Wind speed and power forecasting using Bayesian optimized machine learning models in Gabal Al-Zayt, Egypt.

Area of Science:

• Renewable Energy Systems
• Computational Intelligence
• Meteorological Forecasting

Background:

• Accurate wind speed and power prediction are vital for efficient renewable energy integration.
• Existing forecasting methods face challenges in accuracy across diverse time scales.

Purpose of the Study:

• To compare and evaluate ten machine learning techniques for wind speed and power prediction.
• To identify the most effective models for wind speed prediction (WSP) and wind power prediction (WPP) across various time scales.

Main Methods:

• Utilized a wind speed and power integration prediction system.
• Compared single and ensemble machine learning models, including Light Gradient Boosting Machine (LGBM), Extreme Gradient Boosting, and Bagged Decision Tree (BDT).
• Evaluated model accuracy using metrics: Pearson's correlation coefficient (R), explained variance (EV), mean absolute percentage error (MAPE), mean square error (MSE), and concordance correlation coefficient (CCC).

Main Results:

• For WSP, LGBM, Extreme Gradient Boosting, and BDT demonstrated high accuracy (MAPE: 2.641–12.274%, R: 0.943–0.997).
• For WPP, LGBM and BDT showed strong predictive performance (MAPE: 0.277–186.710%, R: 0.985–1.000).
• Model performance was consistent across different time scales.

Conclusions:

• Light Gradient Boosting Machine and Bagged Decision Tree are highly effective for both wind speed and power forecasting.
• These machine learning models offer reliable solutions for enhancing renewable wind energy applications.