Dynamic Reduced-Order Model Calibration for Industrial RTO
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Industrial process models, such as those used in real-time optimization (RTO) systems, degrade over time due to changes in the underlying process, making it challenging to maintain accurate predictions and requiring extensive efforts to identify and update problematic sub-models, leading to suboptimal operations and profit loss.
Innovation Solution
A sustainable dynamic reduced-order model (SDROM) is constructed by dividing a trained ROM into MISO sub-models, applying dynamic filters, and periodically calibrating with historical and real-time data to monitor and adapt the model, eliminating the need for steady-state detection and enabling timely updates.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If traditional FPM/ROM with steady-state detection is used, then model prediction accuracy can be maintained under stable conditions, but decision-making delays occur and dynamic plant data cannot be utilized during transitions
Solution Approach 1:
The patent transitions from static steady-state detection to dynamic steady-state-free operation by continuously adapting model parameters online. The system uses dynamic parameter adaptation to maintain prediction accuracy during transitions without requiring the process to reach steady-state, enabling real-time utilization of dynamic plant data while eliminating decision-making delays.
Solution Approach 2:
The system continuously adapts model parameters online based on incoming dynamic data, changing parameters in real-time to maintain prediction accuracy during process transitions. This parameter adaptation mechanism allows the model to remain accurate without requiring steady-state conditions, thereby eliminating the time loss associated with waiting for steady-state detection.
2Reliability
If full-scale FPM is used for RTO, then comprehensive process modeling is achieved, but model degradation requires extensive engineering efforts and causes significant interruptions
Solution Approach 1:
The patent segments the full-scale process model into modular functional blocks that can be independently adapted and updated. This segmentation allows specific degraded portions to be identified and recalibrated without requiring complete model reconstruction, significantly reducing engineering efforts and avoiding plant interruptions while maintaining comprehensive process modeling capability.
Solution Approach 2:
The system continuously recalibrates and updates model parameters online, discarding degraded parameter values and recovering accurate values through continuous adaptation. This approach maintains comprehensive process modeling while enabling efficient model degradation recovery without extensive engineering interventions or plant interruptions.
3Reliability
If model is periodically audited and updated, then RTO performance can be sustained, but model degradation still causes loss of benefits between updates
Solution Approach 1:
The patent implements continuous online parameter adaptation that operates continuously between periodic audits, maintaining RTO performance sustainability without benefit loss. The continuous adaptation ensures the model remains accurate at all times, eliminating the performance degradation that occurs during the time intervals between periodic model updates.
Solution Approach 2:
The system incorporates continuous feedback mechanisms that monitor model performance in real-time and trigger parameter adaptations when degradation is detected. This feedback-driven approach sustains RTO performance by continuously correcting model deviations, eliminating the performance loss that would otherwise occur between periodic model audits and updates.
Data Source
AI summary
Embodiments of the present disclosure provide functionality for creating a sustainable dynamic reduced-order model (SDROM) for operating a real-world industrial process. The model is based upon a reduced order model (ROM) trained using data obtained from simulations performed using a first-principles model (FPM) of the real-world industrial process. The trained ROM is divided into multiple-input, single-output (MISO) sub-models, which are partitioned into component terms for incorporation of respective gain factors. The SDROM is deployed online to operate the real-world industrial process with one or more optimization objectives and the SDROM is periodically calibrated and validated using historical operation data.