Performance assessment and monitoring of MPC with mismatch based on covariance benchmark

Model predictive controller has been widely used in the industry process. Model-plant mismatch will directly affect the level of safety and economic benefits. So, Performance assessment and monitoring of MPC with mismatch is very important. A data-based covariance benchmark derived from reference is successfully illustrated for control performance assessment and monitoring the Wood-Berry tower MPC process with mismatch. Choose the outputs without mismatch as the benchmark data and the outputs with mismatch as the monitored date. The worse or better performance directions in the monitored period are identified using generalized eigenvalue analysis. A statistical inference method is further developed to obtain the confidence intervals of the eigenvalues. The covariance-based performance indices within the isolated worse and better performance subspaces are then derived to assess the performance degradation or improvement. These simulations show the feasibility and effectiveness of the method.