulating through an example to verify the effectiveness of these methods through the theoretical research and simulation analysis the following conclusions are drawn: in reducing noise especially colored noise impact on the precision forming part as a noise filter model,the filter parameters are identified with system is a very significant effect;By using a genetic algorithm and particle swarm algorithm combined with classical ident ification methods,can complement each other and fully display their respective advantages,to the problem of nonlinear system identification by noise provides a better identification method;Through BP neural network for data compression method or deep ne ural network to data feature extraction method can fully split the characteristics of the system itself and noise,thus greatly reduces the system under the interference of noise,for accurate modeling system.