Consortium to Develop Open AI Models for the Power Grid

US energy research group EPRI has taken a unique approach to address the power challenges of AI datacentres by setting up the Open Power AI Consortium. This consortium is focused on developing open AI models that can help utilities streamline operations, boost energy efficiency, and improve grid resiliency. The models are being trained on massive libraries of proprietary energy and electrical engineering data from EPRI, with the aim of driving innovation and operational excellence in the industry.

The first version of an open AI model for electric and power systems has already been developed using Nvidia’s H100 GPUs and is expected to be available soon in early access as a NIM microservice. This development marks a significant step forward in leveraging advanced AI tools to tackle industry challenges and position the sector at the forefront of innovation.

As part of its initiatives, the Open Power AI Consortium will create a sandbox environment to develop and validate AI applications in collaboration with startups, academia, national labs, utilities, and technology companies. This collaborative approach is aimed at fostering a culture of innovation and driving the transformation towards a more resilient and affordable energy future.

• Power shortages hold back datacentres
• AI to drive 160% increase in datacentre power
• Selling shovels in the AI goldrush

“Over the next decade, AI has the great potential to revolutionize the power sector by enhancing grid reliability, optimizing asset performance, and enabling more efficient energy management,” said Arshad Mansoor, EPRI’s president and CEO. The consortium, with its diverse membership including energy companies, technology firms, and researchers, is well-positioned to lead this transformation and drive innovation in the industry.

The Executive Advisory Committee of the Open Power AI Consortium comprises executives from over 20 energy companies and leading tech firms such as Duke Energy, Exelon, AWS, Oracle, and Microsoft. The consortium plans to expand its global member base to further enhance collaboration and knowledge sharing in the energy sector.

Global energy consumption is on the rise, with AI datacentres playing a significant role in driving this growth. To meet the increasing demand for electricity, providers are looking to enhance the resiliency of power infrastructure, balance energy sources, and expand grid capacity. AI models developed by the consortium can play a crucial role in enabling utility and energy companies to assess energy needs more efficiently and prepare necessary studies and permits for infrastructure improvements.

By introducing AI agents trained on sector-specific documents, the consortium aims to accelerate processes such as interconnection studies and the preparation of licenses, permits, and utility rate cases. These AI solutions tailored to the unique needs of the power sector have the potential to significantly reduce timelines and improve decision-making in critical areas of energy infrastructure development.

Furthermore, the Open Power AI Consortium is not only focused on developing models and datasets but also on establishing a standardized framework of benchmarks. These benchmarks will help utilities, researchers, and other stakeholders in the energy sector evaluate the performance and reliability of AI technologies, fostering a culture of continuous improvement and innovation in the industry.