Dynamical tuning for MPC using population games: A water supply network application.

Area of Science:

• Control Engineering
• Optimization Theory
• Game Theory

Background:

• Model Predictive Control (MPC) is effective for large-scale systems with multiple constraints and objectives.
• Prioritizing multiple control objectives in MPC is crucial but challenging, especially with time-varying system parameters or disturbances.
• Static prioritization methods may be suboptimal when system dynamics change over time.

Purpose of the Study:

• To develop a dynamical tuning methodology for multi-objective Model Predictive Control (MPC).
• To address the challenge of time-varying priorities in control objectives for large-scale systems.
• To leverage evolutionary game theory for adaptive control strategy optimization.

Main Methods:

• Employing evolutionary game theory to dynamically adjust the prioritization of control objectives within the MPC cost function.
• Implementing a multi-objective MPC framework capable of real-time adaptation.
• Testing the proposed methodology on a large-scale water supply network simulation.

Main Results:

• The proposed dynamical tuning method effectively adapts control priorities in response to periodic time-varying disturbances.
• The evolutionary game theory approach demonstrated superior performance compared to static tuning in the tested water supply network.
• The controller successfully managed multiple operational and physical constraints under dynamic conditions.

Conclusions:

• Evolutionary game theory provides a robust framework for the dynamical tuning of multi-objective MPC.
• The developed methodology enhances control performance for large-scale systems with time-varying characteristics.
• This approach offers a significant improvement over traditional static tuning methods in complex, dynamic environments.