Wind Power Forecasting Data from Four Geographic Sites

5.5 KB of data from four wind farm locations used to benchmark machine learning models for power forecasting. Author Usman Ali published the dataset on figshare in 2026, comparing XGBoost, Random Forest, and Support Vector Regression algorithms. The study reports high accuracy, with R² values of 0.99 for top-performing models across most sites.

Use Cases

• Benchmarking regression algorithms for wind power prediction based on the described model comparison.
• Training forecasting models using features from geographically diverse wind farms mentioned in the study.
• Analyzing the impact of outlier removal methods (Z-score, IQR) on model performance as described.
• Comparing the effectiveness of different SVR kernels (linear, RBF, polynomial) for time-series energy data.

Strengths

• Dataset is openly licensed under CC-BY-4.0.
• Model performance metrics (R² up to 0.99, MAE from 11.10 to 15.94 for XGBoost) are reported, indicating high predictive accuracy.
• Data covers four distinct geographical locations, suggesting some diversity in conditions.

Limitations

• The dataset is very small at 5.5 KB, indicating limited scope.
• Column-level documentation is absent; field semantics must be inferred after download.
• Row count is unknown, which may limit suitability assessment.

Provenance

Source

figshare

Collection Method

Likely collected from operational wind farms, as the description references data from four geographical locations.

Time Range

null

Freshness

Last updated 2026-04-30 17:54:44; freshness should be verified.

Geography

Covers four distinct geographical locations, but specific sites are not named.