Energy Data Analytics Lab: Energy Infrastructure Map of the World through Satellite Data (2018-2019)

Over 15% of humanity has no access to electricity, and far more have unreliable access that precludes most productive energy uses that are beneficial for improving economic prosperity, health and education. Decision-makers require information to determine the optimal strategies for deploying energy resources to decide where to prioritize development and whether that development should be through grid expansion, microgrids or distributed generation. However, two critical data sources for such planning—who has access to electricity and the location of electric infrastructure—are often unavailable or overly time-consuming to collect and maintain.

To address these needs, this Bass Connections project set a goal of working toward creating an energy infrastructure map of the world using satellite imagery. The project team employed state-of-the-art machine learning tools in order to detect critical energy infrastructure in satellite imagery including power distribution lines and off-grid distributed energy generation sources such as solar photovoltaics. Team members approached their analysis from an integrated systems perspective. Instead of modeling individual components of the system independently (generation, distribution, consumption), the team employed techniques that consider these components as part of an integrated whole, comprising a national grid and perhaps local off-grid microgrids. This was the first step toward a functional energy infrastructure map of the world.

Team members used both publicly available data from the U.S. as well as purchased data covering underserved parts of south and central Asia. This work, combined with the research from previous Bass Connections and Data+ projects, provides a toolset that will enable the remote assessment of generation, transmission, distribution, end-use and access to electricity. The team plans to unify these components toward a vision of a methodology for automatically mapping and assessing energy systems in any place on Earth.


Fall 2018 – Spring 2019 

Team Outputs 

Electric Transmission and Distribution Infrastructure Imagery Dataset

Mapping Electricity Infrastructure with Deep Learning (poster by Ben Alexander, Wendell Cathcart, Atsushi Hu, Varun Nair, Lin Zuo, Kyle Bradbury, Leslie Collins, presented at Bass Connections Showcase, Duke University, April 17, 2019)

Automatically Identifying Energy Transmission Infrastructure in Satellite Imagery (poster by Ben Alexander, Wendell Cathcart, Yutao Gong, Atsushi Hu, Varun Nair and Lin Zuo, presented at Duke Research Computing Symposium, Duke University, January 16, 2019)

Interactive exhibit (presented at Energy Week, Duke University, November 2018)

Kyle Bradbury. Solar Array Identification in Adelaide, Australia ($14,626 grant awarded from Australia Energy Market Operator, 2019)


Mapping Electricity Infrastructure with Deep Learning

This Team in the News

Donor Support Spurs Interdisciplinary Research on Pressing Global Challenges

See related teams, A Wider Lens on Energy: Adapting Deep Learning Techniques to Inform Energy Access Decisions (2019-2020) and Energy Data Analytics Lab: Electricity Access in Developing Countries from Aerial Imagery (2017-2018), and a related Data+ summer project, Energy Infrastructure Map of the World (2018).

The team presenting their poster

Team Leaders

  • Kyle Bradbury, Pratt School of Engineering-Electrical & Computer Engineering|Energy Initiative
  • Leslie Collins, Pratt School of Engineering-Electrical & Computer Engineering

/graduate Team Members

  • Wendell Cathcart, Master of Environmental Management, Energy and Environment

/undergraduate Team Members

  • Ben Alexander, Statistical Science (BS)
  • Yutao Gong, Environmental Sciences (BS)
  • Xinchun Hu, Mathematics (BS)
  • Varun Nair, Computer Science (BS)
  • Lin Zuo, Statistical Science (BS), Computer Science (AB2)

/yfaculty/staff Team Members

  • T. Robert Fetter, Nicholas Institute for Environmental Policy Solutions
  • Marc Jeuland, Sanford School of Public Policy
  • Jordan Malof, Pratt School of Engineering-Electrical & Computer Engineering
  • Robyn Meeks, Sanford School of Public Policy
  • Guillermo Sapiro, Pratt School of Engineering-Electrical & Computer Engineering