Decade of Digital Computing
Speeds Transformative Shift to Clean Energy

Sept. 18, 2024 | By Karen Petersen | Contact media relations

In 2013, a decarbonized U.S. energy system was a futuristic idea few in the industry saw as feasible for at least 30 more years.

The most comprehensive analysis of a high-renewable-based U.S. power system ever conducted at the time was the U.S. Department of Energy (DOE) National Renewable Energy Laboratory's (NREL's) Renewable Electricity Futures Study. Results showed that reaching 80% renewable electricity within three decades was economically and technically feasible—and solar and wind could account for a significant fraction of that.

Today the outlook is vastly different. The United States has set its sights on 100% carbon-pollution-free electricity by 2035. And NREL has acquired Kestrel, the third in a series of top-flight supercomputers that catalyzed the remarkable shift from a theoretical longshot to a tangible near-term target in the space of a single decade.

"High-performance computing enabled us to simulate extremely large systems at a fidelity that was unheard of at the time," said Ben Kroposki, director of NREL's Power Systems Engineering Center. "As a result, we are able to understand the impacts of variable renewables on the electric power grid to a level we never had before."

New Era of Clean Energy Innovation Takes Flight

A decade ago, NREL embarked on an epic expansion of its data center in a push to accelerate the energy transition by enabling big data and reducing market risk.

"We are on the threshold," declared Energy Secretary Ernest Moniz in September 2013 as he dedicated NREL's state-of-the-art Energy Systems Integration Facility (ESIF). Minutes earlier, he had pushed the button that powered up Peregrine, the first high-performance computing (HPC) system at the heart of ESIF and the engine that would drive the scientific advances needed to transform energy at a pace and scale few had dared imagine.

Moniz' remarks framed the inflection point. Solar costs had fallen 99%. Wind had edged out natural gas as the leading source of new electric generation capacity. The cost of LED bulbs had dropped to $15, driving the economics of home energy efficiency. And the East Coast was still reeling from Hurricane Sandy, underscoring the urgency of transitioning to a modern, resilient energy system.

A window of opportunity had opened, and with the new capabilities and computing power conferred by ESIF and Peregrine, NREL was poised to soar right through it.

When NREL acquired Peregrine in 2013, high energy costs and intensifying climate impacts had prompted a growing number of U.S. states and territories to adopt renewable portfolio standards aimed at stabilizing costs and building resilience. Hawaii had led the way in 2009, first aiming to meet 70% of the state's energy needs through energy efficiency and renewable energy by 2030 and then upping the ante six years later to 100% by 2045.

Hawaii relied heavily on NREL modeling and analysis to inform its road map. As the state and its utilities considered strategies for adding increasing levels of variable renewable energy generation onto small, geographically isolated island grids, NREL conducted solar and wind integration studies that provided critical insight into how various scenarios would impact grid operations and reliability.

By 2014, Hawaii had established itself as the nation's leading test bed for clean energy innovation. And data-driven insights from the Hawaii Clean Energy Initiative and other small-scale, community-based energy transitions were informing pathways for integrating clean energy technologies into the larger U.S. grid.

At NREL, researchers were leveraging advanced tools and techniques enabled by Peregrine to scale community-level lessons learned and address the technical and economic challenges of integrating renewables into the nation's grid. The laboratory's emerging HPC capabilities created opportunities to explore and de-risk potential solutions that were not practical or possible to experiment with in the real world.

"Supercomputing was transformative because it enabled us to simulate the entire North American grid with massive deployments of renewable technologies," Kroposki said. The first of these large-scale grid integration studies was the Eastern Renewable Generation Integration Study (ERGIS), completed in 2015.

Integration Studies Highlight Opportunities for a Low-Carbon Grid

ERGIS modeled four potential wind and solar photovoltaic (PV) futures and their operational impacts on the power grid of the Eastern United States—one of the largest power systems in the world. The results showed the Eastern Interconnection could accommodate more than 30% wind and solar PV.

