AI Special: ‘Learning’ building controls lift energy efficiency
Arup developed a method for improving HVAC controls at the V&A museum
The most advanced area of the built environment sector in terms of adoption of AI is the software involved in controlling the performance of a building, particularly heating and lighting to boost energy performance.
The Nest intelligent thermostat is well established in the domestic market, but similar controls that learn how buildings behave and then adjust conditions accordingly are not widespread outside the home.
One example at the cutting edge of this work is at the Victoria & Albert Museum (pictured above), where Arup has worked for a number of years.
AI Special: Related articles
Arup’s associate director Carolina Bartram explains the process: “We developed a method for improving existing HVAC controls, using what is known as model predictive control (MPC) and machine learning techniques, to reduce energy consumption and improve internal environmental conditions in the closely controlled gallery environment.
“For the European Galleries project – the first on which we which we have implemented the MPC project – we gathered one year’s worth of data (60+ variables at one-minute intervals) on which to base the predictive control models. Rather than being reactive to current conditions in the space, the controls look at how we know the building behaves, based on data, and optimises the systems (air intake, temperature and humidity of air etc) to control the future conditions.”
Although this project specifically focuses on galleries and maintaining environmental conditions for artefacts, the approach is transferable, she says.
“This approach will also support Arup’s (and the industry’s) goal to deliver better buildings in operation, help close the industrywide recognised ‘performance gap’, and also ensure the learning feeds back into our future designs,” she says.