Satellite images and AI shed light on building energy use

Computer vision algorithms have been developed that can predict energy consumption from meaningful features contained in satellite imagery.

The so-called SCHMEAR (Scalable Construction of Holistic Models for Energy Analysis from Rooftops) approach models and quantifies how much the context around a building contributes to its energy profile.

This compares with traditional approaches, such as either simulating a building’s structural features or relying on basic building information such as age and construction materials used, which are limited in scope and scale.

In addition, they lack local effects such as ‘heat islands’ that absorb heat and reflect it back to the surroundings or ‘canyons’ that can affect airflow.

Have you read?
E.ON launches home energy efficiency drive in Sweden
Webinar Recording: How Digital Twin technology can help homes and buildings perform better

Conversely, the SCHMEAR model can be scaled to larger areas and potentially complete cities as larger scale solutions are sought – and potentially any city anywhere for which satellite data is available.

“Buildings use energy in a holistic world that involves the interactions of not just the systems and people within the building, but also interactions with exterior systems like the streets and the trees outside,” says Stanford University graduate Thomas Dougherty, who developed the model for his doctoral research.

“The fact that a building is in a dynamic environment may change its energy use quite a bit.”

The SCHMEAR computer vision analysis relies on a convolutional neural network, which is a form of AI.

The research found in modelling New York City building energy use that a SCHMEAR model based on a single closeup satellite image provided as much useful information for predicting a building’s energy consumption as the data driven model built from basic curated building data.

A next step is to establish which features of a building contribute to the model’s predictions. In an initial approach using saliency mapping’, Dougherty reports showing that in Manhattan, the region around a building is a much more significant predictor of energy use than it is in a place like Queens, where the buildings are more spaced out.

Another future step is to add different kinds of data such as that from different wavelength satellite instruments or building outlines.

For example, the analysis has shown that the SCHMEAR model is not as good as the data-driven model at taking building height into consideration.