"With Peregrine we could model five-minute time steps for entire years to look at how the electric power grid would react to the variability of wind and solar," Kroposki said.

As Peregrine shouldered the burgeoning computational demand of these highly complex simulations, the associated mathematical challenges spurred the development of new and better NREL tools and models to address wide-ranging energy transition issues and hurdles. The combination of advanced computing and modeling informed solutions and pathways at greatly accelerated rates.

NREL's North American Renewable Integration Study (NARIS) built on the findings of the ERGIS project. Launched in 2016, it was the largest grid integration study ever conducted. Its main focus was to inform grid planners, utilities, industry, policymakers, and other stakeholders about opportunities for cross-continental collaboration in transitioning to a low-carbon North American grid.

Mining consistent datasets from NREL's Renewable Energy Potential model, National Solar Radiation Database, and WIND Toolkit, the NARIS team used the R&D 100 Award-winning Regional Energy Deployment System and Distributed Generation Market Demand model, among other NREL-developed tools, to evaluate the cost, emissions, resource adequacy, and other impacts of multiple future scenarios.

Alongside these large-scale grid integration studies, researchers at ESIF were using NREL's growing suite of tools and models to address a wide range of regional and local energy planning needs, from techno-economic potential studies to renewable energy resource optimization to production cost modeling.

Advanced Computing Yields Clean Energy Breakthroughs

In addition to providing critical decision support for clean energy deployment, advanced computing was spurring basic scientific research and development at NREL, playing a pivotal role in materials discoveries and technological innovations across the laboratory.

By 2017, supercomputing had contributed to significant breakthroughs in clean energy systems and technologies ranging from wind and solar to bioenergy and hydrogen to battery energy storage and fuel cells.

• NREL created the Simulator fOr Wind Farm Applications (SOWFA) framework to investigate the effects of weather patterns, turbulence, and complex terrain on the performance of individual wind turbines and entire wind plants. By allowing users to simulate coordinated, multiturbine control of wind plants, SOWFA aimed to enhance understanding of wind farm physics, helping developers improve the performance and economics of wind turbines and wind farms.

• NREL used HPC and visualization to model the impact of higher levels of wind and solar PV generation on the U.S. power grid. ERGIS modeled the entire Eastern Interconnection, involving more than 5,600 electricity generators and more than 60,000 transmission lines.

• Researchers derived insights into how glycosylation—the natural attachment of sugars to proteins—affects a key cellulase enzyme. This work shed light on enzyme performance in breaking down biomass and plant waste for use in renewable fuels and products.

• Researchers have been studying the degradation mechanism of alkaline exchange membrane (AEM) fuel cells (FCs). AEM-FCs have potential to use much cheaper catalysts than conventional proton exchange membrane FCs. AEMs are the key component to only allow hydroxide ion (OH-) to transport across electrodes. However, AEMs will also react with OH- and cause the AEMs to degrade, which becomes the major challenge for this type of FC. By employing computational modeling, NREL researchers identified the main degradation pathway for AEMs. Based on that insight, researchers used in-silico design to predict several stable AEM structures, and experimental synthesis has confirmed that those AEM structures are significantly more stable.

• The NREL-led Advanced Computer-Aided Battery Engineering Consortium was critical to developing, enhancing, and validating sophisticated multiphysics simulation tools to accelerate the development cycle of batteries while improving their performance, safety, lifespan, and cost. The success of the Computer-Aided Engineering for Electric-Drive Vehicle Batteries project added adaptable battery simulation models to commercial software packages still in use by many battery developers and automakers. NREL researchers continue to leverage computer-aided engineering within the laboratory's robust multiscale modeling of battery physics research portfolio.

