A trio of studies could help Puerto Rico’s energy system weather future storms

When Hurricane Fiona struck Puerto Rico in 2022, it exposed the vulnerabilities of the island’s energy infrastructure. Though only a Category 1 storm, Fiona caused a total blackout across the island, leaving residents without power for days to weeks with far-reaching health, safety, and economic consequences.

Yet in the aftermath, the hurricane also provided a rare opportunity to learn about a power system during an extreme weather event. LUMA Energy, the private power company that since 2021 has been responsible for power distribution and power transmission in Puerto Rico, collected high-resolution outage data in 10-minute intervals as Fiona made landfall on the island.

A team of Princeton engineers is now using that information about Puerto Rico’s energy grid during Hurricane Fiona to help LUMA Energy and other system operators better understand their power grids in the face of increasingly frequent and severe climate extremes, from hurricanes to heat waves.

Now, the Princeton-led team has, in a series of papers, developed models to quantify the risk of catastrophic blackouts—like the one during Hurricane Fiona—and better forecast how climate extremes will impact energy systems. Beyond today’s challenges, the models also reveal the impacts of climate extremes on a future energy system with high levels of renewable energy.

The work could inform grid upgrades to improve its resilience to the extreme weather events that will become more frequent and severe, partially due to climate change. At the same time, the team’s models could help LUMA Energy and other system operators navigate their clean energy targets while maintaining system reliability.

“Part of the motivation for the clean energy transition is to avoid the worst impacts of climate change, including a rise in more frequent and severe extreme weather events,” said research leader Ning Lin, a professor of civil and environmental engineering.

“However, renewables like solar and wind are more exposed to the environment than fossil fuel power sources, making them potentially more vulnerable to those climate extremes. Large-scale integration of renewables may thus induce grid instability. Our work aims to help energy systems navigate the risks of climate extremes while also achieving their clean energy targets.”

CRESCENT: Unpacking the risks of climate extremes to energy systems

Just what happened during Hurricane Fiona?

Outage data reveals that just before the hurricane’s landfall, Puerto Rico’s grid went from over 50% operational to a total blackout in under 10 minutes. This sudden drop points to an event known as a cascading power failure, in which an outage in one grid component causes a chain reaction that leads the entire system to collapse.

In one paper, published March 16 in Nature Communications, the Princeton team developed a model to quantify the risk of such cascading power outages to Puerto Rico’s energy system in the face of hurricanes and other climate extremes. Known as CRESCENT (Climate-induced Renewable Energy System Cascading Event), the physics-based model combines information about climate hazards with grid vulnerability data to predict the likelihood of a catastrophic blackout.

Using CRESCENT, the researchers simulated 1,000 outcomes that could have resulted from a storm with characteristics like Hurricane Fiona. Through the simulations, they identified patterns in how the hurricane impacted the grid, which could help operators identify critical infrastructure and develop strategies for avoiding blackouts.

In doing so, the team uncovered an interesting and unexpected trend: If the transmission lines from Costa Sur, the largest power plant in Puerto Rico, failed early on during the hurricane, the island’s grid was subsequently more resilient to a total blackout than when they failed later in the storm.

“You might think that you would want the most critical power lines to last as long as possible during a hurricane, but we found that if they were going to fail, the system was actually more resilient to subsequent damage when they failed near the beginning,” said first author Luo Xu, an associate research scholar of civil and environmental engineering at Princeton.

Xu explained that having critical components fail early on served to passively de-energize the grid, similar to how grid operators in Texas implemented rolling blackouts to prevent total grid collapse during Winter Storm Uri in 2021. This de-energization allowed the grid to better distribute power imbalances to other, smaller grid components at a time when fewer had been damaged by the storm.

Conversely, as Hurricane Fiona progressed and caused more damage to the grid, the failure of the critical transmission lines sparked imbalances too great for the rest of the grid to withstand, triggering a cascading power failure.

The researchers said information from CRESCENT can support both short-term and long-term grid planning in Puerto Rico.

In the short term, grid operators could use the model to improve the system’s resilience to an incoming hurricane by identifying the grid components most likely to be the source of a cascading power failure.

“With our simulations, we identified certain patterns in which the system was most resilient to a catastrophic blackout,” Xu said. “In this way, our model can help grid operators mitigate the risk of a worst-case scenario.”

On a longer timescale, the model can inform Puerto Rico’s efforts to fully decarbonize its energy system by 2050 as it explicitly considers the vulnerability of renewable energy systems to climate extremes.

