NETL’s world-class artificial intelligence (AI) and machine learning (ML) capabilities are being leveraged to design the cleaner, more efficient power generation systems necessary for near-term decarbonization of the nation’s power sector and economy.

AI refers to machines that can, for a given set of human-defined objectives, learn, predict, and make decisions, only much faster and more efficiently than humans. Most AI applications use ML to find patterns in massive amounts of data. The patterns are then used for making predictions that have numerous applications across the energy landscape.

“A constant challenge of overhauling energy systems is the time and cost involved with running the physical, in-person experiments at power plants and other facilities,” said Chris Guenther, who leads advanced computing and AI work at NETL. “With enhanced AI/ML capabilities, NETL is developing models to accelerate and deploy new technologies through computational approaches that require supercomputer type resources. These approaches reduce much of the uncertainty and costs of maximizing the efficiency of our current power plant fleet and upgrading them to meet current market demands.”

NETL also has extensive capabilities in deep learning, which is a type of machine learning that must be “trained” on established information. Deep learning can then be used to predict useful things, such as how much energy a fuel cell will generate, or if a piece of equipment in a power plant could break.

Another AI-based technique under development at NETL for scientific use is computer vision. While most of the famous work in computer vision has been in photographic imagery, the same principles can also be used to look at scientific data such as microscopic images, X-ray scans, atomic maps and self-driving cars. While they can all give useful information about materials and technologies, the information can be difficult to extract.

Two computer vision tools being researched at NETL, are the convolutional neural network and generative adversarial network. Both are powerful means of getting useful information quickly from data and simulating physics and other uses that could be applied to energy infrastructure such as power plant monitoring, carbon storage projects and transformative power generation, among others.

The Wafer Scale Engine (WSE) is a revolutionary new hardware platform developed by Cerebras Systems Inc. NETL, collaborating with Cerebras Systems, has developed an intuitive application programming interface to solve scientific models on the WSE. NETL and Cerebras are also working together to train and deploy AI and AI hybrid models on the same platform. NETL and Cerebras have demonstrated that this new hardware/software platform can do scientific modeling several hundred to thousands of times faster and with several thousand times lower energy consumption than traditional distributed computing systems like the JOULE 2.0 supercomputer.

“We are continuing to investigate hybridizing this modeling approach with AI on a single device without moving data on or off the system,” Guenther said. “This will allow us to develop high-fidelity science-based modeling applications such as digital twins, cyber-physical security, digital ghosting, real time decision-making, and command and control, which can potentially see widespread adoption as the country undergoes its energy transformation. This is a great example of NETL collaborating with industry.”

NETL is a U.S. Department of Energy national laboratory that drives innovation and delivers technological solutions for an environmentally sustainable and prosperous energy future. By leveraging its world-class talent and research facilities, NETL is ensuring affordable, abundant and reliable energy that drives a robust economy and national security, while developing technologies to manage carbon across the full life cycle, enabling environmental sustainability for all Americans.