Research On Performance Monitoring Of Multivariate Model Predictive Control Loop

As the representative of the advanced control, Model Predictive Control(MPC) has made a great progress in complex industrial process. The study about its performance assessment and diagnosis technology is the new development trends of Model Predictive Control in recent years. In this paper, for performance monitoring of multivariable model predictive control loop, we studied a kind of performance evaluation method for MPC based on the model prediction residual closed-loop potential index and a kind of performance diagnosis approach based on the weighted subspace distance of L2-Hausdorff.The main research works are as follows:Considering that the model prediction residual of MPC controller can effectively reflect the information of model mismatch, we propose a kind of performance evaluation method for MPC based on model prediction residual closed-loop potential index. This method can effectively detect the potential performance change of the MPC controller caused by model mismatch, further detect the possible mismatched control loop, and can also provide technical support for system maintenance in the actual production. We carried out the simulation experiment in the Continuous Stirred Tank Heater(CSTH) system and the proposed method proved to be feasible and valid.For the current problem of locating the degradation source of performance using performance diagnosis method of MPC control loop, we give a kind of performance diagnosis approach based on the weighted subspace distance of L2-Hausdorff. Considering the distribution characteristics of effective data information, this method can use the performance deteriorating subspace to describe the loop characteristics of each deterioration conditions, construct the weighted subspace distance of L2-Hausdorff to measure the similarity of current loop performance model and the known ones, and locate the worsen source leading to loop performance degradation through distance clustering. In the end, we carried out the simulation experiment in the CSTH system to test the feasibility and validity of the method.In order to further verify the application results of the performance evaluation method based on the model prediction residual closed-loop potential index and the performance diagnosis approach based on the weighted subspace distance of L2-Hausdorff, we carried out experimental research about the performance evaluation and diagnosis of MPC control loop in the actual process control experimental device in our laboratory. The experimental results demonstrated that that the research methods about the performance evaluation and diagnosis of MPC control loop have good effects.