Chilled Water Plant Modeling With Bayesian Updates Instead of Retraining
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Chilled water plants face barriers to the commercial application of predictive methods due to high costs, skill availability, equipment variability, limited training data, and ongoing costs for model re-training, which hinder efficient operation and maintenance.
Innovation Solution
A method using online Bayesian linear regression with sub-models for each major component, continuously adapting to system changes without re-training, informed by physics and engineering principles, and employing Bayesian inference to estimate unknown coefficients.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If machine learning methods are deployed for predictive modeling of chilled water plants, then performance and efficiency can be improved, but high costs and complexity of model acquisition and deployment arise
Solution Approach 1:
The patent segments the chilled water plant into multiple independent sub-models, each representing a specific component or subsystem. This segmentation allows each sub-model to be developed, trained, and maintained independently, reducing the overall complexity while maintaining predictive accuracy across the entire plant system.
Solution Approach 2:
The patent develops a universal framework that can be applied across different chilled water plant configurations. The methodology uses standardized modeling approaches and can adapt to various plant types, reducing the need for custom-developed models for each specific plant and thereby reducing costs and complexity.
2Measurement precision
If custom predictive models are developed for each plant, then accuracy improves, but the cost and time required for model development increase
Solution Approach 1:
The patent performs preliminary actions by establishing a standardized framework and template models that can be quickly deployed. Rather than developing models from scratch for each plant, the framework provides pre-configured structures that can be rapidly adapted to specific plant configurations, significantly reducing development time while maintaining accuracy.
Solution Approach 2:
The patent utilizes parameter changes to adapt the universal framework to specific plants. By adjusting key parameters and coefficients based on plant-specific data, the models achieve high accuracy without requiring complete custom development, thus reducing both time and cost.
3Reliability
If predictive models are re-trained to reflect plant changes, then accuracy is maintained, but ongoing costs and complexity increase
Solution Approach 1:
The patent implements feedback mechanisms that continuously monitor plant performance and automatically trigger model updates when changes are detected. This feedback loop ensures model accuracy is maintained while minimizing unnecessary re-training operations, thereby reducing ongoing costs and complexity while maintaining high productivity.
Data Source
AI summary
Predictive methods such as machine learning methods based on neural network technology require large sets of historical training data that are often not available or may not represent the required range of operating scenarios, seasons etc. Accordingly, deployment of machine learning may be impeded or delayed while suitable training data is collected. Further, the characteristics of the plant and its component may change over time due to wear and tear, maintenance events, equipment replacement or upgrades such that the predictive models must be updated (re-trained). Determining an effective schedule for such re-training and the re-training process introduces additional costs, as well as the risk that seasonality and other variables may not be properly captured in the process of re-training these models. Accordingly, it would be beneficial to provide methods and system that mitigate these obstacles with respect to the commercial application of machine-learning and other predictive methods.