Dynamic Model Switching for Process Control Adaptability
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing power plant control systems face challenges in adapting to dynamic processes and disparate data sources, requiring flexible models that can adjust to changing conditions and mix different model types for accurate optimization, while traditional methods are often limited by data sparsity and static model configurations.
Innovation Solution
The introduction of compound, hybrid, and directional change correlation models that allow for the mixing of different model types and switching based on optimization criteria, enabling zooming in or out of fidelity and combining various models for enhanced predictive capabilities and optimization strategies.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If traditional static models are used for process control, then model simplicity is maintained, but adaptability to changing conditions deteriorates
Solution Approach 1:
The patent implements dynamic model selection and switching mechanisms that allow the control system to adapt to changing process conditions in real-time. Multiple models are maintained with different levels of fidelity, and the system dynamically selects or combines models based on current operating conditions, data availability, and performance requirements, resolving the contradiction between adaptability and complexity.
Solution Approach 2:
The patent segments the modeling problem into multiple discrete models with different levels of fidelity and complexity. Instead of using a single static model, the system divides the control space into multiple regions or operating conditions, each with its own optimized model, allowing the system to maintain simplicity where possible while providing adaptability where needed.
2Measurement precision
If multiple model types are combined for higher accuracy, then predictive power is improved, but device complexity increases
Solution Approach 1:
The patent merges multiple model types (first-principles models, neural networks, regression models) into a unified hybrid modeling framework. These models are combined in ways that leverage the strengths of each type while managing complexity through systematic integration strategies, allowing the system to achieve higher predictive accuracy without proportionally increasing overall system complexity.
Solution Approach 2:
The patent applies different model types and levels of complexity to different local regions or operating conditions. Instead of using a single complex model everywhere, the system applies simplified models where sufficient and first-principles models where physical understanding is critical, with neural networks filling gaps where data is available but physical relationships are unknown, thus optimizing the balance between accuracy and complexity locally.
3Measurement precision
If model fidelity is increased for better accuracy, then predictive power improves, but computational requirements and complexity increase
Solution Approach 1:
The patent changes the parameters of model fidelity and complexity based on operating conditions. The system dynamically adjusts which models are active and at what level of detail, using lower-fidelity models during normal operation and switching to higher-fidelity models only when needed for critical decisions or unusual conditions, thus managing computational power requirements while maintaining necessary accuracy.
4Quantity of substance
If traditional testing methods are used to gather training data, then data quality can be controlled, but time and cost increase significantly
Solution Approach 1:
The patent enables the modeling system to self-train by automatically utilizing process data as it becomes available from normal plant operations. Instead of requiring dedicated testing periods or manual data collection campaigns, the system continuously learns from operational data, automatically identifying training opportunities and updating models without disrupting plant operations, thus acquiring training data without significant time loss.
Data Source
AI summary
The present invention presents two new model types and a new method for evaluating a model used in the control application. These include a compound model, a hybrid model and a directional change coefficient model. The present invention allows the mixing of models with different inputs and outputs and the switching between these models based criteria for measuring optimization accuracy. The present invention allows switching between these models. The compound model is a model type that allows zooming in on the process to model parts of the data space with higher fidelity or resolution without loosing the capability to model the complete data space. The modeler does not loose any functionally over a regular neural network, but instead gains the ability to define the conditions when the model should use network weights best matched to the defined local conditions. The hybrid model is an extended version of a compound model. A hybrid model allows the combining of one or more models into a single model for purposes of interrogation or optimization. Within the hybrid model may reside a compound model itself. The directional change model (DCC) allows better evaluation of the predictive capability of Compound Models. It may also be used with any other model type.