Electricity (Hourly)

Overview

Hourly electricity consumption (kW) of 321 clients from 2012 to 2014, as distributed by the Monash Time Series Forecasting Archive (Godahewa et al., 2021). It is the hourly aggregate (Lai et al., 2017) of the UCI ElectricityLoadDiagrams20112014 dataset. Every series has 26,304 points.

../../_images/electricity_hourly_teaser.png

The splits follow the standard forecasting protocol used by Chronos and GluonTS:

Test: the full series (321); the forecast target is the last PREDICTION_LENGTH = 24 points.
Train: the same series without those last 24 points.

ElectricityHourly.FREQ ("h"), PREDICTION_LENGTH (24) and SEASONALITY (24, used to scale MASE) describe the protocol.

Data Structure

Key	Type	Description
`series`	list of float32	Hourly consumption in kW
`item_id`	str	Client id from the Monash file (`T1` … `T321`)
`start`	str	ISO timestamp of the first point, as given by Monash (`2012-01-01T00:00:01`)

Usage Example

import numpy as np
from stable_datasets.timeseries import ElectricityHourly

test = ElectricityHourly(split="test")
H = ElectricityHourly.PREDICTION_LENGTH
y = np.asarray(test[0]["series"])
context, target = y[:-H], y[-H:]

References

Homepage: https://zenodo.org/records/4656140
Original data: https://archive.ics.uci.edu/dataset/321/electricityloaddiagrams20112014
License: CC BY 4.0

Citation

@inproceedings{godahewa2021monash,
  title={Monash Time Series Forecasting Archive},
  author={Godahewa, Rakshitha and Bergmeir, Christoph and Webb, Geoffrey I. and Hyndman, Rob J. and Montero-Manso, Pablo},
  booktitle={Neural Information Processing Systems Track on Datasets and Benchmarks},
  year={2021}
}