EnerGNN: A Graph Neural Network Library for Real-life Complex Energy Systems.
summary
EnerGNN is an open-source python package that models full-scale French transmission data with a hyper heterogeneous multigraph and GNNs built on JAX/Flax. It trains on a year of raw RTE data, reaches TRL 6 for contingency screening, and replaces ~10K simulations replaced by a single GNN forward pass to flag 22 true insecure contingencies, >9,400 secure ones and on average fewer than one missed dangerous case; tertiary voltage control cuts median violations from 37.5 to 7.5 and inference runs under a second; installable via pip install energn.
AI-written summary · about
chapters (10)
- 0:00Introduction to EnergyNN
- 1:04Motivation for AI in power systems
- 2:49Why graph neural networks
- 4:58Hyper heterogeneous multigraphs
- 5:40Package overview and modules
- 8:50Use cases at RTE
- 14:33Pipeline and data conversion
- 15:38Organization and governance
- 17:03Demo: Contingency screening
- 23:37Q&A session
related talks
31mGridFM - Towards Foundation Models for the Electric Grid
24mJet Energy Corrections with GNN Regression using Kubeflow @ CERN
16mPreview: Graph NNs
39mUsing Graph Neural Networks to Embrace the Dependency Within Your Data
23mUsing Graph Nets (GNs) to predict molecular properties
29mGraph Neural Networks + LLM on
01h 30mPractical graph neural networks in Python with TensorFlow and Spektral
16mGraph Network in High Energy Physics