GENCO: A Unified Neural Solver for Power Grid Analysis
GENCO introduces a unified neural solver designed for power flow, optimal power flow, and state estimation, employing a single architecture with a shared network representation. The open-source GridFM Development Framework facilitates low-code synthetic data generation and training, releasing extensive datasets comprising millions of power flow (PF) and optimal power flow (OPF) scenarios. Evaluated on PFDelta and OPFData benchmarks, this work emphasizes the scarcity of foundation models in engineering fields, particularly in power systems. The findings are available on arXiv under the identifier 2608.09921, aiming to streamline data generation and training processes for neural solvers.
Key facts
- GENCO is a unified neural solver for power flow, optimal power flow, and state estimation.
- It uses a single architecture and shared network representation.
- The GridFM Development Framework is open-source and supports low-code synthetic data generation and training.
- Large-scale datasets with millions of PF and OPF scenarios are released.
- GENCO was evaluated on PFDelta and OPFData benchmarks.
- The work is published on arXiv with identifier 2608.09921.
- The paper highlights the lack of foundation models in engineering domains like power systems.
- The framework aims to standardize data generation and training for neural solvers.
Entities
Institutions
- arXiv