Performance based tuning strategies for model predictive controllers

Over the past decade, Model Predictive Control (MPC) has established itself as an industrially important form of advanced control. It differs from classical proportional-integral-derivative (PID) control by employing a model of the process internal to the controller architecture. This model provides the controller with advanced knowledge of the impending impact of operational changes or process upsets and results in improved controller performance. However, the large number of adjustable parameters that alter closed loop performance, make tuning MPC for desired performance a challenging task.; This research effort addresses the important issue of tuning Dynamic Matrix Control (DMC), which is the process industry's standard for MPC. Specific configurations of unconstrained DMC, including one-move single-input single-output (SISO) DMC, multi-move SISO DMC and multiple-input multiple-output (MIMO) DMC, are considered individually. Analytical expressions that compute DMC tuning parameters are derived. These expressions are then used to formulate tuning strategies for the various DMC configurations. Finally, capabilities and limitations of the tuning strategies are investigated through computer simulations.; Successful culmination of this work provides a quantitative understanding of the impact of DMC tuning parameters on closed loop performance and simplifies the task of DMC tuning to a set of step-by-step procedures that are easy-to-use and reliable. Just as tuning rules such as Cohen-Coon, ITAE and IAE proved valuable for PID implementations, the strategies developed in this work are significant because they offer an analogous approach for tuning DMC.