Framework reveals smarter, faster way to retire U.S. coal-fired power plants

Despite the steady decline in coal-fired power generation in the United States, more than 100 power plants still have no plans for retirement. This difference is large enough to derail national climate goals. A new study led by researchers at the University of California, Santa Barbara shows how a targeted, data-driven approach can help accelerate the transition and offers a way forward.

The study, published in Nature Energy, addresses the important question of why so many coal-fired power plants are still operating, if market forces have already forced many to close. Approximately 105 gigawatts of coal-fired power generation, representing 114 power plants, is still scheduled to operate until 2035, despite years of decline, even though a complete phase-out by that date is widely considered essential to achieving the U.S. net-zero emissions goal.

“Coal is complex. There’s no single right way to deal with it,” said Sidney Guthrid ’22, lead author of the study. “Our goal was to build a tool that reflected that complexity and allowed different stakeholders to address different aspects of the problem. There is no straight path, and we wanted to do research that represented that reality.”

In collaboration with Grace C. Wu, an assistant professor in the Environmental Studies Program and the paper’s lead author, Guthrid and his team showed that to achieve these goals, policymakers need to move beyond age-based or one-size-fits-all approaches and focus on the specific circumstances that drive the retirement of particular coal-fired power plants.

To that end, researchers including Jeremy Wayland, Stuart Wayland ’22, and Ranjit Deshmukh, associate professor in the Environmental Science Program and Bren School of Environmental Science and Management, developed a new framework that combines graph theory and topological data analysis to classify the entire U.S. coal fleet into eight distinct groups based on 68 technological, economic, environmental, and sociopolitical factors. It also introduced a “contextual retirement vulnerability” score that measures each factory’s susceptibility to early retirement by comparing it with facilities that have already announced closures.

The framework goes a step further by identifying “retirement archetypes,” or patterns that explain why each group of plants retires. These range from regulatory and health-based factors to adverse economic or political pressures, and provide clear measures that can be applied to similar facilities elsewhere.

“We asked not just why coal-fired power plants are being retired, but how can we do it faster and do it in an efficient and data-driven way,” Guthrid said. “Our framework helps policymakers and advocates identify where they can have the greatest impact.”

The research began as Guthrid’s senior thesis in UCSB’s Environmental Studies Program and evolved into a multi-year collaboration, with support from the campus’ Manalis Leadership Fellowship sponsored by Howard and Lisa Wenger. Wu said the scope and impact of this project is unusual for undergraduate research.

“This is doctoral-level research,” Wu said. “It is highly unusual for a project that began as a senior thesis to reach this level of sophistication and impact. What is interesting is that this framework not only explains which power plants are likely to be retired, but also shows how to accelerate decommissioning using aligned drivers with other decommissioned or soon-to-be decommissioned coal-fired power plants.”

The researchers used a model to group 198 operating coal-fired power plants in the United States into clusters, such as plants with high health impacts, expensive plants, and plants in anti-coal areas, each associated with specific vulnerabilities that could be targeted for policy and advocacy. For example, factories associated with high asthma rates and poor air quality may be prioritized through public health campaigns and environmental regulations, while factories facing economic hardship may respond more effectively to economic incentives and market-based mechanisms.

One notable example is Belews Creek in North Carolina. The plant is a nearly 50-year-old coal-fired power plant with a capacity of 2.49 gigawatts, and the study classifies it as highly likely to be retired and part of Group 0: Mixed Fuel Power Plants. Although the facility can burn up to 50% natural gas, it remains one of the nation’s top sources of particulate pollution, ranking 26th out of 198 in particulate emissions. Financially, it is one of the least profitable factories in the country, with approximately $46 million in debt as of 2020.

Bellews Creek is located in a state that has seen rapid growth in solar power development and the implementation of coal debt securitization policies aimed at helping utilities transition away from uneconomic fossil assets. The authors note that “given the drivers of retirement for this group, advocates can leverage state and utility clean energy goals.”

There were even preliminary discussions about replacing Bellews Creek with a small modular reactor, but the plant’s owner, Duke Energy, has since delayed decommissioning it, highlighting the financial and operational complexities that the UCSB framework aims to untangle.

“We can simplify the approximately 200 plants into distinct groups and combine each with evidence-based strategies,” Wu said. “This is a powerful approach to a geographically diverse and politically fragmented challenge.”

According to their analysis, about 28% of coal-fired power plants without a retirement plan are already at a high probability of closure, potentially a “quick win” for policymakers and advocates. However, it also became clear that the most vulnerable plants were distributed across multiple groups, highlighting the need for diverse strategies to deal with the most persistent plants.

The impact extends beyond coal. The model has the potential to be adaptable to other complex decarbonization challenges because it captures the multidimensional forces that shape real-world decisions, including economics, politics, health, and grid reliability.

Wu, whose research focuses on sustainable energy transition planning, said the framework bridges mathematical science and applied environmental science and has the potential to change the way analysts and decision makers approach energy policy.

“This research employs state-of-the-art mathematical tools and puts them in practitioners’ toolboxes,” she said. “It is flexible, transparent, and reproducible, which is exactly what we need to make smarter, more strategic decisions about the energy transition.”

Guthrid, now co-founder of Los Angeles-based AI and data startup Krv Analytics, said the framework’s open-source design makes it especially valuable.

“The methods we have developed are intended to be used,” he said. “Whether you’re tackling coal, renewable energy, or industrial emissions, the idea is the same: Use the data you have to see where progress will happen first and why.”