Building HVAC Model Training with Automatic Data Selection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing building control systems face challenges in generating accurate predictive models due to complex, nonlinear system dynamics, and current methods lack efficient automated processes for selecting suitable training data during normal operational conditions.
Innovation Solution
A method and system for automatically assessing and selecting training data by evaluating correlations between setpoints and measurements, load conditions, and setpoint change durations to ensure high-quality data for system identification, allowing for online system identification and improved model accuracy.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If system identification is performed using data collected during normal operational conditions, then the system can operate continuously without disruptive experiments, but the quality and suitability of training data for generating accurate predictive models becomes difficult to ensure
Solution Approach 1:
The system performs preliminary assessments of data quality characteristics (correlation, load, setpoint changes) before using the data for training. By evaluating these characteristics in advance and selecting only suitable data segments, the system ensures high-quality training data while collecting data during normal operations, thus resolving the contradiction between continuous operation and model accuracy.
2Manufacturing precision
If multiple assessments of data characteristics are performed to ensure high-quality training data, then model accuracy improves, but the complexity of the data selection process increases
Solution Approach 1:
The data selection process is segmented into multiple independent assessments, each evaluating a specific characteristic (correlation between setpoint and measurement, load on equipment, durations between setpoint changes). By dividing the complex evaluation into separate modular assessments, the system achieves comprehensive data quality checking while maintaining manageable process complexity.
3Manufacturing precision
If training data is carefully selected based on multiple criteria, then the accuracy of predictive models improves, but the time required for data processing and model training increases
Solution Approach 1:
The system extracts and evaluates only the essential characteristics needed for data quality assessment (correlation, load, setpoint change durations) rather than analyzing all possible data attributes. This selective extraction of critical features enables efficient data screening that ensures model accuracy while minimizing processing time.
Data Source
AI summary
A method includes operating equipment in accordance with a setpoint to affect a measurement for a space during a training period, performing a plurality of assessments of different characteristics of data for a segment of the training period, and training a system model using a set of training data. The data for the segment is included in the set of training data in response to passing the plurality of assessments or excluded from the set of training data in response to failing one or more of the plurality of assessments. The method also includes controlling the equipment using the system model.