GE vernova advanced research new vision open 2024 award control number: 3387-1636 title: dynamic grid foundation model-based rapid decision support for fully dynamic grid this competitive…

The Department of Energy Advanced Research Projects Agency awarded GE Vernova Operations LLC a $4,356,581.55 cooperative agreement on July 17, 2026, under ARPA-E's VISION OPEN 2024 funding opportunity (CFDA 81.135) to develop a foundational machine learning model for rapid decision support in dynamic electric grid operations.

The award funds development of DynaGridFM, a groundbreaking foundational model that captures temporal dependencies, grid topology information, and known physical laws to address multiple grid operation tasks with a single machine learning model. The model will rapidly predict consequences of grid conditions and proactively identify control setpoints to alleviate current and anticipated violations. The work combines the foundational model with hardware-accelerated grid optimizers to verify and refine setpoints before human approval for hardware interface. A large language model layer will enable intuitive operator interaction with the foundational model to trigger desired analyses and provide decision support. The effort targets 100X acceleration for intelligent energy distribution using existing and new transmission grid infrastructure.

GE Vernova's Advanced Research Center in Niskayuna, New York performs the work under a period of performance extending from July 17, 2026, through July 16, 2028. The assistance type is cooperative agreement.

GE vernova advanced research new vision open 2024 award control number: 3387-1636 title: dynamic grid foundation model-based rapid decision support for fully dynamic grid this competitive cooperative agreement for a research and development project entitled, dynamic grid foundation model-based rapid decision support for fully dynamic grid is awarded recipient under arpa-e FOA number DE-FOA-0003387(vision open 2024 ) and control 3387-1636. The purpose of this award is a groundbreaking foundational model (FM) for grid operations that seamlessly captures temporal dependencies and grid topology information along with known physical laws. This FM will address various grid operation tasks with a common machine learning (ML) model to rapidly predict consequences as well as to proactively identify setpoints that can alleviate current and anticipated violations. To reliably orchestrate the grid and build user trust, this work introduces a novel combination of FM with hardware accelerated grid optimizers, where the setpoints will be rapidly verified and refined before human approval for interfacing with grid hardware. A large language model (LLM) will be used to enable an intuitive interaction between operator and FM, that can trigger desired analysis to provide decision support. Dynagridfm will provide 100X acceleration for intelligent energy distribution by utilizing existing and new transmission grid (t-grid) infrastructure.