Hybrid Scalable Building Models

Problem Context

By 2050, the entire built environment must be energy neutral. In addition, by 2050 all homes and utility buildings must be "off the gas grid". Although renewable energy sources can help achieve this goal, the energy transition leads to peak loads in terms of energy demand. A solution is needed to combat these high energy demands.

Solution

Building models play a central role in optimizing energy consumption at building level, and (more importantly) in balancing energy supply and demand at a district level. AI based building models are therefore essential in facilitating the energy transition and the transition to an energy-neutral built environment. Unfortunately, to work properly, AI based models will require energy usage data. This data is considered to be private. In order to learn from the energy usage data of homeowners whilst still protecting their privacy, however, it is possible to use federated learning.

The main research objective of this project is to develop a building model which combines a physics knowledge model with a machine learning model. This model will be suitable for predicting energy demand and scalable for all types of building types expected in a smart district.