Wind Power Forecasts from Four Geographic Sites, 5.5 KB

Four datasets from different geographical locations used to analyze machine learning algorithms for wind power prediction. The data was processed by author Usman Ali, with outliers removed using Z-score and IQR methods, and last updated on April 30, 2026. It is a small dataset (5.5 KB) stored in an XLS file format.

Use Cases

• Benchmarking regression algorithms like XGBoost and SVR based on the described R² and MAE evaluation metrics.
• Training models for wind power forecasting based on features from four distinct geographical sites.
• Comparing the performance of different SVR kernels (linear, RBF, polynomial) on time-series energy data.
• Studying the impact of outlier removal techniques (Z-score, IQR) on forecast model performance.

Strengths

• Includes results from four distinct geographical locations, suggesting potential for cross-site analysis.
• Specific performance metrics are reported, such as an R² of 0.99 for XGBoost and MAE values ranging from 11.10 to 15.94.
• Data is shared under a permissive CC-BY-4.0 license.

Limitations

• The dataset is very small (5.5 KB), indicating limited scope or containing only summary results, not raw time-series data.
• Column-level documentation is absent; field semantics must be inferred after download.
• Row count is unknown, which may limit suitability assessment for model training.

Provenance

Source

figshare, author Usman Ali.

Collection Method

Likely compiled from wind farm operational data for a machine learning study, with preprocessing involving Z-score and IQR outlier removal.

Freshness

Last updated 2026-04-30 17:54:58; freshness should be verified.

Geography

Four unspecified geographical locations.