Energy Data Analytics Lab: Electricity Access in Developing Countries from Aerial Imagery (2017-2018)
Access to reliable electricity is strongly correlated with economic prosperity and health. Over 15% of humanity has no access to it at all, and far more have access only to intermittent supplies that preclude most productive energy uses.
Research on sustainable energy transitions being conducted at Duke and elsewhere is trying to better understand the drivers and impacts of electrification on health, land use, the environment and the local economy. However, current methods for assessing access rely almost completely on household surveys or highly aggregated (e.g., national-level) data sources.
This project designed innovative techniques to evaluate electricity access in developing countries through machine learning applied to aerial imagery data. The team filled critical gaps in energy access data collection and evaluation by piloting and then applying automated algorithms for generating spatially-disaggregated data on electricity access in developing countries using aerial imagery. The team also developed new machine learning techniques to extract information on energy infrastructure and consumption using high-resolution aerial imagery.
In 2017-2018, the team focused on producing high resolution estimates of electrification rates in the Indian states of Bihar and Uttar Pradesh. The team curated a dataset of electricity access survey data and satellite imagery for all villages in Bihar and extracted features from satellite bands to be input into an automated classifier. The team trained a gradient-boosted decision tree classifier using extracted features such as light at night, population density, and possible agricultural activity, to predict whether 16,389 individual villages in Bihar were either electrified or un-electrified based on ground truth data from the Indian government’s GARV dataset.
The team then then used this algorithm to predict electrification of villages in the neighboring state of Uttar Pradesh.
Fall 2017 – Spring 2018
Automating Electricity Access Prediction Using Satellite Imagery (poster by Shamikh Hossain, Shijia Hu, Prithvir Jhaveri, Harshvardhan Sanghi, Joe Squillace, Brian Wong, Xiaolan You, Kyle Bradbury, Leslie Collins, T. Robert Fetter, Marc Jeuland, Timothy Johnson), presented at Bass Connections Showcase, April 18, 2018, and Visible Thinking, April 19, 2018
Brian Wong (M.E.M. Program), Bass Connections Award for Outstanding Mentorship
Automating Electricity Access Prediction with Satellite Imagery (Fangge Deng, Shamikh Hossain, Prithvir Jhaveri, Ashley Meuser, Harshvardhan Sanghi, Joe Squillace, Anuj Thakkar, Brian Wong, Xiaolan You, Kyle Bradbury, Leslie Collins, T. Robert Fetter, Marc Jeuland, Timothy Johnson; first prize, 2018 Duke Research Computing Symposium Poster Competition, January 22, 2018)
This Team in the News
See related teams, Energy Data Analytics Lab: Energy Infrastructure Map of the World through Satellite Data (2018-2019) and Energy Data Analytics Lab (2016-2017), and a Data+ summer project, Electricity Access in Developing Countries from Aerial Imagery
- Kyle Bradbury, Energy Initiative
- Leslie Collins, Pratt School of Engineering-Electrical & Computer Engineering
- Timothy Johnson, Nicholas School of the Environment-Earth and Ocean Sciences
/graduate Team Members
Brian Wong, Master of Environmental Management, Global Environmental Change
/undergraduate Team Members
Fangge Deng, Computer Science (AB), Environmental Sci/Policy (AB2)
Shamikh Hossain, Computer Science (BS), Economics (BS2)
Prithvir Jhaveri , Computer Science (BS)
Ashley Meuser, Electrical & Computer Egr(BSE), Computer Science (BS2)
Harshvardhan Sanghi, Mechanical Engineering (BSE), Economics (AB2)
Joseph Squillace, Computer Science (AB)
Anuj Thakkar, Mechanical Engineering (BSE)
Xiolan You, Computer Science (BS), Statistical Science (BS2)
/yfaculty/staff Team Members
T. Robert Fetter, Nicholas Institute for Environmental Policy Solutions
Marc Jeuland, Sanford School of Public Policy
Guillermo Sapiro, Pratt School of Engineering-Electrical & Computer Engineering