One in nine people do not have regular access to electricity, which can have negative impacts on health, educational and economic opportunities, and quality of life. In order to determine the optimal strategy for meeting the needs of communities without access, one must first identify where electricity infrastructure already exists.

This information can help developing countries decide whether to expand the national grid, construct microgrids or provide off-grid solar power. However, current approaches to identifying and mapping energy infrastructure are expensive and time intensive.

Building off the work of the 2018-2019 project team and a Data+ summer project, this team investigated improvements to deep learning models that identify energy infrastructure in satellite imagery. A key adjustment involved creating and integrating synthetic data (a new approach to this problem) into the AI model training process. The end goal is to generate maps of power grid networks that can aid policymakers in implementing effective electrification strategies.

Timing

Fall 2019 – Spring 2020

Team Outputs

Kyle Bradbury. Estimating Solar PV Locations and Capacity in San Diego, CA ($14,971 grant awarded from the Department of Energy, 2019)

Kyle Bradbury. Multi-task Learning for Solar Photovoltaic Array Detection ($30,999 grant awarded from the Department of Energy, 2019)

This Team in the News

Watch: A Wider Lens on Energy - Adapting Deep Learning Techniques to Inform Energy Access Decisions

Faculty Perspectives: Kyle Bradbury

Kyle Bradbury on Improving Global Energy Access through Machine Learning and Collaboration

See related teams, Deep Learning for Rare Energy Infrastructure in Satellite Imagery (2020-2021) and Energy Data Analytics Lab: Energy Infrastructure Map of the World through Satellite Data (2018-2019), and Data+ summer project, Deep Learning for Rare Energy Infrastructure in Satelittle Imagery (2020).