Identiifcation And Control For A Class Of Multirate Systems

Abstract：Incomparable frequency multirate systems (ICFMS) widely exist inindustrial areas, such as petrochemical industry, communication system andbiomedical system, which have wide prospect of application. Multiratesystems have two or more sampling frequency, and own a more complicatedmathematical model than the traditional discrete system. Multirate systemshave nonlinear, large noise and strong coupling characteristics. It is difficultto modeling and control by the traditional methods. The study of modelingand control in multirate systems has attracted wide attention both at home andabroad. Therefore, how to identify and control multirate systems effectively isnot only significant in theory, but also potentially values in practice use. Thus,this dissertation is supported by the National Natural Science Foundation(NO.60905034) and the Natural Science Foundation of Zhejiang province(NO.Y1080950), used data preprocessing idea based on feature, the leastsquares iterative identification principle, support vector machine inversionmethod and multi-model method, the identification and control of ICFMS hasbeen studied. The major contributions of this thesis are as following:(1) The algorithm of data preprocessing of ICFMS based on feature ispresented, which convert the ICFMS into CFMS. Simulation results confirmedthat the data preprocessing method is correct and effective.(2) The identification of two-output ICFMS is studied. For the unknownvariables in system identification, we replace the unmeasurable variables bytheir iterative estimates by the iterative identification idea. The least squaresbased iterative algorithm is studied using the iterative method. Thedissertation also considers the noise’s influence on identification performance.(3) The decoupling control of two-output ICFMS is studied. Then itpresents a decoupling control method based on support vector machineinversion method. The inverse system controller is designed by the study andapproach ability of the support vector machine in small sample. For thecoupling problem in different loop, we linearize the system by the inversion method.(4) The decoupling control of300MW fossil power unit is studied. Then itpresents a coordinate control method based on support vector machineinversion method. In order to solve the non-linear characteristics of unit underthe varying load, a new coordinate control method is designed by themulti-model and support vector machine inversion method. Finally, we testconvergence properties of the proposed algorithm involved by simulationexamples.