Kyle Bradbury is managing director of the Energy Data Analytics Lab at the Duke University Energy Initiative. Recently, the Rhodes Information Initiative at Duke (iiD) asked him to explain what he works on and how he involves students through the Data+ and Bass Connections programs. Here are excerpts from their conversation:
Locating Energy Infrastructure Using Satellite Imagery
My research focuses on how we can make energy systems more affordable, accessible, reliable, and clean using machine learning and data analysis tools. The team that I work with, we’re working on questions around understanding where energy infrastructure is using satellite imagery. One of the challenges in this space is called geographic domain adaptation. If I have an algorithm that’s able to find solar panels in California and I train my algorithm there, how am I able to then use that to find solar panels in Africa, Asia, or Europe? Being able to transfer that can really increase the impact of the research that we’re doing, but it leads to a lot of challenging technical issues.
Another research area that I’ve been working on with other members of the team is looking at how we can use data to address some of the challenges in the energy access space. Right now there are close to a billion people around the world that don’t have access to electricity, but we don’t necessarily know which specific communities lack access, and we don’t always know where the grid infrastructure is that could potentially provide access to electricity.
iiD has been a fantastic resource, especially with their program Data+. Data+ is a ten-week summer program for undergraduates to deeply engage in a data-focused research project. Over the last few years, we’ve engaged numerous undergraduates to help us with our research. They produced datasets and laid the foundation for dozens of research papers that have been able to answer some of these really challenging questions at the intersection of energy systems and machine learning.
This year, Bradbury is leading a Bass Connections project team, A Wider Lens on Energy: Adapting Deep Learning Techniques to Inform Energy Access Decisions, which builds on the work of last summer’s related Data+ project.
Video by the Rhodes Information Initiative at Duke (iiD)