GridShedWatch: A Benchmark for Early Warning of Grid Load Sheds

GridShedWatch is an open benchmark that scores early warnings of US grid supply emergencies the way weather services score tornado warnings. It covers every firm load shed of at least 250 MW caused by a supply shortfall from 2020 to 2026 (8 events), with frozen hourly public data for 48 grids, 39 of which had no emergency, so false alarms are counted where nothing happened. Every method alerts at the same rate and is tested against a seasonal chance baseline. Of 11 reference methods, four (three learned forecasters and Hierarchical Temporal Memory) beat chance with data replayed as it happened. A control built only on the operators’ own day-ahead forecast catches all 8 load sheds but also alerts on equally cold grids that coped, and with a day’s real-time data delay no method is shown to beat chance. Telling a coming shortfall from a cold week is the open problem the benchmark measures. Code, data and results: https://github.com/gotham29/gridshedwatch

Authors

Publication Details

Journal
Zenodo (CERN European Organization for Nuclear Research)
Published
2026-11-16
DOI
https://doi.org/10.5281/zenodo.24288272
Primary Topic
Energy Load and Power Forecasting
Type
preprint
Controls
|||
ALL TIME
JAN
FEB
MAR
APR
MAY
JUN
JUL
AUG
SEP
OCT
NOV
preprint

GridShedWatch: A Benchmark for Early Warning of Grid Load Sheds

Samuel Heiserman
4 links
Zenodo (CERN European Organization for Nuclear Research)
Energy Load and Power Forecasting
preprint

GridShedWatch: A Benchmark for Early Warning of Grid Load Sheds

Samuel Heiserman
preprint en

Abstract

GridShedWatch is an open benchmark that scores early warnings of US grid supply emergencies the way weather services score tornado warnings. It covers every firm load shed of at least 250 MW caused by a supply shortfall from 2020 to 2026 (8 events), with frozen hourly public data for 48 grids, 39 of which had no emergency, so false alarms are counted where nothing happened. Every method alerts at the same rate and is tested against a seasonal chance baseline. Of 11 reference methods, four (three learned forecasters and Hierarchical Temporal Memory) beat chance with data replayed as it happened. A control built only on the operators’ own day-ahead forecast catches all 8 load sheds but also alerts on equally cold grids that coped, and with a day’s real-time data delay no method is shown to beat chance. Telling a coming shortfall from a cold week is the open problem the benchmark measures. Code, data and results: https://github.com/gotham29/gridshedwatch

Zenodo (CERN European Organization for Nuclear Research)
Affordable and clean energy
Energy Load and Power Forecasting
AI Navigator

Ask Laika to Summarize, Analyze, and Connect papers live on the map.

Summarize Papers & Methodologies

Extract key findings, datasets, and comparative methods across publications.

Benchmark Rankings & Visual Analytics

Rank top research institutions, authors, funders, topics, and journals by Field-Weighted Citation Impact (FWCI) and paper volume with instant charts.

Connect Distant Disciplines

Bridge topological clusters on the map to find hidden collaborative intersections.

Laika access coming soon