Wayne State University
Name of the energy conservation measure (ECM)
AI-enabled HVAC Control
Purpose of ECM:
The Artificial Intelligence (AI)-enabled Control for heating, ventilation, and air condition (HVAC) system is a new measure. The team’s proposed Energy Conservation Measure (ECM) synchronizes HVAC control with the occupancy pattern of the building’s zones.
According to the U.S. Energy Information Administrative (EIA) space heating accounted for 38% of delivered energy in buildings. Also, in 2020, electricity use for cooling the interior of buildings by the U.S. residential and commercial sectors was about 392 billion kWh, which accounts for about 10% of total U.S. electricity consumption in 2020.
In addition, commercial buildings will expand by 34% on a square footage basis. Building Management Systems (BMS) have been used extensively to control on/off HVAC equipment time schedules based on outdoor temperatures and other variables. But the existing BMS does not capture the real-time building occupancy and the dynamic behavior of occupants. The team’s proposal is a significant opportunity to increase building energy efficiency by adjusting HVAC control with the actual occupancy patterns of the building.
Description of ECM:
The Internet of Things (IoT) sensors and actuators installed in modern buildings provide a wealth of data about human activities and movements. AI can turn the data collected from IoT sensors into valuable information, which can be used for pro-active control of the HVAC system.
For instance, when a zone is not occupied, the HVAC set point can be relaxed by 2-3° F to save energy. Also, predicting the occupancy of a zone can provide opportunities for pre-heating or precooling to avoid large thermal peak loads, while ensuring occupants’ comfort.
In this project, the team developed new AI algorithms to learn occupants’ behavior and predict their activity and patterns based on the historical data collected from IoT sensors as well as social and environmental determinant data from publicly available sources, such as Points of Interest and social media check-ins. Then, they characterized zonal activity levels in the building, and control the HVAC system, both spatially and temporally, to optimize energy efficiency without compromising occupants’ comfort. The learning starts at the sensor level (edge) to optimally control the HVAC system in real-time or near real-time.
Long-term learning will be performed at the Energy Management System (EMS) to predict zonal occupancy patterns and implement predictive HVAC control to further optimize energy efficiency.