Transformer-GNN fusion with gradient surgery for electricity price forecasting

Bingxiang Wu, Pengcheng Yang, Yiming He, Huayang Liang · 2026

Read Transformer-GNN fusion with gradient surgery for electricity price forecasting by Bingxiang Wu, Pengcheng Yang, Yiming He, Huayang Liang (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/PMC13518813.txt.

Description

Bingxiang Wu, Pengcheng Yang, Yiming He, and Huayang Liang propose a 2026 hybrid model for short-term electricity price forecasting. Good forecasts help markets run and grids stay reliable, yet prices depend on long time patterns plus network connections that most models handle separately. Training mixed data streams together can also create conflicting updates.

Their framework joins a Transformer encoder for time patterns, a graph neural network for network structure, and a gradient surgery method that calms training conflicts. Market records from PJM were fused with simulated sensor data from an IEEE 39-bus grid. Across five runs the fused model posted the lowest average error, with tighter uncertainty than the strongest rival baseline.

Catalog record

CREATORBingxiang Wu, Pengcheng Yang, Yiming He, Huayang Liang
YEAR2026
COLLECTIONPAPERS
FORMATTXT
FILE SIZE50914
LICENSEhttps://creativecommons.org/licenses/by/4.0
RIGHTSCC BY
SOURCEEurope PMC

Source: https://europepmc.org/article/MED/PMC13518813

Browse the full library · About the Tower of Babel