Building housing energy efficiency models from local sensing

Decarbonising domestic heating is a central challenge for the sustainability agenda - requiring fundamental changes in home heating systems and major upgrades in the thermal performance of the existing building stock to reduce energy consumption.

This research will develop city-wide indicators using a combination of urban sensing technologies, digital footprint data and machine learning (ML) models.

Through this project, we will collaborate with a start-up company in London (SatVU, Global Satellite VU Ltd) and develop a new deep learning methodology to explore the use of very high-resolution thermal satellite imagery (3.5m/pixel) to estimate and predict the building energy efficiency.

In the analysis, we will combine with a range of property-related data (Zoopla, ROS, EPC, etc.), held by UBDC Housing & Neighbourhoods research groups, and other sources such as Google Street View and aerial thermal/LiDAR surveys.

Using ML, we will build models that let us scale-up the estimates - from a sample of properties scanned by SatVU - to the wider city. We will also use the smart metre data and indoor environment measurement data collected from the Understanding Society Innovation panelto validate our model performance.

All these measures will help us obtain a holistic picture of the energy efficiency of housing stock in the deprived neighbourhoods and serve as a valuable data product and practise for larger-scale deployment.

Aims and objectives

This project aims to develop indicators to support efforts to raise the efficiency of our domestic building stock.

The project team will:

  • Understand the energy efficiency and energy usage of the domestic building stock in deprived areas in Greater Glasgow.
  • Generate a heat loss index for major cities in the UK to identify areas with the most urgent need for decarbonisation and energy efficiency improvement. This will provide evidence and suggestions for housing retrofitting planning.
  • Create open-source code and ML models for processing high-resolution remotely sensed and street-level images. This methodology has the potential to extend to other cities.


Lead: Dr Qunshan Zhao (Urban Analytics at UBDC)

Glasgow team:

Dr Qiaosi Li (PDRA in Urban Sensing, Glasgow)

Cambridge team:

Maoran Sun (

Ronita Bardhan (

SatVU team:

Manuel Esperon (

Jointly funded by