Mixed-kernel least square support vector machine predictive control based on improved free search algorithm for nonlinear systems

Abstract

Many controlled objects in the actual industrial process are nonlinear systems, and the traditional control theory cannot achieve very good control effect. Based on swarm intelligence optimization algorithm, the nonlinear prediction and predictive control algorithm, this paper put forwards a nonlinear systems predictive control method based on the mixed-kernel least square support vector machine (LSSVM) model and improved free search (IFS) algorithm. The mixed-kernel LSSVM combines the advantages of radial basis function (RBF) and the Polynomial function, which can achieve a better prediction and modeling accuracy. The optimal parameters of the mixed-kernel LSSVM are obtained by IFS algorithm. The proposed predictive control method utilizes mixed-kernel LSSVM to estimate the nonlinear systems model and forecast the output of the system. The output error is reduced through output feedback and error correction. The rolling optimization of control variables are obtained by IFS algorithm. This predictive control method can be used to design effective controllers for nonlinear systems with unknown mathematical models. The stability analysis shows that the control method is asymptotically stable. The simulation experiment of single input and single output, multiple input multiple output and continuous stirred tank reactor nonlinear systems are performed. The validity of the proposed control method is also verified by an actual electric heating furnace system. The simulation and practical experiment results show that the proposed predictive control method has good tracking signal and anti-interference ability.