Datasets:
dataset_info:
features:
- name: timestamp
dtype: timestamp[us]
- name: id
sequence: string
- name: past_values
sequence:
sequence: float32
- name: past_observed_mask
sequence:
sequence: bool
- name: static_categorical_values
sequence: float32
- name: static_real_values
sequence: float64
- name: future_values
sequence:
sequence: float32
- name: future_observed_mask
sequence:
sequence: bool
splits:
- name: train
num_bytes: 61803996485
num_examples: 8306132
- name: validation
num_bytes: 8813372653
num_examples: 1184471
- name: test
num_bytes: 17811677540
num_examples: 2393795
download_size: 2106288708
dataset_size: 88429046678
configs:
- config_name: default
data_files:
- split: train
path: data/train-*
- split: validation
path: data/validation-*
- split: test
path: data/test-*
license: cc
task_categories:
- time-series-forecasting
pretty_name: ONS Capacity Factor Dataset
Dataset Card for ONS-Capacity-Factor-Dataset-Full
Dataset Summary
The ONS-Capacity-Factor-Dataset-Full provides hourly data for wind and solar power plants in Brazil. These values are published by the Operador Nacional do Sistema Elétrico (ONS) — the Brazilian National Electric System Operator — which is responsible for coordinating and controlling electricity generation and transmission in the National Interconnected System (SIN).
The dataset includes data from 2009 to 2025, with files provided yearly from 2009 to 2021 and monthly from 2022 onwards. Each entry represents the capacity factor of a plant or group of plants, calculated as the ratio between the actual generation and the installed capacity.
This dataset is useful for research in renewable energy forecasting, time series analysis, energy systems modeling, and applications of artificial intelligence in the power sector.
Supported Tasks and Benchmarks
- Time Series Forecasting: Predict future capacity factor values for wind and solar energy.
- Renewable Energy Modeling: Analyze the performance of renewable plants over time.
- ML for Energy Systems: Train and evaluate machine learning models for grid optimization and resource planning.
Languages
The dataset contains metadata and headers in Portuguese, as provided by ONS. However, variable names are intuitive and can be easily adapted to English in preprocessing.
Dataset Structure
Data Fields
Each file includes the following fields (translated from Portuguese):
timestamp: Start of the time series (YYYY-MM-DD HH:MM).id: Id of the time series.past_values: The context values of shape(512, 1).past_observed mask: The mask of context values of shape(512, 1).static_categorical_values: If it is wind or solar energy(0 or 1).static_real_values: Latitude and longitude.future_values: The target sequence.future_observerd_mask: The mask of target sequence.
Note: Some years or months may lack some of these fields depending on the ONS publication format.
File Formats
- Parquet (
.parquet)
Data Splits
There are no predefined train/test/validation splits. Users are encouraged to split the data according to their application (e.g., temporal split for forecasting).
Licensing Information
This dataset is licensed under the Creative Commons Attribution License (CC-BY 4.0). You are free to share and adapt the data, provided that appropriate credit is given to ONS.
More information: http://opendefinition.org/od/2.1/pt-br
Source
- Website: ONS Historical Operation Data
- Maintainer: ONS (Operador Nacional do Sistema Elétrico)
- Contact: [email protected]
Citation
If you use this dataset in your research, please cite ONS as the data provider:
@misc{
ons_capacity_factor,
title = {ONS Capacity Factor Dataset},
author = {Operador Nacional do Sistema Elétrico (ONS)},
year = {2025},
url = {http://www.ons.org.br}
}
Acknowledgements
This dataset is made available through the ONS Open Data Initiative and the AWS Open Data Sponsorship Program: https://registry.opendata.aws/ons-opendata-portal/