A model mismatch assessment method of MPC by decussation

Area of Science:

• Control Engineering
• Process Systems Engineering
• Chemical Engineering

Background:

• Model Predictive Control (MPC) systems are crucial for process optimization and stability.
• Accurate assessment of Model Plant Mismatch (MPM) is essential for reliable MPC performance.
• Existing methods for MPM assessment in multivariable systems require improvement for precise sub-model identification.

Purpose of the Study:

• To propose and validate a novel method for assessing Model Plant Mismatch (MPM) in multivariable MPC systems.
• To accurately locate sub-model mismatches within complex multivariable control structures.
• To enhance the robustness and reliability of MPC applications through improved MPM assessment.

Main Methods:

• Development of a Correlation Analysis between Input and Disturbance (CAID) index to quantify MPM.
• Integration of CAID with Model Quality Index (MQI) for comprehensive MPM assessment.
• Utilizing Principle Component Analysis (PCA) to construct a synthetical disturbance variable for multivariable systems.
• Introduction of a decussation method combining synthetical CAID and MQI for precise mismatch localization.

Main Results:

• The proposed CAID method effectively assesses MPM in multivariable MPC systems.
• The synthetical CAID index, derived using PCA, enables accurate identification of sub-model mismatches.
• Validation on the Wood-Berry distillation process and an industrial air separation process demonstrates the method's efficacy.

Conclusions:

• The combined CAID and MQI approach provides a robust framework for MPM assessment in multivariable MPC.
• The decussation method offers precise localization of sub-model mismatches, improving system diagnostics.
• This work contributes to the advancement of reliable and high-performing MPC systems in industrial applications.