Research And Application Of Generalized Predictive Control Of Nonlinear Systems

All actual systems have a certain degree of nonlinear features. Though mostweak-nonlinear system can be solved by linear method, however,for some systems withstrong nonlinear features,high noise and heavy time delays, normal theory fails. Theexisting GPC based on Hammerstein model could divide a nonlinear system into twoparts(linear and nonlinear part) and solve separately by approximate method. However,it is hard to be widely used because of problems of mass calculation, hard to model andlow stability.This paper presented a new method of generalized Hammerstein model based onCARIMA model to these problems. This method approximate output by Taylor serieswith increment minimized model of dynamic approximate and identify algorithmparameters by recursive adaptive prediction. This method avoids solving complexDiophantine equation,which could simplify calculation in a large scale. What’s more,thealgorithm promotes control accuracy of system by using recursive adaptive predictionmethod for algorithm parameters,which has a better real-time performance, anti noiseability and robustness.This paper also simplified Hammerstein model by neglecting items of havinginsignificant effect on system based on original model when recursively solvinginput,which simplified calculation further.GPC based on Generalized Hammerstein model proposed on this paper could solveproblems of long-time optimization calculation,recursive predict of controllerparameters and noise estimation in control law,which has good control effect. Computersimulation shows that improved Hammerstein model has a better performance.