Predictive Functional Control Technology And Its Application In Three Tank-water Control

Model Predictive Control (MPC) is a control algorithm to predict the futurebehavior of the controlled system using the dynamic models and to optimize control ofthe model. It has been widely used in practical industrial control, and it is one of themost successful technologies in advanced industry control. In the practical industrialapplications, model predictive control algorithm based on linear model is already quitemature. But the actual controlled object is often highly non-linear, the problem thatnon-linear system control is still the research difficulty and hotspots of the currentcontrol field. With the expansion of industrial production scale and the ascension of thecontrolled object’s complexity, it is difficult to obtain accurate identification model forthe controlled system.Support Vector Machine (SVM) is based on the principle ofminimized structural risk, it has excellent performance in accuracy and generalizationability when dealing with nonlinear controlled object classification and regressionproblems. This paper aimed at the nonlinear characteristics of the controlled object, andstudied, with the combination of the intelligent modeling method based on Libsvm andpredictive functional control to solve nonlinear control problems. The main contents areas follows:(1) Access to a large number of documents, we have a further understanding to thedevelopment trend and possessed results of model predictive control. Especially weconducted a more in-depth study on predictive function control, recognizing theadvancement of predictive function in solving the linear model predictive control ofsuperiority.(2) Support vector machines have outstanding performance in nonlinear systemmodeling, this paper mainly studied Libsvm toolbox based on support vector machine.The toolbox can solve the nonlinear system modeling problem.(3) We have studied the nonlinear predictive function algorithm which is based onLibsvm. The method mainly takes Libsvm regression to create a nonlinear predictionmodel of the system, then the created nonlinear model is linear processed, last step ispredictive function controlling to the linearized model. The simulation results show thatthe performance of the method of controlling is good, and there is a strong suppression ability to interference.