Closed-Loop MPC Step Testing with Giveaway Tolerance Control
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Multivariable Predictive Control (MPC) performance degrades over time due to changes in processes, primarily due to poor model accuracy, requiring frequent recalibration and causing interruptions in industrial operations, with previous methods often leading to overreaction and biased models.
Innovation Solution
A non-invasive closed-loop step testing approach that maintains process variables within optimal operating ranges by relaxing economic optimization, allowing for perturbation while minimizing disruptions, using a computer system with a MPC controller and target optimizer to adjust configurations and calculate new target values, and adding perturbation signals to the dynamic move plan.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If conventional re-test and re-identification approach is used to improve model accuracy, then model accuracy is improved, but it causes significant interruption to normal operation and requires weeks of intensive work
Solution Approach 1:
The system performs preliminary calibration by collecting process data and identifying model parameters during normal operation before actual process changes occur. This preliminary action prepares the model for future changes, allowing quick adaptation when process modifications happen, thus maintaining both accuracy and operational continuity
Solution Approach 2:
The MPC system automatically performs model calibration using its own collected process data without requiring external intervention. The system self-updates its dynamic models by utilizing accumulated operational data, eliminating the need for separate re-identification tests and maintaining continuous operation while improving model accuracy
2Reliability
If the process controller calculates new steady-state targets every control cycle to eliminate offset, then offset removal is achieved, but it causes unwarranted overreaction to plant noise and excessive feedback correlation
Solution Approach 1:
Instead of calculating new steady-state targets every single control cycle, the system applies partial action by updating targets at selected intervals or only when significant process changes occur. This reduces the excessive feedback correlation and overreaction to noise while still maintaining adequate offset elimination through periodic updates
Solution Approach 2:
The system implements periodic calculation of new steady-state targets rather than continuous calculation at every control cycle. This periodic action reduces feedback correlation in collected data sets while maintaining the integration action that removes offset, making the data suitable for model identification purposes
3Productivity
If purely closed-loop data is used for model identification to maintain continuous operation, then operational continuity is maintained, but the identified models are significantly biased and contain large errors
Solution Approach 1:
The system introduces an intermediary calibration process that uses collected closed-loop data to identify model parameters without requiring process shutdown. This intermediary step processes the closed-loop data through system identification algorithms, transforming it into accurate models while maintaining operational continuity throughout the calibration process
4Measurement precision
If process perturbation is applied to generate useful data for modeling purpose, then model identification accuracy is improved, but it causes loss of optimal operation performance
Solution Approach 1:
The system uses feedback from the MPC controller to guide the perturbation process. The controller continuously monitors process variables and adjusts perturbation magnitude and direction to maximize information gain for model identification while keeping process variables within acceptable ranges, thus maintaining near-optimal operation performance during calibration
Data Source
AI summary
A controller has improved closed-loop step testing of a dynamic process of an industrial processing plant. The controller performs economic optimization relaxation on process variables, such that operating range of the variables (MVs and CVs) during the testing are not skewed by variations in optimization cost factors. The controller employs computer-implemented methods and systems that receive a user-defined giveaway tolerance representing an allowable range between a current process variable value and a target process variable value. In response to the variables not meeting the giveaway tolerance, embodiments adjust the MPC controller configuration to drive the variables inside the tolerance, while relaxing optimization of the variables already meeting the giveaway tolerance. Using the adjusted configuration, embodiments calculate a new set of targets and generate a dynamic move plan from the new target. Embodiments add perturbation signals for the testing to the move plan in accordance with the adjusted configuration.