Self-Training Thermodynamic Model for Building System Commissioning
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing building automation systems lack a unified operational model for managing energy interrelationships between various building components, struggle with model-free control loops that are difficult to manage and optimize in complex systems, and require manual, time-consuming commissioning processes that are limited to known models and occupancy-free training periods.
Innovation Solution
A system for in-situ control model training using a thermodynamic model with neurons and machine learning to adaptively tune parameters based on sensor data, allowing self-commissioning and optimizing building systems without requiring offline testing or manual intervention.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of manufacture
If model-free control loops are used to manage building systems, then implementation is simple, but the system becomes difficult to manage and optimize as complexity increases
Solution Approach 1:
The patent creates a digital twin (virtual copy) of the building system that mirrors the physical system's behavior. This digital model can be trained and optimized independently, then applied to control the actual building system. The copying approach allows complex optimization to occur in the virtual model without directly complicating the physical system's operation.
Solution Approach 2:
The trained thermodynamic model serves as an intermediary between sensor data and control decisions. Instead of directly controlling the building system based on raw sensor data, the system uses the trained model to interpret sensor inputs and generate control commands, simplifying the management complexity while maintaining optimization capability.
2Reliability
If automated commissioning is performed during occupancy-free training periods with artificial test regimes, then training can be conducted, but retro-commissioning and continuous commissioning are limited
Solution Approach 1:
The patent enables the commissioning system to adapt dynamically to different operational states. The thermodynamic model can be trained during occupancy-free periods using artificial test regimes, then continuously refined during occupied periods using actual building data. This dynamic approach allows the system to transition between different commissioning modes based on building occupancy and operational conditions.
Solution Approach 2:
The system maintains continuous commissioning capability by seamlessly transitioning between offline training modes and online refinement modes. The thermodynamic model undergoes initial comprehensive training during occupancy-free periods, then continues to adapt and improve continuously during building operation, ensuring uninterrupted optimization capability.
3Reliability
If manual commissioning through user analysis of building automation system data is performed, then commissioning can be conducted, but it requires a knowledgeable person and is very time-consuming
Solution Approach 1:
The system performs self-commissioning through automated training of the thermodynamic model. The model automatically learns from sensor data and building operations without requiring manual intervention or expert analysis. This self-service approach eliminates the need for knowledgeable commissioning personnel while dramatically reducing commissioning time.
Solution Approach 2:
The patent replaces manual expert analysis with automated machine learning algorithms. Instead of relying on human experts to analyze building automation system data, the system uses trained thermodynamic models and neural networks to automatically interpret data, identify patterns, and optimize control parameters, substituting mechanical human effort with automated computational processes.
4Ease of manufacture
If digital model based approaches are used for known models defined a-priori, then modeling can be performed, but the approach is limited in scope and cannot scale to complex ad hoc arrangements
Solution Approach 1:
The patent segments the building system into multiple thermodynamic zones and components, each represented by neural network units. This segmentation allows the model to handle complex ad hoc arrangements by breaking down intricate building topologies into manageable segments that can be independently modeled and then integrated, enabling scalability to arbitrary building configurations.
Solution Approach 2:
The system uses parameter-based neural network models that can adapt to different building configurations by changing parameters rather than requiring complete remodelling. The thermodynamic model adjusts its parameters based on the specific building topology, allowing it to scale from simple to complex ad hoc arrangements while maintaining modeling capability.
Data Source
AI summary
Using processes and methods described herein, a digital twin of a physical space can train itself using sensors and other information available from the building. In some embodiments, a system to be controlled comprises a controller that is connected to sensors. This controller also has a thermodynamic model of the system to be controlled within memory associated with the controller. The thermodynamic model has neurons that represent distinct pieces of a controlled space, such as a piece of equipment or a thermodynamically coherent section of a building, such as a window. The neurons represent these distinct pieces of the controlled space using parameter values and equations that model physical behavior of state with reference to the distinct piece of the controlled state. A machine learning process refines the thermodynamic model by modifying the parameter values of the neurons, using sensor data gathered from the system to be controlled as ground truth to be matched by behavior of the thermodynamic model. The thermodynamic model may be warmed up by running the model using state data as input.