A Framework for Implementation of Dynamic Discrepancy Reduced-Order Modeling in Advanced Process Control.

Area of Science:

• Chemical Engineering
• Control Systems Engineering
• Computational Modeling

Background:

• Advanced process control relies on accurate dynamic models.
• Reduced-order models offer computational efficiency but often sacrifice accuracy.
• Existing methods for model mismatch correction are limited.

Purpose of the Study:

• To introduce a novel dynamic discrepancy reduced-order modeling framework.
• To enhance model accuracy and computational performance in process control.
• To develop a gray-box modeling approach for advanced process control applications.

Main Methods:

• Constructing gray-box models combining first-principles and black-box components.
• Focusing on discrepancies in the rates of change within the reduced-order model.
• Utilizing moving horizon estimation for data generation and Bayesian inference for calibration.

Main Results:

• The proposed dynamic discrepancy approach compensates for lost dynamic information in reduced models.
• Improved model accuracy and computational performance were demonstrated.
• The framework was validated using a Fischer-Tropsch synthesis reactor simulation.

Conclusions:

• The novel framework effectively implements dynamic discrepancy reduced-order modeling.
• This approach enhances the suitability of reduced-order models for model predictive control (MPC).
• The method offers a balance between computational complexity and model accuracy for advanced process control.