Do topology-aware grid foundation models survive feeder reconfiguration? SimBench MV study: the topology encoding is what transfers (2-8% voltage error on unseen feeders, flat across switching states) — with zero-shot, DC-baseline, and scale gaps named plainly.
topology pytorch power-systems smart-grid power-flow pandapower network-reconfiguration graph-neural-networks self-supervised-learning pytorch-geometric positional-encoding simbench distribution-grid foundation-models
-
Updated
Jul 21, 2026 - Jupyter Notebook