How green is your building New AI tells you instantly

A new AI can predict the energy efficiency of non-domestic buildings almost instantly.

The AI system, created by researchers at Loughborough University together with consultant Cundall, generates a prediction of building emissions rates (BER), used to calculate the energy performance of buildings.

Current methods of producing BERs involve inputting hundreds of variables by hand and take hours or days per building.

Dr Georgina Cosma and Kareem Ahmed of Loughborough’s School of Science, with the support of Edwin Wealend, head of research and innovation at Cundall, trained the AI on data from government energy performance assessments.

In the UK, an energy performance certificate must be completed to assess the energy efficiency of every building sold, rented or constructed. It gives the building a rating from A+ to G and recommends ways to improve it.

Producing these ratings is a slow process because of the time it takes to generate BERs. But the new system takes less than a second to generate a BER with as few as 27 variables, and with little loss of accuracy.

The research focused on non-domestic buildings – training the AI on 81,137 data records from offices, shops, factories, schools, hospitals, restaurants, and cultural institutions – because they tend to be the worst energy offenders. Even though they make up just 8% of Britain’s buildings, they produce 20% of our total carbon emissions.

Being able to assess the energy efficiency of these buildings in seconds will make it much easier to get that number down and move us closer to the goal of net-zero.