IBM's Fastest Supercomputer Will Be Used To Find Better Ways To Produce Green Electricity
Energy giant General Electric will be using one of the world's most powerful supercomputers, IBM's Summit, to run two new research projects that could boost the production of cleaner power.
Energy giant General Electric (GE) will be using one of the world's most powerful supercomputers, IBM's Summit, to run two new research projects that could boost the production of cleaner power.
Last month, the US Department of Energy (DOE), which hosts Summit in Oak Ridge National Laboratory, awarded a total of more than 7 million node hours on the supercomputer to 20 research teams, two of which belong to GE Research.
The Summit supercomputing system is the second most powerful in the world, behind the Fugaku supercomputer located in Japan. Built by IBM, Summit boasts system power equivalent to 70 million iPhone 11s, which scientists can leverage to run large computations such as simulating systems' behavior or solving complex physics problems.
GE has now lifted the veil on the two projects that were selected by the DOE to run on Summit, and they will both address sticking points in the generation of renewable energy.
One team, led by GE researcher Jing Li, received 240,000 node hours to advance research in the field of offshore wind power. Using the Summit supercomputer, Li hopes to be able to run complex simulations to study new ways of controlling and operating offshore turbines to best optimize wind production.
In particular, Li's team will be looking at a wind phenomenon known as coastal low-level jets, which occur along many coastlines and can affect the performance and reliability of offshore wind turbines. Thanks to high-fidelity computational models, the researchers will simulate interactions between wind farms and coastal low-level jets, to inform future, more-efficient designs for the farms.
The findings will also be used to guide the DOE's ExaWind project, which is designed to accelerate the US's deployment of onshore and offshore wind plants.
Doing so requires a precise understanding of the ways that natural wind phenomena interact with the built infrastructure. Simulating these interactions, however, comes at a large computational cost because of the many factors at play. Most research projects are currently only able to predict the behavior of a small number of turbines.
The ExaWind project is aiming to generate predictive simulations of wind farms with tens of megawatt-scale wind turbines dispersed over an area of many square kilometers with complex terrain—a computation that could involve simulations with up to 100 billion grid points.
The huge computing power that has been granted to Li's team with Summit, therefore, is a promising step toward achieving the ExaWind challenge.
GE researcher Michal Osusky was also awarded another 256,000 node hours on Summit for a separate research project that focuses on applying machine-learning methods to improve the design of physical machines such as jet engines or power-generation turbines.
Combining machine learning and simulation, Osusky's team could mimic real-world engines quickly and run virtual tests to verify designs faster than with conventional means.
"These simulations would provide unprecedented insight into what's happening in these complex machines, way beyond what is possible through today's experimental tests," said Osusky. "The hope is we can utilize a platform like this to accelerate the discovery and validation process for cleaner, more-efficient engine designs that further promote our decarbonization goals."