Design And Implementation Of MPC System With Remote Model Error Detection

Model Predictive Control(MPC) came up as a model-based control method in the late1970s, which has been used in industrial fields such as petroleum, chemical, electricity and aircraft industries. Dynamic Matrix Control, Model Algorithmic Control, Predictive Control and Internal Model Control, all of these are developed on the base of Model Predictive Control. However, the precondition of using these new control strategies is that engineers master the model predictive control theory and system identification very well. Dynamic model plays an important role because its accuracy affects the control performance directly. Thus, modeling and identifying industry plants is the difficulty of industrial model predictive control projects. The deviation can be caused by:slow changing of industrial process objects; aging of sensor instrument; and changes in the production environment. Diagnosing and modifying mismatch models in time during the industrial processes can reduce the high cost due to maintenance dynamic model and the skill requirement of MPC for users.This thesis depended on the mismatch of dynamic models, studies the design and implementation of the integrated system with multi-function such as MPC operation control, system identification and model detection. The development process has great realistic significance in MPC applied extension of process control. An implementation scheme of the combination of Remote Model Detection and MPC System has been put forward in this paper. On the basis of digesting the diagnostic algorithm which uses additional small amplitude test signals, experimental client and remote detection service are set up by using a centralized architecture of distributed system design idea. XML-RPC communications network management interface takes charge of data exchange among nodes and realizes the physical separation between model error detection system and controller in industry field which provides a simple and effective approach for remote model error detection. The remote implementation of the integration of MPC operation control, system identification and model detection can help users to identify the mismatch models selectively according to remote model detection results. It ensures the MPC control performance, reduces the extra cost of the maintenance and provides a better guarantee of economic, efficient and safe operation in industrial process so that the practicality of the MPC product can be fully promoted.