WATCH: Energy Infrastructure Detection with Satellites: Synthetic Imagery for Finding Wind Turbines

Duke students' Bass Connections research on energy access and data analytics comes together in a final energy presentation on synthetic imagery used to improve automated wind turbine detection in satellite imagery, especially when applied to diverse locations.

Efforts to ensure energy access across the globe are often hampered by a lack of critical information to guide decision-making and electricity system planning. Information on village-level electricity access and reliability, as well as the location and characteristics of power system infrastructure, is especially scarce. Decision-makers require this information to determine the optimal strategies for deploying energy resources, like where to prioritize development and whether electrification should be accomplished through grid expansion, micro-grids, or distributed generation.

During the 2020-2021 school year, a Bass Connections research team at Duke University aimed to develop deep learning techniques that can automatically and rapidly scan massive volumes of remotely sensed data, such as satellite imagery, to develop detailed maps of energy infrastructure. These deep learning approaches may provide powerful tools for researchers, policy-makers, and governments to collect energy systems information. This video captures the Bass Connections team's end-of-year presentation in April 2021.

The team used machine learning to create a model that detected wind turbines solely from satellite imagery by training it first with real images of turbines. Since these images are scarce and in practice the machine learning techniques need to be applied to different locations than from where the training data are available, this approach was compared to data resulting from a model which also was trained on synthetic images of wind turbines. Synthetic images, while they might look real to the machine, are generated images and are not genuine photos. Feeding the model synthetic images of wind turbines increased the accuracy or "average precision" of the predicted turbine location.

Bass Connections is a unique Duke University program that brings together faculty, postdocs, graduate students, undergraduates, and external partners to tackle complex societal challenges in interdisciplinary research teams.

Student Team Members: Ada Ye (T'23), Jessie Ou (T'22), Wendy Zhang (T'21), Eddy Lin (T'22), Tyler Feldman (T'23), and Jose Moscoso (MIDS '21)

Faculty Team Leaders: Kyle Bradbury (Pratt School of Engineering and Managing Director of the Energy Data Analytics Lab at the Duke University Energy Initiative) and Jordan Malof (Pratt School of Engineering)

Learn more about the project:

• Bass Connections
• The website created by the team