AI identifies ‘hard to decarbonise’ houses in GB

Cambridge University researchers have developed an open data-based AI solution to identify hard to decarbonise houses.

The decarbonisation of buildings is a key component of the drive for net zero but many buildings may be hard to decarbonise for a variety of reasons, including age, structure and location among others.

To address this and with the aim to identify buildings that could be prioritised for retrofitting and other decarbonising measures, the researchers from Cambridge University’s Department of Architecture have developed a new ‘deep learning’ model that they say can identify such buildings with 90% precision.

And this rate is expected to increase as more and more data is added.

Have you read?
Push for safe incorporation of AI in energy sector
Mytilineos: Decision-making will define utilities of the future

“This is the first time that AI has been trained to identify hard to decarbonise buildings using open source data to achieve this,” says Dr Ronita Bardhan, who leads Cambridge’s Sustainable Design Group.

“Policymakers need to know how many houses they have to decarbonise, but they often lack the resources to perform detail audits on every house. Our model can direct them to high priority houses, saving them precious time and resources.”

Hard to decarbonise houses are estimated to be responsible for over a quarter of all direct housing emissions but are rarely identified or targeted for improvement.

The new AI model, which was published in the journal ‘Sustainable Cities and Society’, also is expected to helps authorities to understand the geographical distribution of the hard to decarbonise houses and enable them to target and deploy interventions efficiently.

The model was ‘trained’ using a variety of data for the home city of Cambridge, including energy performance certificates as well as data from street view and aerial view images, land surface temperature and building stock.

In total, their model identified 700 hard to decarbonise houses and 635 non-hard to decarbonise houses.

With the model trained, it should be able to be applied in other British cities as well as elsewhere even where datasets are very patchy.

The researchers are currently working on a more advanced framework which will bring additional data layers relating to factors including energy use, poverty levels and thermal images of building facades.

The model is capable of identifying specific parts of buildings, such as roofs and windows, which are losing most heat, and whether a building is old or modern, but the level of detail and accuracy is expected to be significantly increased.

They also are training AI models based on other UK cities using thermal images of buildings, and are collaborating with a space products-based organisation to benefit from higher resolution thermal images from new satellites.