powergraph-gnn-research

What problem does it solve?

Research teams need a reproducible, physics-informed, topology-aware GNN framework tailored to power grids, enabling PF/OPF and cascade analysis on the PowerGraph benchmark.

Core Features & Use Cases

• Data ingestion from the PowerGraph benchmark and conversion to PyG Data objects.
• Self-supervised pretraining (grid-specific SSL) to bootstrap representations for low-label regimes.
• Multi-task training with shared physics-guided encoder for PF, OPF, and cascade prediction, plus cascade explanation evaluation against ground-truth masks.
• Reproducibility anchors including blocked splits, config templates, and reference docs.

Quick Start

Run the end-to-end PowerGraph GNN research pipeline to pretrain the encoder and run downstream PF/OPF/cascade experiments with explanation evaluation.