Generalized Predictive Control Based On Multiple Kernel Fuzzy Least Squares Support Vector Machines

A large number of industrial produce processes have inherently nonlinear, time-variantand uncertain characteristics, however the control performance of linear model predictivecontrol is often not satisfactory. Thus, the research on nonlinear predictive control hasbecome an important issue in control field. Aim at nonlinear decontrolled plants at largethat exist in industrial processes, this thesis brings forward arithmetic of generalizedpredictive control algorithm based on multiple kernel fuzzy least squares support vectormachines Main contents as follows:1) A new algorithm based multiple kernel fuzzy least squares support vectormachines is proposed, and used for predictive model. Then by combining the generalizedpredictive control, a generalized predictive control algorithm based on multiple kernelfuzzy least squares support vector machines is proposed. And the simulation results showthe effectiveness of the presented method.2) A class of nonlinear system is replaced by a time varying linear system.Thecontroller directly is designed based on multiple kernel fuzzy least squares support vectormachines, and the controller parameters is adjusted adaptively by using the estimation ofgeneralized error. The convergence and stability of the algorithm are proved in theory.The algorithm could effectively increase the noise immunity of least square supportvector machine. And an equivalent kernel is built by linear weighted combination of multikernels to reduce the dependence of modeling accuracy on kernel function and parameters.