What is it about?
Buildings use a large amount of energy, but many cities do not have enough detailed building data to easily compare energy performance or identify where improvements are needed. This study explores whether satellite images, weather information, and basic building data can be used together to estimate how much energy buildings use each month. We developed an artificial intelligence model that looks at buildings and their surrounding urban context at different satellite image scales. This allows the model to consider not only the building itself, but also nearby features such as roads, trees, shadows, density, and surrounding land use. The study also uses explainable AI methods to show which visual features the model relies on, making the results easier to understand. The approach was tested using buildings in Washington, D.C., and showed strong performance in estimating monthly energy use. This work can help cities, researchers, and building managers better understand building energy patterns, especially in places where detailed benchmarking data are limited.
Featured Image
Photo by Dmitrii E. on Unsplash
Why is it important?
This work is important because it offers a more scalable and transparent way to estimate building energy use. Traditional energy benchmarking often depends on detailed building records that are not available in many cities. By using widely available satellite imagery and weather data, this study shows that AI can help fill this data gap. A key contribution is that the model not only makes predictions, but it also helps explain which parts of the surrounding urban environment may influence energy use. This can support more informed decisions about building retrofits, energy policy, urban design, and carbon reduction strategies.
Perspectives
This study shows that building energy benchmarking can move beyond spreadsheets and isolated building records. By combining satellite imagery, weather data, building information, and explainable AI, we can better understand how buildings perform within their real urban context. This perspective is especially valuable for cities seeking practical, data-driven tools to reduce energy use and greenhouse gas emissions, even when local benchmarking datasets are incomplete or unavailable.
Tian Li
University of Nebraska-Lincoln
Read the Original
This page is a summary of: XAI energy benchmarking with multi-scale satellite and tabular fusion, Building Simulation, June 2026, Tsinghua University Press,
DOI: 10.1007/s12273-026-1437-9.
You can read the full text:
Contributors
The following have contributed to this page
