MetaGrid
A topology-aware Transformer model that trains a single parameter set through cross-market meta-learning, enabling rapid adaptation to unseen power grids for electricity price forecasting. It explicitly models the grid graph structure to capture spatial dependencies.
Timeline (last 90 days)
MetaGrid is a topology-aware Transformer that, through cross-market meta-learning, enables a single parameter set to adapt to unseen power grids.
On leak-free benchmarks for five U.S. ISOs, MetaGrid achieved the lowest average MAE, improving 3.1% over the best baseline and 7.7% over training from scratch with the same architecture.
In the MetaGrid experiments, the five U.S. ISOs used the same inputs: electricity price, day-ahead load forecast, and calendar features.
All Facts (3)
MetaGrid is a topology-aware Transformer that, through cross-market meta-learning, enables a single parameter set to adapt to unseen power grids.
50%UnverifiedOn leak-free benchmarks for five U.S. ISOs, MetaGrid achieved the lowest average MAE, improving 3.1% over the best baseline and 7.7% over training from scratch with the same architecture.
50%UnverifiedIn the MetaGrid experiments, the five U.S. ISOs used the same inputs: electricity price, day-ahead load forecast, and calendar features.
50%