Explainable Multi-Horizon Wind Power Forecasting via Aquila-Optimized Machine Learning Models

Mostafa A. Abdelnaby, Computers and Control Systems Engineering Department, Faculty of Engineering, Mansoura University, Egypt.Follow
Nahla B. Abdel-Hamid, Computers and Control Systems Engineering Department, Faculty of Engineering, Mansoura University, Egypt.
Eman M. El-Gendy, Computers and Control Systems Engineering Department, Faculty of Engineering, Mansoura University, Egypt
Mahmoud M. Saafan, Computers and Control Systems Engineering Department, Faculty of Engineering, Mansoura University, Egypt.

Subject Area

Computer and Control Systems Engineering

Article Type

Original Study

Abstract

The inherently unpredictable nature of wind energy necessitates the development of sophisticated forecasting models to ensure grid stability and optimal distribution. In this research, we propose a novel and systematic approach to wind power forecasting (WPF) across diverse timescales. This approach leverages the power of various machine learning (ML) models, metaheuristic hyperparameter optimization, and utilizes explainable artificial intelligence (XAI). The developed methodology is based on the Aquila optimizer (AO), capable of automatically adjusting different ML models for four different time periods (30 minutes, 6 hours, 24 hours, and 36 hours) on the data collected from the Gabal El-Zayt wind power plant in Egypt. The improved models demonstrate superior accuracy with near-perfect R² values , outperforming existing models, including complex hybrid deep learning (DL) methods, with improvements of up to 9.65% in R² metrics and significantly lower error values. Finally, to achieve greater predictive efficiency, SHapley Additive exPlanations (SHAP) interpretability analyses show that the models are physically valid, with wind speed being a critical factor, and their behavior is dependent on time horizon patterns. Thus, this research presents a robust, clear, and accurate forecasting method that specifically addresses the challenges of operating variable renewable energy sources in the modern power grid.

Keywords

Wind Power Forecasting, Aquila Optimizer, Machine Learning, Multi-Horizon Forecasting, SHAP, Renewable Energy Systems

Creative Commons License

This work is licensed under a Creative Commons Attribution 4.0 License.

Recommended Citation

Abdelnaby, Mostafa A.; Abdel-Hamid, Nahla B.; El-Gendy, Eman M.; and Saafan, Mahmoud M. (2026) "Explainable Multi-Horizon Wind Power Forecasting via Aquila-Optimized Machine Learning Models," Mansoura Engineering Journal: Vol. 51 : Iss. 5 , Article 2.
Available at: https://doi.org/10.58491/2735-4202.3467

Download