Research On Hammerstein Model-based Nonlinear Model Predictive Control

Linear model-based model predictive control has been widely used in industrial processes,with substantial profits obtained. But nonlinearity exists widely in industry. For a plant withstrong nonlinearity, traditional linear model predictive control can’t obtain satisfactory resultsany more. In this case nonlinear model and predictive control algorithm must be considered.Hammerstein model is one kind of nonlinear model, which can describe many industrial plants,so it is widely applied and attracts many researchers. With the wide application of modelpredictive control technology, it is realized gradually that though designed as requested, thepredictive controller will deteriorate because of many factors such as model mismatch,unmeasured disturbance, equipment faults and so on. In such cases if we don’t redesign thecontroller, the control system performance will become worse. So performance assessment ofpredictive controllers is of great significance too.This thesis first introduces the Hammerstein model structure and researches on Hammersteinmodel identification briefly, then focuses on identification, control and performance assessmentof the nonlinear Hammerstein model predictive control with constrains, and the realization ofadvanced control algorithms based on MATLAB and OPC interface. The major results of thethesis are as follows：1. A dual recursive extended least squares identification method is proposed for theHammerstein model with ARMAX linear subsystem. At first the nonlinear static part and lineardynamic part of the model are decoupled with input signals specially designed, then theparameters of each part are estimated respectively with recursive extended least squares.2. The nonlinear Hammerstein model predictive controller with constraints is realized, and thedesign performance index is adopted to assess the performance of the control system. 3. Using OPC interface technology, the advanced control platform is established for datacommunication between MATLAB and S7-300PLC, which ensured the experimental researchon the identification, control and system assessment of the nonlinear system.