The Department of Energy has entered the artificial intelligence fray, publishing a series of reports and announcements on the technology’s use in the energy transition this week.
- The top line: On Monday, DOE made good on an executive order issued six months ago, which compelled several federal agencies to conduct research into the development and security implications of AI for their work. Acknowledging both the potential and the risks of using the technology, the department launched a website showcasing the tools and resources it has already set up; and DOE also shared plans for collaboration with external stakeholders, ostensibly to discuss the AI and load growth conundrum.
- Market grounding: The executive order issued in October 2023 was wide in scope, and prompted DOE to assess, alongside the Department of Homeland Security, the threats that AI could pose to the safety of critical infrastructure like the power grid. But it also catalyzed research, including by the National Labs housed within DOE, that involves developing DOE’s computing capabilities and creating AI testbeds for clean energy deployment. This comes as many industries, including the utility sector, begin to explore using the AI — though research from Latitude Intelligence suggests that most utilities don’t yet have a coordinated approach to integrating the technology into their work.
The department’s new website will showcase and house the AI tools and foundation it is developing. These include, for instance, WildfireGPT, a large language model agent hosted by the Argonne National Laboratory that delivers insights on wildfire risk for engineers, infrastructure operators, and others.
DOE also announced its investment of $13 million in an initiative specifically to develop software to better the siting and permitting process, dubbed VoltAIc. The initiative aims to offer AI as a resource to streamline siting and permitting not only at the federal level, but also at the state and local levels, where DOE has limited power.
The department has already partnered with the Pacific Northwest National Laboratory to develop PolicyAI, a policy-specific LLM test bed for developing software to better reviews under the National Environmental Policy Act and similar processes. (This came just a day before the Biden administration finalized new rules to speed up environmental reviews for clean energy projects, and less than a week after the administration unveiled rules to reform the federal permitting process, with a view toward easing the country’s transmission backlog.)
The research prompted by the executive order yielded at least four reports, out this week. These include a “first-ever” view on how AI could be used to modernize the grid in the near-term, as well as a framework rooted in research from the National Laboratories on long-term opportunities for using the technology in areas such as nuclear power, carbon management, and energy storage over the next decade.
As directed by the executive order, the department’s Office of Cybersecurity, Energy Security, and Emergency Response collaborated with the Lawrence Livermore National Laboratory to publish an evaluation of potential benefits and risks of AI for critical energy infrastructure.
And finally, the Office of Electricity and the Pacific Northwest National Laboratory co-published a look at specific applications for machine learning and artificial intelligence in the power system.
Overall, the reports point to AI’s unique processing capabilities, which they say should allow us to disentangle the research, development, and operational challenges that would otherwise take years to address.
Going forward
However, some of the department’s most influential work on AI may be yet to come.
In light of the huge load growth that is already coming as a result of data centers devouring more energy than ever (largely due to AI itself), as well as electrification, and domestic manufacturing, DOE seems to be planning a discussion group of sorts. According to Monday’s announcement, in the coming months the department will bring together utilities, developers, data center owners, and regulators in places experiencing particularly large load growth.
And the department will also separately convene experts and energy stakeholders to assess further the threats that AI could pose to the grid — whether by “unintentional failure, intentional compromise, or malicious use” — and the technology’s potential to improve grid resilience.
Meanwhile, the Energy Advisory Board chartered a working group on powering AI and data center infrastructure, which is expected to make recommendations on how to meet the energy demand for AI by June.
