Automating Decarbonization Through AI
Artificial intelligence, algorithms, and software are emerging as powerful tools for building a more sustainable future. To lead the development of a new field that applies computing and data science to reduce carbon emissions across major infrastructure sectors, UMass Amherst has been awarded $12 million over five years by the U.S. National Science Foundation.
This new field of computational decarbonization, or CoDec, aims to coordinate and automate decarbonization efforts in computing, electricity, buildings, and transportation. Ultimately, the project seeks to optimize carbon efficiency by aligning energy-intensive activities with the availability of clean energy.
David Irwin, professor of electrical and computer engineering, is a key researcher in the project. Irwin’s work focuses on sustainable computing, smart grid technologies, and the interface between computing systems and energy infrastructure. He emphasizes the importance of flexibility in infrastructure, especially in computing.
“The different infrastructures we’re looking at in this project all have different dimensions of flexibility in terms of time and space,” Irwin explains. “For example, the heating in a building can’t be shifted in space—the building can’t move. However, computing is uniquely flexible across time, space, and performance. You can move running a computing job or serving a web page from New York to California. And you can do it very quickly. But all of these infrastructure systems have some flexibility that we’ll be looking to exploit.”
Another leading contributor is Jimi Oke, assistant professor of civil and environmental engineering, whose research combines systems modeling, optimization, and data science to address sustainability challenges in transportation and infrastructure. At UMass Amherst, Oke has worked on a wide variety of dynamic modeling projects, including schematically mapping the adoption of micromobility (bicycles, e-bikes, e-scooters) as well as applying machine learning to predict the factors that will lead to accidents on the road.
According to Oke, “The transportation sector accounts for nearly 30% of total US emissions. The trucking industry, which moves the majority of our domestic freight, is responsible for about a quarter of this share. The convergence of advances in truck automation and electrification promises to yield significant benefits in terms of decarbonization, air quality, and efficiency.
“However, there are as yet no definitive pathways to achieve these decarbonization goals while also integrating AI and accounting for automation. In this project, our team plans to leverage optimization and AI-augmented simulation tools to find robust strategies addressing facilities, charging, batteries, behavior and the several uncertainties that play into these decisions. We will also investigate synergies resulting from prioritizing corridors with potentially deeply decarbonized power grids.”
The rest of the UMass Amherst collaborators on the project are Manning College of Information and Computer Sciences faculty members Prashant Shenoy (PI), Mohammad Hajiesmaili, Ramesh Sitaraman, Neena Thota, and W. Richard Adrion.
The CoDec team was assembled with an interdisciplinary approach in mind, which is essential for addressing the issue of infrastructure interdependencies. As an example of such interdependence, while working from home reduces transportation emissions, it can increase residential energy use. The CoDec team aims to resolve these kinds of trade-offs through data and computational methods.
Irwin elaborates on the human aspect of automating such trade-offs: “If the performance from a website goes down a little bit, that might be okay, but if I can’t keep my home warm, that’s not okay. The system is going to have to be more flexible to handle the increase in demand and prioritize different uses of energy.”
Together, the team will lead the CoDec initiative through a research framework based on sensing, optimization, and implementation. The team will measure carbon footprints across various infrastructure systems, design AI models to match energy use with clean energy availability, and ultimately develop software that can implement these optimizations in real-world systems.