Harnessing AI for a Greener Grid: The Pioneering Work of Bolun Xu 

With seed funds from the Data Science Institute, Bolun Xu is driving the transition to renewable power. 

As the world races to wean itself off fossil fuels and move towards renewable energy, integrating variable energy sources like solar and wind into the grid is a complex puzzle. Large-scale batteries are essential for storing up wind and solar for later use, but the success of these batteries in reducing costs and emissions is determined by more than the technology: Electricity markets decide when, how, and even if energy sources are widely adopted and used.

Although current market designs might support the initial use of energy storage, they can struggle to balance economic and environmental goals as storage systems expand. How can we ensure that the solutions we adopt are both effective and financially viable, now and in the future? 

At Columbia University, Bolun Xu, a member of the Data Science Institute and Assistant Professor in the Department of Earth and Environmental Engineering, is working to address these issues, developing computational tools to lead the way toward a grid that is both climate-ready and cost-effective.

His work has garnered major funding and recognition, including a National Science Foundation (NSF) CAREER award, a powerful vote of confidence in a young investigator’s research.  

But when discussing his success, Xu says one award that helped propel his research stands out: a Data Science Institute Seed Funds grant in 2022. These grants support promising faculty research collaborations that bring data science and AI to new domain areas.

“The Data Science Institute Seed Funds award was my first grant at Columbia, and all the others followed,” said Xu, who joined the Columbia University faculty in 2020. “That project has branched off into at least five related projects.” 

A Catalyst for Innovation: Data Science Institute (DSI) Seed Fund

The DSI Seed Funds were established in 2018, and since then they have supported 38 projects exploring a range of topics across fields and disciplines, from assessing the psychological impact of no-knock search warrants on communities to developing new AI-based tools to improve detection of disease. In the past two years, Seed Funds have supported projects in 18 departments across 9 schools.

Xu’s grant supported a collaboration with Upmanu Lall, a visiting professor and the Director of the Columbia Water Center at the Climate School, to advance research on the value of energy storage in mitigating climate impacts on the electric grid — an area that’s become central to plans to decarbonize electricity. Although the tool has not been deployed, it performed very well in simulations. 

Their work involved a deep dive into the economics of energy storage. Using data from New York, Xu and his colleagues devised an AI algorithm for energy price arbitrage, where electricity is bought at low prices and discharged when prices are high, which lowers the cost of electricity overall. The tool, which performed very well in simulations, has not been deployed.   

Scaling Up: From Theory to Real World Application 

Since the initial DSI funding, Xu’s research has expanded, contributing to ambitious, large-scale projects that apply his theories to real-world challenges. A forthcoming project will investigate integrating hydro storage, which is an energy storage system that uses gravity and water to store and generate electricity, into the power grid in New Mexico. 

“This is exactly the topic we studied with the DSI Seed Funds, but applied to a real pumped hydro storage project,” said Xu.

In addition to his work on pumped hydro storage, Xu is conducting research in collaboration with California’s power system operator, developing models to optimize the deployment of dispatch storage resources on the state’s grid, and working with Johns Hopkins University to assess the emissions impact of home battery systems used to store energy for personal use. Other projects explore integrating power systems and climate models, studying consumer behaviors in power systems, optimizing carbon capture and utilization processes, and advancing market design and operational optimization in energy storage. 

The Data Science Institute Seed Funds award was my first grant at Columbia, and all the others followed. That project has branched off into at least five related projects.

– Bolun Xu, Data Science Institute member and Assistant Professor in the Department of Earth and Environmental Engineering

All of these projects draw on the initial research launched through the DSI Seed Funds, and even rely on the high performance computing resources Xu was able to acquire through that grant.

“Bolun is an outstanding, super-focused young researcher who provides an excellent demonstration of how seed funds can help build not just a winning proposal but also define a career,” says Lall.

The Next Data Challenge: Bridging the Gap Between Climate Models and Energy Research

While much of Xu’s current research builds on what he was able to accomplish with seed funds, and he and his team have established themselves as leaders in combining energy storage models with machine learning techniques, one of the more vexing aspects of Xu’s initial Seed Funds proposal – incorporating climate modeling data so his models can take climate change into account–has proven more difficult. 

Because climate change, and the resulting changes in weather patterns, can affect both renewable energy supply (in the form of sunshine and wind) and demand (often in the form of heating and cooling during extreme weather events), it is important that his models take changing weather into account. 

But as Xu discovered, this is not as straightforward a task as one might expect. 

This is an important challenge to solve because as climate change makes weather more volatile, this impacts both energy supply (in the form of sunshine and wind) and demand (often in the form of heating and cooling during extreme weather). 

While the data for power systems modeling requires high space and time resolution, the data from climate models tends to be in more coarse scales, limiting its applicability in high-resolution power system models. Without the ability to integrate this data, Xu has worked to combine power system studies with specific climate-related data points, a challenging workaround until he can develop a systematic solution.

“Bridging that gap in granularity is one of our primary goals,” said Xu. “We need to be able to answer critical questions: What happens if the Earth is one degree warmer in 20 years? Two degrees? Three? Each scenario impacts energy storage decisions in profound ways. We’ve made significant strides, but we are far from done. The future of energy storage—and the stability of our grid—depends on finding solutions to these complex problems.”