Creating Artificial Worlds with AI to Improve Energy Access Data (2021-2022)
Critical information for energy access decision-making and electricity system planning is not universally available, including information on village-level electricity access and reliability as well as the location and characteristics of power system infrastructure. Decision-makers require information to determine the optimal strategies for deploying energy resources to decide where to prioritize development and how electrification should be accomplished.
Previous work has shown that these information gaps can be filled by applying machine learning techniques to satellite and aerial imagery. However, there are major challenges that remain with scaling these techniques to apply across large data of varying geographies and many categories of objects. To be effective, more robust techniques need to be developed to assess infrastructure relevant for sustainable transitions.
This project is a continuation of the 2020-2021 team. In past projects, team members have begun to develop deep learning models that can detect energy infrastructure in satellite imagery. However, these techniques struggle when they are applied in settings different from the training data, such as different geographies, sensor modalities and season of data collection.
Building on previous teams’ work, this year’s team will use 3D modeling to represent energy infrastructure in satellite imagery, expanding the methods for generating synthetic imagery to include generative models. Team members will investigate techniques for creating realistic training data with even less information by creating synthetically generated data. The team will create a labeled synthetic energy infrastructure remote sensing dataset generation tool and dataset.
The goals of this project are to:
- Apply synthetic data generation techniques including generative models and style transfer for creating synthetic overhead imagery
- Compare these techniques with traditional, high-effort techniques for training data collection and recent 3D modeling approaches
- Share these tools and any generated data in an open-source repository to encourage use by other researchers and decision makers
Learn more about this project team by viewing the team's video.
Dataset of geographically diverse synthetic overhead imagery; tool shared on GitHub for synthetic generation of overhead imagery; project website and final presentation; conference or journal paper
Fall 2021 – Spring 2022
- Fall 2021: Complete intensive short course on data analytics, energy systems, energy access, research techniques and project management; compete with other students in Kaggle machine learning competition; develop goals and roles
- Spring 2022: Execute project; meet with guest speakers from industry and academia; refine research results; create presentations and posters
This Team in the News
See related teams, Tracking Climate Change with Satellites and Artificial Intelligence (2022-2023) and Deep Learning for Rare Energy Infrastructure in Satellite Imagery (2020-2021), and Data+ summer project, Tracking Climate Change Impacts with Satellites and AI (2022).
Image: Wind farm, by pfly, licensed under CC BY 2.0
- Kyle Bradbury, Pratt School of Engineering-Electrical & Computer Engineering|Energy Initiative
- Jordan Malof, Pratt School of Engineering-Electrical & Computer Engineering
/graduate Team Members
Saksham Jain, Electrical/Computer Engg-MS
Ben Ren, Electrical/Computer Engg-PHD
Boya Sun, Interdisciplinary Data Science - Masters
Katherine Wu, Master of Engineering Mgmt-MEG
/undergraduate Team Members
Caleb Kornfein, Statistical Science (BS), Computer Science (BS2)
Alexander Kumar, Computer Science (BS)
Caroline Tang, Mathematics (BS), Statistical Science (BS2)
/yfaculty/staff Team Members
Leslie Collins, Pratt School of Engineering-Electrical & Computer Engineering
T. Robert Fetter, Nicholas Institute for Environmental Policy Solutions
Marc Jeuland, Sanford School of Public Policy
Luana Lima, Nicholas School of the Environment-Environmental Sciences and Policy
Robyn Meeks, Sanford School of Public Policy