Thermal-Adaptation-Behavior-Based Thermal Sensation Evaluation Model with Surveillance Cameras

Area of Science:

• Building energy efficiency
• Artificial intelligence in smart buildings
• Human-computer interaction

Background:

• The construction sector consumes nearly 30% of global energy, with HVAC systems being major energy users.
• Manual control of HVAC systems limits energy efficiency and occupant comfort.
• Real-time adjustments based on occupant thermal comfort can significantly improve building energy performance.

Purpose of the Study:

• To propose a non-intrusive model for dynamic inference of occupant thermal comfort using indoor surveillance camera data.
• To enhance energy efficiency and enable dynamic decision-making in intelligent building energy management.

Main Methods:

• Development of a two-stream transformer-augmented adaptive graph convolutional network.
• Identification of people's heat-related adaptive behaviors from video data.
• Training and testing on a dataset of 16 distinct temperature adaptation behaviors.

Main Results:

• The proposed model achieved a behavior recognition accuracy of 96.56%.
• The transformer augmentation significantly improved the accuracy of detecting thermal adaptation behaviors.
• Demonstrated the potential for substantial energy savings and emission reductions.

Conclusions:

• The developed model offers a viable solution for real-time, non-intrusive monitoring of occupant thermal comfort.
• Enables intelligent buildings to dynamically adjust HVAC systems for optimal energy efficiency and occupant well-being.
• Paves the way for advanced energy management strategies in smart buildings.