GridSFM is an open-source framework for AC Optimal Power Flow (AC-OPF) optimization. It is designed to determine the cost-minimizing generator dispatch while satisfying all physical and operational constraints of a power grid. The project includes a data pipeline that transforms grid topologies into solved scenarios using Hugging Face loaders, as well as a neural surrogate model capable of performing fast AC-OPF inference on those scenarios.
Key features and components:
- Neural surrogate models for rapid AC-OPF inference
- A Python-based data pipeline to convert topologies into .pyg.json scenarios
- A browser-based viewer for inspecting network topology and OPF results
- A Julia-driven pipeline that processes raw topologies through controlled perturbations such as load scaling, generator outages, and voltage squeezing
Key features and components:
- Neural surrogate models for rapid AC-OPF inference
- A Python-based data pipeline to convert topologies into .pyg.json scenarios
- A browser-based viewer for inspecting network topology and OPF results
- A Julia-driven pipeline that processes raw topologies through controlled perturbations such as load scaling, generator outages, and voltage squeezing
Microsoft Research has released an open dataset providing approximate transmission topology of the U.S. power grid, constructed entirely from publicly available data such as OpenStreetMap and EIA statistics. This work addresses the difficulty researchers face when accessing critical infrastructure information by providing geographically grounded and electrically coherent models that support AC optimal power flow (AC-OPF) analysis. The pipeline allows for large-scale studies ranging from individual states to the entire Eastern Interconnection, enabling physics-based investigations into grid resilience and capacity without relying on proprietary datasets.
