Making Buildings Smarter: How AI Can Reduce Energy Waste

In the United States buildings account for around 40% of total energy consumption and 75% of electricity use. Heating, ventilation and air conditioning (HVAC) systems alone account for more than half of a commercial building’s energy consumption. 

HVAC systems have relied on static building management systems and manual adjustments that use fixed setpoint, schedules and periodic readjustments. But these inefficient controls have led to a large amount of energy waste. With cities decarbonizing and as renewable energy gets more integrated into day to day life, buildings need to become more intelligent in how they consume energy. 

These concepts are explored in a new article published in the Institute of Electrical and Electronic Engineers, co-authored by You Lin, Postdoctoral at MIT in Machine Learning Applications and Energy Systems; Leslie Norford, Professor in the Department of Architecture at MIT; Jeremy Gregory, Executive Director of the MIT Climate and Sustainability Consortium; and Audun Botterud, Principal Research Scientist in the MIT Laboratory for Information and Decision Systems. They discuss how AI-based automation systems can manage HVAC systems to adjust energy use based on occupancy, weather and energy prices to reduce waste, lower cost and create new revenue through a new service. 

In the article the researchers highlight an AI-based system they developed and piloted in a few buildings on the MIT campus in collaboration with the MIT Department of Facilities. The system uses graph learning and reinforcement learning to capture room occupancy, local weather and thermal interactions across zones. The pilot projects have shown significant energy savings by adjusting heating and cooling set points. 

These pilot implementations in MIT buildings delivered about 40% average energy savings with day to day results ranging from about 20% to 70% depending on weather conditions like the outside temperature or amount of sunshine. This AI solution not only reduces energy consumption, but also saves money. The energy savings roughly translate to $3,000 in monthly savings for a campus building of approximately 45,000 SF when compared to the original settings. 

“AI gives us the ability to make these systems much more responsive, learning from occupancy, weather and energy prices to continuously adjust how a building uses energy. Our work shows that this can translate into substantial energy savings while maintaining occupant comfort,” said lead author You Lin. “This enables buildings to intelligently manage how and when they use energy.”

This work demonstrates how commercial buildings can contribute to decarbonization efforts by linking academic innovations with deployable building management system-compatible software. It paves the way for AI-driven building controls and for these systems to work closely with the electric grid. As these kinds of systems continue to evolve, smarter building control management systems could not only help with efficiency but also be flexible, resilient and active participants in a clean energy future. 

The researcher’s future work will focus on scaling the multi-space HVAC optimization framework to a broader portfolio of commercial buildings, integrating wireless, interactive, and personalized occupant-comfort technologies, validating participation in real electricity markets to generate new revenue streams, and ultimately developing an industry-scale platform that unifies building control, occupant interaction, flexibility aggregation, and market participation.

Read more here