A PSO-based optimal tuning strategy for constrained multivariable predictive controllers with model uncertainty.

Area of Science:

• Control Engineering
• Optimization Techniques
• Process Systems Engineering

Background:

• Model uncertainty poses significant challenges for Model Predictive Control (MPC) performance.
• Tuning constrained MPC algorithms requires robust methods to handle worst-case scenarios.
• Multi-objective performance criteria are essential for complex industrial processes.

Purpose of the Study:

• To develop an optimal tuning method for constrained Model Predictive Control (MPC) algorithms.
• To address model uncertainty by incorporating worst-case control scenarios.
• To apply a robust optimization technique for solving the tuning problem.

Main Methods:

• Formulation of the tuning problem using the Morari resiliency index and condition number.
• Inclusion of a nonlinear multi-objective performance criterion.
• Application of a modified particle swarm optimization (PSO) technique to solve the mixed-integer nonlinear optimization problem.

Main Results:

• Successful development of a PSO-based method for tuning constrained MPC with model uncertainty.
• Demonstration of the method's effectiveness on the Shell heavy oil fractionator process.
• Validation of the worst-case control scenario approach for robust MPC tuning.

Conclusions:

• The proposed PSO-based tuning method offers an effective solution for constrained MPC under model uncertainty.
• The approach provides a robust framework for optimizing control performance in challenging industrial applications.
• This work contributes to the advancement of advanced process control strategies.