Duke students make sense of data to advance energy solutions

Posted On:

Friday, Aug 30, 2019 - 11:27 am

Seventeen Duke students spent their summer solving real-world data puzzles related to smart meters, wholesale energy markets, oil and gas production, energy access, and even Duke University’s own energy use. 

Students in the 10-week Data+ program learned how to marshal, analyze, and visualize data while working in small teams to apply their skills to authentic challenges. Organized by the Rhodes Information Initiative at Duke, the program also exposes students to data science professionals and encourages idea exchange among more than two dozen student teams working in diverse sectors. 

Learn about the six energy teams' work: 

Marco Gonzalez in front of his Data+ presentation

Duke undergraduate Marco Gonzalez (Civil and Environmental Engineering) and graduate students Mengjie Xu (Biostatistics) and Sophia Zu (Statistical Science) spent the summer of 2019 advancing Duke University's efforts to achieve carbon neutrality by 2024. Their Data+ research team processed and analyzed utility data from buildings across campus, creating reports that show historical usage trends, benchmarks for comparison, and practical recommendations.  

Marco reports that “Data+ has definitely made me a better collaborator and problem solver.” He’s excited that a 2019-2020 Bass Connections team will pick up the project, creating easy-to-interpret energy dashboards for clusters of buildings across campus.

Project advisors included Dr. Billy Pizer (Sanford School of Public Policy) and Tim Johnson (Nicholas School of the Environment) as well as Sustainable Duke and Duke Facilities staff.

Teammates Alec and Cass standing in front of their Data+ presentation

Electricity prices in wholesale energy markets are highly volatile. In summer 2019, Duke University undergraduates Alec Ashforth (Economics) and Cassandra Turk (Economics, Mathematics) and recent alum Paichen Li (MEM’19) sought to help their client Tether Energy minimize the risks of trading electricity in California.

Using historical data, this Data+ team identified extreme price events and determined their causes (e.g., air quality, heat, solar production adjustments). Their dataset can now be used to help Tether Energy create a model to predict future occurrences.

Cassandra says the project was a good opportunity to practice collaborative skills: “Alec and I spent a lot of time bouncing ideas off of each other. A lot of these ideas were, frankly, awful, but that's the real magic of working together. We're both comfortable with not feeling like we have to be the smartest person in the room, and it makes us a lot more productive. When we don't know or understand something, we ask each other... When we're overwhelmed with work, we're not afraid to tell each other so, and help out accordingly.”

Eric Butter of Tether Energy advised the team. “Our Data+ students brought enthusiasm and energy to a very important and time-consuming data management project,” he said. “We are very thankful for their work, and look forward to engaging more students in the years to come!”

Victoria and Aaron standing in front of their Data+ presention

Producing oil and gas in the North Sea, off the coast of the United Kingdom, requires a lease to extract resources from beneath the ocean floor and companies bid for those rights. For this Data+ project sponsored by ExxonMobil in summer 2019, Duke University undergraduates Victoria Worsham (Economics, Mathematics) and Aaron Chai (Economics, Computer Science) set out to understand why these leases are acquired and who benefits.

Victoria and Aaron say the project honed their communication skills as they created a single, tabular database of U.K. bid history and work programs. They were working with datasets in many formats—and often had to contact the creators with questions. 

One lesson they’ve learned? “Data can be messy,” Victoria pointed out, “but once it’s cleaned and in a better format, it can be quite enjoyable to run regressions!”

Artem Streltsov of the Energy Data Analytics Lab at the Duke University Energy Initiative served as a mentor to the team.

Dr. Bradbury, Varun, Paul, Fanjie, Andy, and Dr. Malof posing for a Data+ team photo

If you’re trying to design efficient pathways towards electrification in the developing world, it’s helpful to be able to identify existing energy infrastructure. Researchers at Duke have been working on ways to use machine learning to do this using satellite imagery.

But geography matters. An algorithm that learns to find transmission lines in the desert might fail in a forest, since the areas look different.

That’s why a team of Duke undergraduates spent the summer of 2019 exploring a new way to adapt algorithms to new geographic areas: using synthetic data rather than real satellite data.

Jichen “Andy” Yang (Computer Science, Electrical and Computer Engineering) says, “I loved this project because it combined my majors—electrical and computer engineering and computer science—with my minor in environmental science. We got to work with the latest technology, using deep learning techniques to better understand energy and the environment.” 

Yang, Varun Nair (Economics, Physics), and Paul Rhee (Computer Science) worked with graduate student Fanjie Kong (Biomedical Engineering) and faculty advisors Dr. Kyle Bradbury (Energy Initiative) and Dr. Jordan Malof (Pratt School of Engineering) on this Data+ project affiliated with the Energy Data Analytics Lab at the Energy Initiative. A 2019-2020 Bass Connections team will build on their work.

Jason and Josh standing in front of their Data+ poster

Producing oil and gas in the Gulf of Mexico requires rights to extract these resources from beneath the ocean floor and companies bid into the market for those rights. The top bids are sometimes significantly larger than the next highest bids, but it’s not always clear why.

Duke master’s student Jiacheng “Jason” Fan (Economics) and Yueru “Josh” Li of Haverford College delved into this mystery as part of a Data+ project sponsored by ExxonMobil this summer.

One lesson they’ve learned? “When it comes to datasets, sometimes what is missing is as important—if not more important—than what’s already in front of you,” says Josh.

Hyeongyul Roh of the Energy Data Analytics Lab at the Duke University Energy Initiative served as a mentor to the team. 

Jiwoo Song and Jessie Ou standing in front of their Data+ poster

Electricity theft costs utility companies worldwide more than $25 billion each year—and it’s a major issue in developing countries. What if algorithms could help detect anomalies in the energy consumption data collected by smart meters? In summer 2019, Duke undergraduates Jessie Ou (Computer Science) and Jiwoo Song (Chemistry, Mechanical Engineering) worked with graduate students Bernard Coles (Sociology) and Zhenxuan Wang (Environmental Economics and Policy) to apply machine learning techniques to smart meter data collected in Kyrgyzstan. Jessie says their project faced some obstacles: missing data, data from sources that were overlapping or didn’t match, and the lack of labels for data. Still, the team got a great start on developing a new way to identify abnormal consumption trends and possible fraud. The Data+ project team—affiliated with the Energy Access Project at Duke—was advised by Dr. Robyn Meeks (Sanford School of Public Policy).

Dr. Kyle Bradbury, managing director of the Energy Data Analytics Lab at the Duke University Energy Initiative, worked with Dr. Paul Bendich (Rhodes Information Initiative at Duke) to develop several of this year’s six energy projects, including three sponsored by ExxonMobil and Tether Energy. 

“Each year I’m impressed by what our Duke undergraduates are able to accomplish in just 10 weeks,” he said. “They learn about the application area, they learn about the tools needed to analyze relevant data and how to apply those tools, and they learn to communicate what they’ve accomplished to an interdisciplinary audience. They help the project leaders and sponsors think through energy-relevant questions more deeply because of the teams’ hard work.”

Bradbury points out that Data+ also enables graduate students to work on energy industry challenges as project managers and develop skills in team-building and client-based work. “Duke is working to increase opportunities for graduate and professional school students to apply their education to diverse career options, and Data+ does that very well. It’s truly a win-win-win program for industry partners, undergraduates, and graduate students.”

Interested in partnering with Duke on an energy Data+ project? Contact Kyle Bradbury