(J-373) Wearable AI-enabled thermal comfort sensor for human-in-the-loop control of the HVAC system for energy-efficient building

Introduction:: Energy consumption is a serious problem since it is generating over 60% of greenhouse gas emissions. Building energy consumption occupies over one-third of the total energy demand, with Heating, Ventilation and Air Conditioning (HVAC) systems accounting for half of total energy consumption in buildings. However, current HVAC systems are operated by a set-point control which only considers ambient temperature, leading to suboptimal energy usage and reduced thermal comfort for occupants.

Recent advances in artificial intelligence (AI) have led to the development of thermal comfort prediction technologies, but, there has been no validation of these technologies with closed-loop HVAC control system as well as they only utilized commercial wrist band of single skin temperature measurement which limits its thermal comfort prediction accuracy.

Here, we developed an AI-Human-In-the-Loop (HIL)-HVAC control system based on AI-enabled wearable temperature sensor that operates in real-time which adapts to thermophysiological conditions to improve occupant’s thermal comfort and reduce power consumption. The system employs a soft wearable sensor that measures skin temperature at three points on the wrist, providing a more accurate prediction of thermal comfort than conventional single-point measurements. The system also employs wireless communication with AI-HIL-HVAC control system to enable real-time temperature detection and control of indoor temperature.

Demonstration of the system showed its ability to optimize HIL-HVAC control based on the thermal comfort status of occupant and reduce 20% of power consumption compared to conventional set-point HVAC control system. This work represents a potential solution towards sustainable building energy management and improved occupant comfort.




Materials and Methods:: 1. Construction of AI-HIL-HVAC control system based on AI-enabled wearable temperature sensor

The AI-HIL-HVAC control system is composed of a flexible wearable wrist skin temperature sensor, external ambient temperature and humidity sensors, an AI model, and HVAC system. The wearable sensor comprises three thermistors, regulators, wireless charging coils, a BLE SoC and a lithium polymer battery, which can wirelessly transmit the three skin temperatures measured by thermistors on upper wrist, ulnar artery and radial artery to the external cloud server which is connected to the AI HVAC control system using BLE communication, while the external sensors transfers the ambient thermal conditions. Additionally, information on sex, age and clothing insulation are gathered via a straightforward user survey through a smartphone app. The AI model for thermal comfort prediction was developed by field experiments collecting described environmental and skin physiological conditions of occupant and various machine learning algorithms are compared using five-fold cross validation.



2. Evaluation of AI-HIL-HVAC control system based on AI-enabled wearable temperature sensor

Mechanical robustness and electrical reliability of the wearable temperature sensor is both verified for long-term operation in daily life using a finite element method (FEM) and mechanical distortion experiments including bending, stretching and twisting conditions. Also, statistical analysis of thermal comfort prediction with various machine learning models as well as the demonstration for the HIL-HVAC control validating our concept was carried out, measuring both the HVAC power consumption and individual heart rate variability using an ECG monitored with a wireless heart rate recording device.


Results, Conclusions, and Discussions:: Our Soft wearable sensor ensures consistent electrical performance for accurate skin temperature measurement offering a resolution of 0.05°C, even when it undergoes mechanical deformation of bending, stretching and twisting which verifies that our device can be practicality used in daily life. Utilizing this AI-enabled wearable temperature sensor, the model's accuracy increased up to 93.9% with Area Under the Curve (AUC) values of receiver operating characteristic (ROC) and precision recall (PR) curves over 0.9 for each class which is comparable to the best accuracy of state-of-the-art thermal comfort prediction models. Also, the statistical analysis proved that our three-point skin temperature measurements can increase accuracy by 2.9% from a single-point measurement without sex, age and clothing insulation features and by 11.2% without external sensor information, indicating that the multi-sensing implementation can enhance the performance of thermal comfort prediction for more efficient HVAC system operation compared to traditional methods that uses commercial wristbands with single-point skin temperature measurement. The result ensures that our AI wearable sensor-based thermal comfort prediction model could be utilized in HVAC control systems with precise thermal comfort prediction.

From the actual field experiments of comparing conventional set-point-HVAC control and our HIL-HVAC control system, the HIL-HVAC control decreased the total power consumption by 20% compared to the set-point HVAC control (131 kWh power consumption for set-point-HVAC control vs. 105 kWh for HIL-HVAC control for 6-hour operation) as well as the average HRV value decreased in range of thermal comfort condition for occupant compared to the previously collected HRV values in various thermal condition. These results verify that using our AI-HIL-HVAC control system, the occupant’s thermal comfort is improved even with efficient HVAC system operation.

To summarize, the suggested AI-HIL-HVAC control system based on AI-enabled wearable temperature sensor can improve thermal comfort prediction accuracy and reduces the power consumption of HVAC control systems by utilizing novel AI-enabled wearable three-point wrist skin temperature sensors which ensures long-term reliable operation in daily life. These results demonstrate the potential use of the system in buildings for energy-efficient and occupant-comfortable HVAC operation.


Acknowledgements (Optional): :

References (Optional): :