MPC Step Testing with Optimization Relaxation for Model Accuracy
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 of closed-loop step testing often leading to process variable drift and suboptimal operation.
Innovation Solution
A computer-implemented method and system that performs non-invasive closed-loop step testing by adjusting the MPC controller's configuration to maintain process variables within defined tolerances, relaxing economic optimization to minimize perturbation impact, allowing for accurate data collection while maintaining optimal operating ranges, thereby reducing the need for re-tuning and run-time intervention.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If conventional closed-loop step testing is performed with MPC controller actively optimizing economic objectives, then model identification data can be collected, but process variables drift away from optimal targets causing suboptimal operation and excessive feedback correlation
Solution Approach 1:
The system dynamically adjusts the MPC controller configuration between two states: during normal operation, the controller actively optimizes economic objectives with tight constraints; during step testing, the controller configuration is relaxed to allow larger perturbations while maintaining safety constraints. This dynamic switching resolves the contradiction by adapting the control strategy to the current operational mode.
Solution Approach 2:
The system implements periodic step testing intervals where the MPC controller temporarily relaxes economic optimization to collect identification data, then returns to normal optimization mode. This periodic switching between testing and operation modes allows both model accuracy improvement and maintained process optimality during respective phases.
2Productivity
If MPC controller actively optimizes economic objectives during step testing, then optimal operation is maintained, but process perturbation is insufficient for accurate model identification
Solution Approach 1:
During step testing, the system applies partial relaxation of the economic optimization objective function, allowing process variables to deviate more than normally permitted. This excessive perturbation action generates sufficient excitation for accurate model identification while still operating within safety constraints, resolving the contradiction between optimality and identification accuracy.
3Measurement precision
If frequent model recalibration is performed to maintain MPC performance, then model accuracy is improved, but operational interruptions increase
Solution Approach 1:
The system performs self-calibration during normal operation by continuously collecting process data and automatically updating model parameters without requiring external intervention or process shutdown. This self-service approach maintains high model accuracy while eliminating operational interruptions associated with manual recalibration.
Solution Approach 2:
The system implements continuous model updating during normal operation rather than periodic offline recalibration. This continuous action maintains model accuracy throughout operation without interrupting production, resolving the contradiction between accuracy and operational continuity.
4Measurement precision
If process variables are allowed to drift during step testing to maximize perturbation effect, then model identification data quality improves, but process operation performance degrades
Solution Approach 1:
The system applies different quality standards to different process variables during step testing: safety-critical variables maintain tight constraints to protect operation performance, while non-critical variables allow larger deviations to improve identification data quality. This localized differentiation resolves the contradiction by optimizing each variable's perturbation level according to its importance.
Data Source
Figure 1A
Figure 1B
Figure 1C
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, the methods and systems adjust the MFC controller configuration to drive the variables inside the tolerance, while relaxing optimization of the variables already meeting the giveaway tolerance. Using the adjusted configuration, the methods and systems calculate a new set of targets and generate a dynamic move plan from the new target. The methods and systems add perturbation signals for the testing to the move plan in accordance with the adjusted configuration.