Improving Control, Energy Efficiency, and Comfort of Digital Twin Building Environments Using Machine Learning Transformers

Digital Twins, virtual representations of physical systems, are extremely useful for gaining valuable insights and making predictions by modeling the environment and sensor systems of a building. Digital twins allow us to maximize productivity by making better control decisions and reducing our carbon footprint. However, it is difficult to model a physical building precisely in a digital twin. We propose using machine learning to optimize and control the building’s digital twin for improved energy efficiency while maintaining occupants’ recommended thermal comfort levels. Our work provides a framework of digital twin and machine learning to enhance the performance of HVAC and non-HVAC building digital twins, improving energy efficiency by approximately 8% in an HVAC building and approximately 20% in a non-HVAC building across an entire year, while also ensuring occupants comfort
is maintained and improved for both building types.