U.S. Supercomputer to Support Renewable Energy Innovation Coming in 2023

A common kestrel, which the Energy Department’s National Renewable Energy Laboratory newest supercomputer, is named after

A common kestrel, which the Energy Department’s National Renewable Energy Laboratory newest supercomputer, is named after Andyworks/istockphoto

HPE will build Kestrel, named for the American falcon, for an Energy Department lab.

The Energy Department’s National Renewable Energy Laboratory, or NREL, is set to gain a powerful new supercomputer that will support huge, data-intensive workloads and necessary research and development to help the nation prepare for future clean energy-aligned needs.

Anticipated to operate at approximately 44 petaflops, that fresh machine—named Kestrel—will boast more than five times greater performance potential than NREL’s existing system.

“Based on the recently published top500 list, Kestrel is expected to be among the top 10 computers in the U.S. when fully deployed,” NREL Laboratory Program Manager for Advanced Computing Dr. Kristin Munch told Nextgov in an email.

Previous generations of the lab’s high performance computing systems, Peregrine and Eagle, were named after birds. Sticking to that theme, Kestrel’s name comes from that of the American falcon. Once it’s fully built, likely in 2023, the supercomputer will be hosted at NREL’s Energy Systems Integration Facility data center in Golden, Colorado.

“We can’t share the funding value of the contract,” Munch explained. “The Kestrel procurement was a publicly available solicitation through SAM.gov and the subcontract was awarded as firm fixed price for the design, build, delivery and on-going maintenance of the system.”

Hewlett Packard Enterprise is the contractor the department tapped to produce Kestrel, which will be built using the HPE Cray EX next-generation supercomputing platform. Among other components, the system will include Intel processors and Nvidia graphics processing units to accelerate artificial intelligence applications. 

Munch noted that “NREL’s advanced computing influence spans several common themes across the Office of Energy Efficiency and Renewable Energy,” such as “materials discovery, process modeling, fluid dynamics, resource mapping, and analysis of large-scale systems with real-time optimization.” 

Kestrel will eventually allow for new and advanced modeling, simulation, AI and analytics capabilities around those areas, and it’ll boost research across the lab’s portfolio.

For instance, NREL officials are leading a range of AI studies. Some are motivated by autonomous vehicle applications, or systems for machine-guided inverse design. Now, “Kestrel’s GPU-accelerated capability,” according to Munch, “will enable new directions in AI research for energy innovation.”

NREL also hosts the Advanced Research on Integrated Energy Systems platform. Deemed ARIES for short, that tool helps federal experts understand the impact and get the most value from the millions of new devices—like those for electric vehicles, renewable generation, hydrogen, energy storage, and grid-interactive efficient buildings—that are being connected to the grid daily.

“High performance computing, and the new Kestrel computer, will amplify the scale of the research capabilities of ARIES by enabling a broad spectrum of modeling and simulation tools for distributed generation, transmission, distribution, capacity expansion, power flow dynamics, and grid load forecasting and modeling,” Munch said. 

Further, the lab conducts a variety of jurisdictional planning studies that the program manager said wrestle with the challenges of melding the plethora of clean energy technologies now available with aspirations to reduce their carbon footprints. 

“These jurisdictional analyses aimed at clean electricity for most of the population requires new computing capabilities to scale these approaches to state, regional, and national levels,” she noted. “Getting to a net zero economy will additionally require more robust analysis with rigorous treatment of cyber, climate resilience, environmental justice, and economy-wide modeling, all of which requires an exponential increase in computational intensity and in the amounts of data to be analyzed.”