• In 2016, NREL's commercial vehicle researchers helped some of the world's largest companies—including FedEx, Coca Cola, Frito-Lay, and AT&T—deploy more than 450 fully electric delivery trucks across the United States. They used NREL's supercomputing resources to analyze more than 5 million miles of driving data and 1 million hours of electric vehicle charging data they collected in the field. These data formed the foundation of NREL's FleetREDI analysis platform, the world's only secure, public, and anonymized data analysis and insights pipeline for commercial vehicle decarbonization.

In the ensuing years, a procession of basic and materials science advances has helped pave the way for NREL to tackle increasingly complex energy transition challenges, with HPC driving innovative discovery. For example:

• Microstructure modeling of lithium-ion batteries expounded on the link between lithium-ion battery electrode microstructures and performance, leading to the open-source MATBOX toolbox, which supports microscale and macroscale electrochemical modeling.

• NREL analysis of solar thermochemical hydrogen production, the water-splitting technology that may produce hydrogen more efficiently, identified the potential of perovskite materials to play a role in renewable hydrogen production.

• Short-range order—the organized arrangement of molecules or atoms in disordered solutions—is important in fields like semiconductor physics. Researchers simulated perfect short-range order in a solid solution alloy phase, demonstrating quality electronic properties and signaling superior carrier transport. The findings may support the search for new materials, such as superconductors or thermoelectrics.

Peregrine-supported projects have also played a pivotal role in helping industry and partners clear hurdles to commercializing technology innovations, accelerating paths to market.

Increasingly, NREL's cleantech innovations, reliable datasets, game-changing modeling tools, and groundbreaking energy analysis helped inform the clean energy visions and goals of states, territories, Tribes, and communities. Los Angeles was on the leading edge, as the largest U.S. city to envision a 100% clean energy future.

LA Taps NREL To Explore Pathways to 100% as HPC Demands Mount

In 2017, the Los Angeles Department of Water and Power (LADWP) looked to NREL to lead the Los Angeles 100% Renewable Energy Study (LA100). LADWP sought to leverage NREL's decades of experience studying high-renewable power systems at various scales, from the Renewable Electricity Futures Study to ERGIS and NARIS.

"LA100 is a first-of-its-kind objective, rigorous, highly detailed, and science-based study to analyze potential pathways the LA community can take to achieve a 100% clean energy future," said Dr. Jaquelin Cochran, manager of NREL's Grid Systems Analysis Group and principal investigator of the LA100 Study.

The ambitious, multiyear project tapped into diverse capabilities across NREL's research programs, including:

• Detailed electricity demand modeling

• Power system investments and operations analysis

• Economic impact analysis

• Distribution grid modeling

• Life-cycle greenhouse gas emissions analysis.

As the research ramped up, demand for HPC exploded. Soon Peregrine, the world's largest supercomputer dedicated to energy efficiency and renewable energy, was oversubscribed.

Eagle Swoops in, Boosting Speed and Broadening Insights

Eagle landed at ESIF in mid-2018 and went online in 2019, five years after Peregrine. Equipped with the latest Intel Xeon processors and boasting a peak performance of 8.0 petaflops, NREL's next-generation Hewlett Packard Enterprise supercomputer could perform 8 million-billion calculations per second, more than tripling the laboratory's computing capability. The system also included a high-speed parallel centralized file system and data storage components.

With Eagle in the house, NREL researchers were able to run increasingly detailed models that simulated complex processes, systems, and phenomena at higher speeds, gaining crosscutting insights that informed energy transition pathways in LA and beyond.

"A power system transition of this scale benefits from complex analysis to provide deep insights for electrification, clean mobility, and power-sector decarbonization, coupled with implications for environmental justice, air quality, and economics," Cochran said. "To that end, the LA100 study charts a methodology that can be used to replicate, build upon, and scale up this type of analysis for other questions and jurisdictions."

Results from LA100 revealed LADWP could achieve its 100% renewable electricity goal by 2045—or even 2035—while maintaining reliable power for LA customers. By the time LA100 was released in 2021, hundreds of other jurisdictions, institutions, companies, utilities, and state and federal agencies had turned to NREL for data-driven clean energy decision support as they embarked on energy transitions of their own.

