Performance Monitoring for Model Predictive Control

Megan Zagrobelny

The goal of performance monitoring is to determine whether or not a control system is performing acceptably. This task requires analyzing large quantities of input and output data to determine whether or not the loop has acceptable behavior. When there are a large number of loops, it becomes increasingly important to have a systematic and automated method of evaluating the performance of each loop. In addition, performance monitoring methods should aid in diagnosing the problems behind unacceptable performance. As model predictive control has become more popular in industrial applications, it is important to develop monitoring and diagnosis methods that can be used for MPC systems, which often include a large number of interacting variables. While many conventional monitoring methods consider only how to minimize the output variance, a good MPC monitoring tool will consider all of the goals of the MPC, including limiting the amount of control action and any constraints on the system.