MPC Controller Robustness Against Near Collinearity
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Model Predictive Control (MPC) controllers in industrial processes face performance issues due to model uncertainties and near collinearity among variables, leading to oscillations, constraint violations, and undesirable operating points.
Innovation Solution
A method to enhance the robustness of MPC controllers by introducing a user-defined robustness factor, adjusting steady-state targets, and implementing a dynamic move plan to avoid weak directional movements, while incorporating an economic objective function with tolerance and constraint relaxation to stabilize the process.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If MPC controller uses standard control calculation, then controller structure is simple, but controller performance deteriorates due to model uncertainties and near collinearity causing oscillations and constraint violations
Solution Approach 1:
The patent implements dynamic adjustment of control targets and move suppression factors based on real-time detection of near-collinear relationships and model uncertainty. The controller adapts its calculation approach by modifying steady-state targets and move plans when collinearity is detected, rather than using a fixed control strategy. This dynamic adaptation resolves the contradiction by making the control calculation complexity variable - simple when not needed, complex only when model uncertainty and near-collinearity are present.
Solution Approach 2:
The patent changes key control parameters including steady-state targets, move suppression factors, and economic objective function weights based on detected system conditions. When near-collinearity is detected, the controller modifies these parameters to avoid problematic control moves. This parameter adaptation allows the controller to maintain reliability under uncertain conditions while avoiding unnecessary complexity when conditions are favorable.
2Stability of the object's composition
If MPC controller adjusts steady-state targets to avoid weak directional movements, then oscillations are reduced, but calculation complexity increases
Solution Approach 1:
The patent performs preliminary detection of near-collinear relationships and preliminary calculation of modified steady-state targets before executing control moves. By identifying problematic directions in advance and pre-calculating alternative targets, the controller avoids oscillations without requiring complex real-time adjustments during execution. This preliminary action resolves the contradiction by shifting computational complexity to the planning phase while maintaining stable execution.
Solution Approach 2:
The patent introduces an intermediary detection and adjustment mechanism that identifies near-collinear relationships and mediates between the ideal control target and the practically achievable target. This intermediary layer calculates modified steady-state targets that avoid weak directional movements while maintaining overall control objectives. The intermediary resolves the contradiction by providing a computational bridge that achieves stability without excessive complexity.
3Reliability
If MPC controller implements constraint relaxation with robustness factor, then constraint violations are reduced, but control precision decreases
Solution Approach 1:
The patent applies partial constraint relaxation only in specific situations where near-collinearity and model uncertainty are detected, rather than uniformly relaxing all constraints. The robustness factor is applied selectively to modify only the affected constraints, maintaining precision for unaffected constraints. This partial action resolves the contradiction by applying relaxation only where necessary to prevent violations, preserving control precision elsewhere.
Solution Approach 2:
The patent applies different levels of constraint strictness locally based on detected system conditions. When near-collinearity is detected in specific variable relationships, the controller applies relaxed constraints only to those affected variables while maintaining strict constraints on other variables. This local differentiation resolves the contradiction by protecting constraint satisfaction where needed while preserving control precision where possible.
4Reliability
If MPC controller detects and avoids near collinear relationships, then undesirable operating points are avoided, but detection and calculation complexity increases
Solution Approach 1:
The patent replaces complex mathematical detection of near-collinear relationships with a practical computational test based on the determinant of the gain matrix or a similar numerical criterion. Instead of requiring sophisticated mathematical analysis, the controller uses a straightforward computational check that is easy to implement and evaluate in real-time. This substitution resolves the contradiction by replacing difficult detection with a simple, reliable computational test.
Solution Approach 2:
The patent uses a simple, computationally inexpensive test for detecting near-collinearity that can be performed quickly at each control cycle. The detection mechanism is designed to be computationally lightweight, requiring minimal processing resources and time. This inexpensive detection approach resolves the contradiction by providing reliable operating point selection without significant computational burden.
Data Source
Figure 1
Figure 2
Figure 3
AI summary
A method, apparatus, and computer program product for increasing closed-loop stability in a MPC controller controlling a process where there are significant uncertainties in the model used by the controller. This invention focuses on the improvement of the robustness of the steady-state target calculation. This is achieved through the use of a user defined robustness factor, which is then used to calculate an economic objective function giveaway tolerance and controlled variable constraint violation tolerance. The calculation engine uses these tolerances to find a solution that minimize the target changes between control cycles and prevent weak direction moves caused by near collinearity in the model. If the controller continues to exhibit large variations in the process, it can slow down the manipulated variable movement to stabilize the process.