Adaptive Multivariable MPC Controller Using Matrix Row Elimination
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current Model Predictive Control (MPC) systems face challenges in adapting to nonlinear processes and handling infeasible linear programs, particularly due to the inherent correlation of PID dynamics, which limits their ability to create adaptive controllers for complex multivariable processes like chemical manufacturing and oil refining.
Innovation Solution
A method is developed to remove PID dynamics from the MPC controller by interchanging final control element positions with PID controller set points using matrix row elimination, creating an open-loop finite impulse response model that allows for adaptive control and off-line simulator development, enabling the creation of an adaptive multivariable controller without requiring new identification testing.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If PID dynamics are included in the MPC controller model, then the controller can handle correlated variables, but the controller cannot be adapted to different configurations without re-identification testing
Solution Approach 1:
The patent extracts and removes PID dynamics from the process model by using matrix row elimination to interchange final control element positions with PID controller set points. This separation allows the MPC controller to operate with a simplified model that does not include PID dynamics, enabling easier adaptation to different configurations without requiring re-identification testing.
Solution Approach 2:
The patent segments the controller into two distinct parts: the MPC controller that handles multivariable optimization and the PID controllers that handle individual loop control. This segmentation is achieved by removing the correlated PID dynamics from the MPC model, allowing each component to function independently with reduced complexity.
2Reliability
If standard identification tests are used, then the dynamic model captures PID dynamics, but this prevents creation of MPC controllers free of PID dynamics
Solution Approach 1:
Instead of accepting PID dynamics as an inherent part of the identification process, the patent inverts the approach by using matrix row elimination to remove these dynamics from the model. This inversion allows the creation of MPC controllers that are free of PID dynamics while maintaining model accuracy for the underlying process.
Solution Approach 2:
The patent performs preliminary model transformation during the identification phase by interchanging variables using matrix row elimination. This preliminary action removes PID dynamics from the model before the MPC controller is deployed, avoiding the need for later modifications or re-identification when configuration changes are needed.
3Manufacturing precision
If PID controller configuration is fixed during identification, then the model reflects actual process behavior, but the controller cannot be retuned or reconfigured
Solution Approach 1:
The patent extracts PID controller configuration dependencies from the process model by removing PID dynamics through variable interchange. This extraction allows the MPC controller to be retuned or reconfigured independently of the original PID settings, providing flexibility while maintaining control precision through the adapted model.
Data Source
AI summary
A method is disclosed for developing and using a high-speed adaptive multivariable controller by removing the dynamics of the PID controllers from a Model Predictive Controller that was developed using identification testing of a process. The resulting multivariable controller, based on final control elements as the manipulated variables is then used in a novel control adaptation with all of the PID controllers switched to manual.