Open Modeling Architecture for Model Predictive Control
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Model predictive control systems often simplify process models for computational efficiency, compromising model quality and controller performance, and may not secure proprietary information from entities responsible for the process models.
Innovation Solution
An open modeling architecture decouples the execution frequencies of optimization engines and model modules, allowing model owners to maintain their models in appropriate languages without revealing implementation details, using unified access modules to adapt and interface outputs for the optimization engine.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If the MPC system significantly simplifies the process model for computational efficiency, then the computational efficiency is improved, but the model quality and controller performance deteriorate
Solution Approach 1:
The system segments the modeling and control functions into separate modules: model modules that maintain high-fidelity process models and an optimization module that handles real-time control decisions. This segmentation allows each module to operate at its optimal complexity level without compromising the other.
Solution Approach 2:
The patent introduces an intermediary layer (the optimization module) that bridges the high-fidelity model modules and the control system. This intermediary handles the computational burden of real-time optimization while allowing the model modules to maintain their detailed representations without being simplified.
2Productivity
If the MPC system accesses and maintains the process models, then the controller can optimize performance, but the proprietary information security deteriorates
Solution Approach 1:
The patent extracts the process models from the MPC controller and places them in separate, independently secured model modules. This extraction allows the controller to access model outputs without having direct access to or custody of the proprietary model implementations, thereby securing sensitive information.
Solution Approach 2:
The optimization module serves as an intermediary that interfaces with model modules without requiring access to their internal implementations. This intermediary layer enables the controller to benefit from accurate models while maintaining security boundaries that protect proprietary information.
3Ease of operation
If the system uses a unified interface for all model modules, then the system integration is simplified, but the ability to maintain models in different languages and formats deteriorates
Solution Approach 1:
The patent introduces unified access modules as intermediaries that translate between diverse model module interfaces and the optimization module's requirements. This intermediary layer maintains a standardized interface for integration while allowing model modules to be implemented in different languages and formats without compromising system coherence.
Solution Approach 2:
The system allows each model module to have its own specific interface characteristics and implementation details (local quality) while the unified access modules provide the necessary adaptation. This enables each component to operate in its optimal form while maintaining overall system integration.
Data Source
AI summary
In one embodiment, a model predictive control system for an industrial process includes a processor to execute an optimization module to determine manipulated variables for the process over a control horizon based on simulations performed using an objective function with an optimized process model and to control the process using the manipulated variables, to execute model modules including mathematical representations of a response or parameters of the process. The implementation details of the model modules are hidden from and inaccessible to the optimization module. The processor executes unified access modules (UAM). A first UAM interfaces between a first subset of the model modules and the optimization module and adapts output of the first subset for the optimization module, and a second UAM interfaces between a second subset of the model modules and the first subset and adapts output of the second subset for the first subset.