You are here

Gauging Renewable Energy Generation Using Satellite Imagery

Duke-World Resources Institute team among five finalists in the GBDX Sustainability Challenge

Researchers connected with Duke's Energy Data Analytics Lab and Applied Machine Learning Lab, working in partnership with the World Resources Institute, are among five finalist teams in the GBDX for Sustainability Challenge. 

The GBDX for Sustainability Challenge is sponsored by DigitalGlobe, a Maxar Technologies company, with support from Amazon Web Services. The competition is intended to spur innovation with the potential to address the United Nations' 18 Sustainable Development Goals.

More than 70 teams proposed bold ideas using geospatial big data to drive progress toward the Sustainable Development Goals. Five teams—including Duke's—have been named finalists. For the next two months, each finalist team will have access to Digital Globe's geospatial big data platform (GBDX) and 100+ petabyte image library, one of the largest collections of satellite imagery data available to the public or private sectors. In April, judges will review their progress and select an overall winner. 

The Duke/World Resources Institute team will undertake a project using high-resolution satellite imagery and computer vision to build an open database of global power plants.

"Providing clean, affordable energy is one of the most important challenges the world faces today, as highlighted in Sustainable Development Goal 7 (Affordable and Clean Energy)," noted the team's proposal. "Achieving that goal will require a range of actions, many of which depend on a detailed analysis of the electric power sector. Unfortunately, data about the power sector is limited or lacking in many countries, or only available in proprietary formats."

The team will apply computer vision algorithms to high-resolution satellite imagery provided by DigitalGlobe to geolocate and characterize wind and solar power generation facilities around the world. Information about these power plants is especially important to show how rapidly the power sector is changing. The data will be aggregated and released in open format, to support applications including energy sector transition planning, grid integration of renewable power, and clean energy public education. 

"We’ve developed successful techniques for automatically identifying distributed generation resources in a research environment, using publicly available data," explained Kyle Bradbury, managing director of Duke's Energy Data Analytics Lab. "Access to the massive DigitalGlobe archive will enable us to test our algorithms on a much larger scale and on more recent satellite imagery data."

The Duke / World Resources Institute teams will be competing with four other finalists:

  • UNICEF, Big Pixel Initiative at UC San Diego, Development Seed - "Mapping Schools to Reduce the Digital Divide";
  • Ernst & Young, Cornell University, Metabolic, Geodan, Trinity College Dublin, University College Dublin, University of Amsterdam, University of Toronto - "Global Green City Watch";
  • London School of Economics, University of Oxford - "Monitoring Economic Activity in Large-Scale Industrial Areas"; and
  • - "Deforestation Intelligence for Cerrado."

Duke University's Energy Data Analytics Lab is developing and applying advanced data analytics tools to advance an accessible, affordable, reliable, and clean energy system. The lab is a partnership among the Duke University Energy Initiative (its institutional home), the Information Initiative at Duke, and the Social Science Research Institute

The Applied Machine Learning Lab is based in Duke's Pratt School of Engineering

The World Resources Institute is a global research organization working on six critical goals that the world must achieve this decade in order to secure a sustainable future: climate, energy, food, forests, water, cities & transport.

Questions about the Energy Data Analytics Lab? Contact managing director Dr. Kyle Bradbury