Abstract
Studies have shown that the actual energy consumption of buildings once built and in operation is often far greater than the energy consumption predictions made during design—leading to the term “performance gap.” An alternative to traditional, building physics based, prediction methods is an approach based on real-world data, where behavior is learned through observations. Display energy certificates are a source of observed building “behavior” in the United Kingdom, and machine learning, a subset of artificial intelligence, can predict global behavior in complex systems, such as buildings. In view of this, artificial neural networks, a machine learning technique, were trained to predict annual thermal (gas) and electrical energy use of building designs, based on a range of collected design and briefing parameters. As a demonstrative case, the research focused on school design in England. Mean absolute percentage errors of 22.9% and 22.5% for annual thermal and electrical energy use predictions, respectively, were achieved. This is an improvement of 9.1% for the prediction of annual thermal energy use and 24.5% for the prediction of annual electrical energy use when compared to sources evidencing the current performance gap.
Funding
This research has been supported by funding from the UK Engineering and Physical Sciences Research Council (EPSRC) via UCL's EngD Centre in Virtual Environments, Imaging and Visualisation (VEIV) with additional funding and support from AHR (formerly known as Aedas Architects).
Notes
1 The DEC dataset went through a data cleaning process to erase potentially erroneous data (Paterson Citation2017).