kevykibbz's picture
Create README.md
5aacd1a verified

Wind Turbine SCADA Data For Early Fault Detection

About the Dataset

This dataset, originally published as "CARE to Compare: Wind Turbine Anomaly Detection Dataset," contains real-world SCADA data from wind turbines. It is designed for testing and developing anomaly detection algorithms for wind energy systems.

Dataset Overview

  • Duration: 89 years of cumulative operating data
  • Turbines: 36 wind turbines across 3 wind farms
  • Datasets: 95 total datasets
    • 44 contain labeled anomaly events
    • 51 represent normal behavior
  • Resolution: 10-minute SCADA data intervals
  • Feature Count:
    • Wind Farm A: 86 features
    • Wind Farm B: 257 features
    • Wind Farm C: 957 features
  • Data Structure: Each dataset contains 1 year of training data plus 4-98 days of prediction data

Key Features

  • Detailed fault information: Most comprehensive publicly available wind turbine fault dataset
  • Status-based labels: Ensures high-quality training data
  • Labeled anomaly events: 44 labeled time frames of faults
  • Normal operation datasets: 51 time series with no faults
  • Comprehensive sensor descriptions: Includes units and descriptions for all measured variables

Potential Applications

  • Anomaly detection algorithm development and testing
  • Predictive maintenance modeling
  • Wind turbine performance analysis
  • SCADA-based condition monitoring
  • Benchmark testing using the CARE score (Coverage, Accuracy, Reliability, and Earliness)

Citation

If you use this dataset, please cite the original paper:

Güuck, C., Roelofs, C.M.A., & Faulstich, S. (2024). CARE to Compare: A real-world dataset for anomaly detection in wind turbine data. arXiv:2404.10320

Original Source

This dataset was originally published at:

Modification Note

The only modification made to the original dataset is the replacement of semicolons with commas as CSV separators to facilitate easier data loading and analysis.

License

This dataset is shared under the Creative Commons Attribution-ShareAlike 4.0 International (CC BY-SA 4.0) license.