Regression-Optimization Predictive Control for Nonlinear Process Constraints
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing model predictive control methods struggle with process dynamics and non-linearity in manufacturing and processing industries, requiring complex constraints that are not effectively addressed by approximate dynamic programming, Markov decision processes, and reinforcement learning.
Innovation Solution
A computer-implemented method using multiple step look-ahead regression models with tuned look-back windows, generating optimization constraints and variables, and solving an optimization model to produce actions through a rolling horizon procedure.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If model predictive control uses first principles system models with local linearization, then it can handle process dynamics, but it requires complex constraints and simplifications that reduce accuracy
Solution Approach 1:
The patent replaces traditional first principles system models with data-driven regression models trained on historical process data. This substitution eliminates the need for complex analytical models and local linearization, while capturing non-linear process dynamics through learned relationships from data.
Solution Approach 2:
The patent creates simplified copies of complex process dynamics through regression models that replicate system behavior without requiring underlying physical principles. These models capture essential dynamics while being computationally tractable for real-time control.
2Extent of automation
If approximate dynamic programming, Markov decision process, or reinforcement learning techniques are used, then automation is improved, but they have limited ability to consume complex constraints
Solution Approach 1:
The patent segments the control problem into two parts: data-driven regression models capture process dynamics while separate optimization constraints explicitly handle complex operational constraints. This segmentation allows automated control to effectively consume and satisfy multiple constraints simultaneously.
Solution Approach 2:
The patent creates a universal optimization framework that can handle diverse constraint types (equality, inequality, integer, binary) within a single mathematical programming formulation, making the automated control system versatile across different process constraints.
3Ease of manufacture
If traditional control methods are used for non-linear processes, then implementation is simpler, but they cannot effectively address process dynamics and non-linearity
Solution Approach 1:
The patent substitutes traditional control theory approaches with data-driven regression models that automatically adapt to non-linear process dynamics. This substitution maintains implementation simplicity through standard machine learning workflows while dramatically improving control performance for non-linear processes.
Data Source
AI summary
A computer-implemented method, computer program product, and computer system for automated model predictive control. The computer system trains multiple step look-ahead regression models, using historical states and historical actions for a to-be-optimized system, for each timestep of a past time horizon. Regression models may be either linear or nonlinear in order to capture process dynamics and nonlinearity. The computer system generates optimization constraints for each timestep of a future time horizon. The computer system generates optimization variables, based on the multiple step look-ahead regression models, for each timestep of the future time horizon. The computer system constructs a mixed integer linear programming based optimization model that includes an objective function, the optimization constraints, and the optimization variables. Nonlinear regression models are converted into piecewise linear approximation functions. The computer system solves the optimization model to produce actions for the to-be-optimized system, over the future time horizon, and recommend commitment-look-ahead actions.