GitHub - zhouhaoyi/ETDataset: The Electricity Transformer dataset is collected to support the further investigation on the long sequence forecasting problem.

Abstraction: Two-year real-world electricity transformer dataset for long-sequence forecasting

Key points:

• Dataset from 2 to 69 electricity transformer stations in China, July 2016–July 2018
• Three variants: ETT-small (2 stations), ETT-large (39 stations), ETT-full (69 stations)
• Each ETT-small point recorded every minute; hourly variants also provided; 70,080 data points per series
• 8 features per point: date, oil temperature (target), and 6 external power-load features (HUFL, HULL, MUFL, MULL, LUFL, LULL)
• Data exhibits short-term daily periodicity (24h), long-term weekly pattern (7 days), and seasonal trends
• Associated with the Informer model (AAAI 2021 Best Paper Award); cite arXiv:2012.07436

Connections: Informer · Time Series Forecasting · Long Sequence Modeling

Source: https://github.com/zhouhaoyi/ETDataset