For instance, the model found that selectively adding grid-forming energy storage alongside renewables can significantly reduce the risk of a system-wide blackout. Energy storage becomes particularly important as the penetration of behind-the-meter renewables, such as rooftop solar, increases above 45%—a threshold estimated by the study, above which the risk of a system-wide blackout becomes increasingly likely.

“Intermittent renewables like solar and wind lack the inertia that is usually provided by a spinning turbine under traditional power generation,” said co-author H. Vincent Poor, the Michael Henry Strater University Professor of Electrical Engineering. “Renewables-dominated systems thus require accompanying large-scale storage capabilities or other stabilizing mechanisms to ride through extreme weather events.”

REDUCER: Planning day-ahead grid operations for climate extremes

To effectively manage the power grid, operators need to predict how much energy will be needed at every hour of the day, on every day of the year.

Under normal circumstances, this estimation is relatively straightforward. A day in advance, operators use information from weather forecasts, historical patterns, and expected consumer behavior to decide how many and which generators to run during each hour of the following day to ensure that energy demand and supply are closely matched.

In the face of hurricanes and other climate extremes, however, energy demand forecasts and grid operation strategies can be wildly inaccurate, often leading to service interruptions or the use of expensive emergency generators.

In another paper, published May 14 in Proceedings of the National Academy of Sciences, the Princeton team developed a model to help grid operators better plan their next-day energy supply in advance of climate extremes like hurricanes.

Known as REDUCER (Risk-aware Electricity Dispatch Under Climate Extremes with Renewable integration), the model outperformed existing approaches at capturing potential energy demand losses during climate extremes and managing the associated risks. When applied to Puerto Rico’s power grid before Hurricane Fiona, for instance, it reduced operational costs by 20% compared to leading day-ahead operation strategies and avoided relying on nearly a gigawatt of flexible energy dispatch.

“This prevents the disaster’s impacts from compounding by allowing grid operators and consumers to focus on other recovery issues,” said Poor. “It also translates into a more reliable energy supply overall, and, ultimately, a lower cost of electricity.”

REDUCER outcompeted similar models during Hurricane Fiona because it incorporated risks to the energy system’s extensive distribution networks alongside its transmission network.

If an electric grid were like a road system, the transmission network would be the major interstates that carry vehicles at high speeds over long distances, while the distribution network would be the slower-moving local roads that bring cars to homes and businesses.

However, most day-ahead operation models only capture risks to the transmission network because the sprawling distribution network, which includes rooftop solar, contains so many uncertainties that it is too computationally intensive to incorporate.

“Generally, there’s been no way for operators to incorporate that level of uncertainty into existing models for next-day hourly operations,” said Xu, who is also first author of this paper. “It’s just too computationally intensive to get a quick enough answer.”

By factoring in risks to the distribution network, REDUCER matched energy supply and demand better than existing models—requiring less dispatch of emergency energy sources—while employing advanced optimization techniques to return results over 10 times faster than state-of-the-art open-source and commercial solvers.

“REDUCER could be solved in under 20 minutes for Puerto Rico, while the commercial solver took between one and three hours, even for a small grid like Puerto Rico,” said Xu. “It would be basically impossible to extend that model to a larger, more complex grid.”

The Princeton team’s model especially shined when considering an energy system with high levels of rooftop solar, a key piece of Puerto Rico’s energy transition strategy. While REDUCER became more cost-effective in simulations with higher levels of rooftop solar, conventional day-ahead operation models were up to twice as expensive in high-renewables scenarios compared to the low-renewables baseline scenario.

“As climate extremes intensify and renewable energy adoption grows, tools like REDUCER are going to become increasingly important for informing real-time decision-making during extreme events to avoid large-scale impacts,” said Lin.

Planning for a future of climate extremes

Lin and Xu said that CRESCENT and REDUCER address challenges that grid operators will continue to face as climate change increases the frequency and severity of extreme weather events.

And while the team used Puerto Rico’s grid during Hurricane Fiona as a test case for their models, the researchers emphasized that the models could be tailored to a variety of grid configurations and climate extremes.

“Climate change is happening everywhere, not just Puerto Rico,” said Lin. She pointed to a 2024 perspective in Nature Reviews Electrical Engineering highlighting that from 2011 to 2021, the United States alone saw a 78% increase in weather-related power outages compared to the preceding decade.

“We hope that our work can help energy systems everywhere to adapt to the risks posed by climate extremes, whether they be hurricanes or other hazards,” Lin said.