Building control system with adaptive online system identification
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing building control systems face challenges in accurately modeling and predicting system dynamics, leading to suboptimal performance in maintaining occupant comfort and optimizing energy usage due to the complexity and non-linearity of building systems, and the need for frequent model updates to account for changes in disturbances and physical properties.
Innovation Solution
The implementation of an adaptive online system identification method that utilizes a combination of light and intensive online system identification frameworks, which leverage old models with new data to generate accurate predictive models, focusing on heat disturbances and Kalman gain adjustments, allowing for frequent updates without constant re-identification of all models, and using model predictive control to optimize HVAC operations.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If frequent model updates are performed to account for changes in disturbances and physical properties, then the accuracy of predictive modeling is improved, but the computational cost and system complexity increase
Solution Approach 1:
The system segments the model update process into two distinct approaches: light online system identification for routine updates and intensive online system identification for comprehensive recalibration. This segmentation allows the system to perform frequent lightweight updates without the full computational burden of intensive identification, thereby improving prediction accuracy while managing system complexity.
Solution Approach 2:
The system dynamically selects between light and intensive online system identification based on operational conditions and performance requirements. This dynamic approach enables the system to adapt its computational intensity to current needs, performing frequent updates when necessary while avoiding unnecessary computational overhead during stable periods, thus resolving the contradiction between accuracy and complexity.
2Productivity
If light online system identification is used for frequent updates, then computational cost is reduced, but the ability to capture complex system dynamics may be limited
Solution Approach 1:
The system dynamically transitions between light and intensive online system identification modes based on detected system conditions. When complex dynamics are detected or performance degradation is observed, the system switches to intensive identification to ensure accurate modeling, while using light identification during normal operation to maintain computational efficiency.
Solution Approach 2:
The system incorporates feedback mechanisms that monitor prediction accuracy and system behavior. When feedback indicates that light identification is insufficient for capturing current system dynamics, the system triggers intensive identification to recalibrate the model, ensuring reliability while maintaining overall computational efficiency through selective use of intensive processing.
3Measurement precision
If intensive online system identification is performed, then comprehensive model recalibration is achieved, but the computational resources and time required increase significantly
Solution Approach 1:
The system segments model updates into light and intensive modes, performing comprehensive intensive identification only when necessary rather than continuously. This segmentation allows the system to maintain accurate models through frequent light updates while reserving intensive recalibration for situations where it is truly needed, thereby reducing overall time loss while maintaining recalibration accuracy when required.
Solution Approach 2:
The system implements periodic intensive online system identification at scheduled intervals or triggered by specific conditions, rather than performing it continuously. This periodic approach ensures comprehensive model recalibration occurs regularly enough to maintain accuracy while minimizing the total time and computational resources devoted to intensive identification, thus resolving the contradiction between recalibration accuracy and time loss.
Data Source
AI summary
A controller for equipment that operates to affect a variable state or condition of a building including one or more processors and non-transitory computer-readable media storing instructions that, when executed by the processors, cause the processors to perform operations. The operations include generating a new predictive model using training data associated with one or more durations selected from a time period and selected to satisfy a set of criteria. The predictive model models system dynamics of the building during the time period. The operations include storing the new predictive model in a database including predictive models that model the system dynamics of the building and include comparing performance of the new predictive model and the predictive models stored by the database to select a particular predictive model for controlling the equipment. The operations include using the particular predictive model to generate and provide control signals to the equipment.