PowerModelsGAT-AI: Physics-Informed Graph Attention for Multi-System Power Flow with Continual Learning

CHIDOZIE EZEAKUNNE, JOSE E. TABAREZ, REEJU POKHAREL, ANUP PANDEY · 2026

Read PowerModelsGAT-AI: Physics-Informed Graph Attention for Multi-System Power Flow with Continual Learning by CHIDOZIE EZEAKUNNE, JOSE E. TABAREZ, REEJU POKHAREL, ANUP PANDEY (2026) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2603-16879.txt.

Description

Ezeakunne and co-authors present PowerModelsGAT-AI, a graph attention network that solves power flow for many grid systems at once. Classic Newton solvers can be slow under stress, and most learned models train on one system only. Their model predicts bus voltages and generator output, with masking that respects bus types. A physics based loss penalizes power mismatch. Trained on 13 benchmark systems under double outage conditions, it reached a mean voltage error of 0.89 percent. Continual learning tests kept new system errors below 2 percent, where plain fine tuning forgot badly.

Catalog record

CREATORCHIDOZIE EZEAKUNNE, JOSE E. TABAREZ, REEJU POKHAREL, ANUP PANDEY
YEAR2026
COLLECTIONPAPERS
FORMATTXT
FILE SIZE87383
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2603.16879

Browse the full library · About the Tower of Babel