Mycelium model boosts electrical grid task performance across networks
Mycelium: A Generalizable Cross-Grid Multi-Task Model for Electrical Distribution Systems
Machine Learning
Summary
Electrical grids are complicated networks that need careful monitoring to work properly. The paper’s authors created a new kind of model called Mycelium that can understand many different electrical grid tasks at once, even on networks it hasn’t seen before. They built a way to represent grids as graphs and trained their model with both real and simulated grid data. Mycelium learns from the physics behind how electricity works, helping it perform better than other models tailored to single tasks.
What this means in practice
- •For utility grid operators: Improve monitoring and fault detection for diverse electrical distribution networks using a single model that generalizes across grids.
- •For smart grid system developers: Develop integrated software that performs multiple grid analysis tasks efficiently by embedding physical and topological grid knowledge.
Authors
Zhengyang Wei, Shourya Bose, Helgi Hilmarsson, Elena Carnio, Dhruv Suri
Abstract
Electrical distribution grid operations require inference across heterogeneous networks from sparse, noisy, and incomplete time series measurements. In this work, we identify challenges and explore solutions towards a unified model that can perform diverse tasks grounded in the physics of the electric grid and generalize to unseen distribution networks. We define a unified grid ontology that represents variable sized distribution networks as heterogeneous graphs while preserving native topology, asset types, and electrical relationships across networks. We develop a physics based data simulation pipeline that combines reference and procedurally generated distribution networks with network reconfigurations, fault scenarios, and configurable sensing conditions. We present Mycelium, a heterogeneous graph transformer with structure aware communication edges and electrical reference features that encode network position and nominal phase orientation, together with task specific temporal readouts which generate per task outputs. We train Mycelium on reference as well as synthetic grids, and study its generalization on benchmark networks completely excluded from training and validation. Mycelium is observed to outperform task specific neural baselines on most reported benchmark metrics. Architectural ablations and the aforementioned studies reveal Mycelium's capability to learn representations of the underlying physics which serves to enhance cross-task performance, thereby addressing a significant challenge in unified grid models.