Dallas Fort Worth (DFW) International Airport was among them. DFW sought NREL's help exploring the potential of leading-edge transportation technologies to improve the efficiency and affordability of the energy used to move people and goods through one of the busiest airports in the world.

The initiative, dubbed Athena, drew upon NREL's expertise in computational science, vehicle technologies, and mobility systems to develop and run sophisticated models to inform long-term investments at DFW. Athena culled data from individuals, traffic, freight routes, flight schedules, autonomous vehicles, and other sources to develop a "digital twin" model of DFW.

Using data-driven statistical modeling and artificial intelligence (AI), DFW's digital twin simulated future scenarios and mobility options to help DFW and other transportation hubs understand the impacts of transformative technologies and optimize energy use and costs.

As energy transitions gained momentum, the need to solve more problems faster increased. By the end of 2021, almost 300 NREL projects were using Eagle to advance a range of research priorities, including:

• Materials science discoveries involving chemical and material compositions

• Computational fluid dynamics simulations for wind turbines

• Grid dynamics studies to advance grid modernization

• Traffic optimization efforts

• Electric vehicle and battery technology improvements.

Allocation requests in 2021 exceeded Eagle's capacity by more than 200%, and plans to replace it were already well underway, along with upgrades to the Insight Center.

Kestrel Brings a 100% Clean Energy Future Into Focus

When Eagle's replacement, named Kestrel, came online in early 2023, it delivered five times more computing capacity with approximately 44 petaflops of computing power. The new Hewlett Packard Enterprise system also features a bigger and better set of graphics processing units, further increasing capacity and accelerating the time to solution for models.

Like the small but mighty falcon it is named after, Kestrel is sharp-sighted, smart, and fast. Those qualities will be instrumental in NREL’s ability to leverage cutting-edge capabilities, from AI and machine learning to advanced data visualization, to build on the progress of the past decade and augment the impact of successive HPC investments.

Combined with the latest NREL-developed modeling innovations, such as Advanced Research on Integrated Energy Systems, Simulation and Emulation for Advanced Systems, and Sienna, these new HPC capabilities will enable NREL researchers to bring a 100% carbon-pollution-free future into focus with speed and precision.

"We now have unprecedented access to data information measurements coming from electric power grids," Kroposki said. "And we can use advanced machine learning techniques to extract salient points from these massive amounts of data, enhance how we model power systems, and learn responses to situations to help power system operators improve operations in real time."

Onward and Upward

Kroposki has seen NREL's grid integration work evolve from adding tiny amounts of wind and solar to small stand-alone systems to charting viable pathways to 100% renewable electricity.

"When I started, I'd get laughed out of rooms for even bringing up 2% wind and solar," he said. "And in the last decade, we've seen tremendous strides and utilities announcing 100% clean energy goals."

A decade ago, little in the existing energy landscape foreshadowed such a seismic shift, much less the meteoric rise of renewables. Coal supplied the lion's share of U.S. electricity. By contrast, solar and wind generated less than 3% combined. Even in the best-case scenario NREL had considered at that point, renewable energy might generate 80% of U.S. electricity by 2050.

Now the findings of the 2012 Renewable Energy Futures Study—which struck many industry experts as overly optimistic then—seem conservative in light of recent scientific advances and current market realities. Today, a quarter of U.S. electricity comes from renewable resources, primarily solar and wind. And coal's share has plummeted by more than half, a trend that is expected to continue.

We have crossed the threshold, and the journey continues. There are obstacles to overcome and decisions to be made—smart decisions driven by hard data. And time is of the essence.

As HPC systems continue to scale by orders of magnitude in the decade ahead, NREL researchers can look forward to adding even more complexity to their models and simulations, solving bigger problems faster, and increasing the momentum of the energy transition.

Learn more about computational research at NREL.