Statistical Online Model Quality Monitoring for Closed-loop Control System

The performance of model-based control system depends a lot on the model fidelity. The mismatch between the model and the true plant plays an important role in poor control quality of the closed-loop control system. In this paper, a new statistical online model quality monitoring method for closed-loop systems is presented, which monitors model quality only requiring the real-time routine operation data. The proposed method adopts the moving window method to online update model evaluation index which is obtained from the closed-loop input/output data. Then the statistical process control (SPC) technique is used to monitor model quality by Shewhart control chart. The approximate statistical distribution of the model evaluation index can be obtained by adaptive Least absolute shrinkage and selection operator (Lasso) method. The simulated examples are presented to verify the effectiveness of the proposed monitoring method.