Real-time comparison of a number of predictive controllers

Area of Science:

• Control Engineering
• Automation Systems
• Systems Theory

Background:

• Model Predictive Control (MPC) encompasses various algorithms influenced by system matrix conditionality and cost functions.
• Existing MPC schemes vary in complexity and performance characteristics.

Purpose of the Study:

• To introduce and compare newer MPC schemes like extended predictive control (EPC) and shifted MPC against established methods.
• To evaluate the control performance of different MPC algorithms on slow and fast-reacting systems.

Main Methods:

• Comparative analysis of MPC algorithms including EPC and shifted MPC.
• Simulation and evaluation of closed-loop responses for disturbance rejection and setpoint tracking.
• Assessment of system matrix ill-conditionality impact on controller performance.

Main Results:

• All evaluated MPC controllers demonstrated effective disturbance rejection and setpoint tracking.
• Extended predictive control (EPC) and generalized predictive control showed enhanced performance, particularly in mitigating system matrix ill-conditionality.
• Differences in control performance were explained by the inherent structures of the compared MPC schemes.

Conclusions:

• MPC algorithms designed to address system matrix ill-conditionality, such as EPC, yield superior control performance.
• The choice of MPC scheme significantly impacts performance, especially in systems with varying dynamics.
• Further research into robust MPC formulations for ill-conditioned systems is warranted.