Artificial intelligence can determine whether a building is energy efficient by analysing existing data. The technology could be used to get a broad picture of the energy efficiency of buildings in different countries.
at Stanford University in California and his colleagues trained and tested an AI on remote-sensing and public data for almost 40,000 buildings in the UK. This included Google Street View images, aerial images and satellite-based measurements of building heat loss, as well as information about individual building dimensions and construction materials.
These buildings also had energy performance certifications, which range from 鈥淎鈥� as the best to 鈥淕鈥� as the worst. To simplify the task, the researchers classified buildings with certifications A to D as energy efficient, and those with grades E to G as not energy efficient. That distinction could make a practical difference, for example, for whether it makes economic sense to install a heat pump in a certain building, as the technology is expensive and works best in buildings that are already energy efficient.
Advertisement
The researchers then challenged the AI to predict the energy ratings of 3700 of the buildings using just the other types of data. It achieved a performance score of around 70 per cent. By comparison, a model that simply predicted the same thing each time only achieved a score of 45 per cent.
While the UK and countries in the European Union already often require building owners to get energy-performance certificates through a physical inspection, the US and many other countries do not. 鈥淭he benefit of using these images is you will be able to extend [the AI analysis] to a global level,鈥� says at the University of Glasgow in the UK, who has done on using Street View images to estimate building energy efficiency. 鈥淵ou get Street View images in most countries.鈥�
Higher-resolution thermal infrared imagery from new satellites could be used to boost AI predictions, says Zhao. Another approach, which both Zhao and Mayer are investigating, includes using laser-based scans of buildings鈥� physical facades.
Having a clearer picture of building energy efficiency is vital at a time when residential and commercial buildings consume about 40 per cent of all energy used in the US, UK and European Union. 鈥淚 hope that this remotely sensed approach can be one piece in helping us allocate investments for decarbonisation in a smart way,鈥� says Mayer.
Reference:
