DOE’s NETL seeks opportunity to augment AI/ML tools in the energy workspace

The Energy Department’s National Energy Technology Laboratory is seizing the opportunity to integrate artificial intelligence and machine learning tools.

Derace Lauderdale

March 19, 2024 2:18 pm

4 min read

Download audio

The Energy Department’s National Energy Technology Laboratory is seizing the opportunity to integrate artificial intelligence and machine learning tools into its activities to overcome challenges related to carbon management and new energy resources. NETL has numerous other AI/ML projects in process, and climate resiliency is just one of the earliest opportunities.

“There’s a lot of exciting opportunity for AI and machine learning in the applied energy space. We have quite a large portfolio of this type of work here at NETL already aligning to the core competencies and application spaces that I enumerated, but in particular for things like advanced materials, innovating and identifying opportunities for materials to address infrastructure resiliency, new materials needs for things like carbon capture and dealing with climate and mission related activities. So understanding atmospheric modeling, understanding climate processes, understanding emissions sources and how to mitigate those and predict them and prevent potential deleterious impacts from those types of things. There’s opportunity for artificial intelligence and machine learning to play a role in any number of activities that we’re involved in,” Kelly Rose, senior fellow for computational science and engineering at the National Energy Technology Lab said.

When it comes to staring an AI project, agencies need to gather enough data to begin creating and build a team with IT and data experts to execute their plan. Nearly three years ago, NETL launched the Science-based AI/ML Institute (SAMI) with the focus to be more tactical and strategic in its AI/ML implementation. With so many tools in the AI/ML toolkit it can be confusing to learn. The SAMI institute crosscuts NETL’s teams and directorates as a way to help people train and learn the toolkit.

“Research teams that are already doing this type of activity are showing how they’ve gotten there. And we also have a tech team, a consulting advisory group within the institute that research teams across the org can reach out to and say, ‘hey, can we talk to you about how to get started?’ Or ‘we’re interested in this specific thing in the AI/ML space, how would we responsibly pursue this? Does it make sense?’” Rose said on Federal Monthly Insights – Operationalizing AI. “We’re on this path here at NETL and there’s a number of really exciting examples of how that has been already paying off in benefiting researchers here at the lab and with our collaborators, industry regulators and other R&D entities. It’s exciting to see it mirrored with the guidance that’s coming out with the AI executive order and other guidance coming from the federal government on large.”

NETL currently performs with a software suite called Multiphase Flow with Interphase Exchanges (MFiX), a multiphase fluid flow model used for designing next-generation energy systems. NETL’s MFiX software has also been used for troubleshooting multiphase flow reactors for fossil, bio, nuclear and solar energy processing and nuclear waste treatment.

“There’s a large collaborative consortium that NETL has helped nurture from more conventional advanced computing, high performance computing (HPC) modes. But for the last ten years, the researchers here at NETL have been integrating AI machine learning-based methods into the capabilities of MFIX for applications, like [Computational Fluid Dynamics (CFD)] modeling for power plant generation, optimization and forecasting analysis of how to help conventional coal plant power plants convert to natural gas combustion or even convert to alternative energy uses,” Rose said.

NETL also has one of their sub teams collaborating with tech sectors to develop a cutting-edge AI computer model that will be more sufficient than the supercomputers currently in use. The goal is to use a wafer scale, engine-based approach to speed up the process of solving engineering problems 470 times over that of a traditional high-performance computer using less energy.

“The other piece of a specific example of what the subset of this group has been doing is they’re actually looking at cutting edge advanced AI computing and modeling to accelerate simulations. So the more fundamental computational modeling element, the actual computing effort itself. As supercomputers have continued to scale and grow bigger and bigger and bigger to keep up with this demand, to scale compute not just for AI, but for advanced simulation needs as well in the high performance computing domain, there’s only so much we can scale towards. It’s putting demands on how these systems operate,” Rose told the Federal Drive with Tom Temin.

With AI/ML advancing, NETL is doing everything they can to recognize and have assurance with these complex systems, and help people understand that it is reliable.

“There’s people that are always excited about change, there’s people that are curious about change, and there’s people that resist change because they don’t understand it or they like the way things are. That’s human nature,” Rose said. “We refine on things and help improve the state of knowledge or state of technology. But we also are up for scrutiny. And as new information emerges or as new perspectives come into the play, we can improve, or we can help people understand why this is actually trustworthy. It’s not going to be perfect, but the ability to detect and innovate at scales we’ve never had the opportunity to do before. That’s what’s exciting.”