What question did this study set out to answer?

This research aims to enhance wind energy production forecasting using advanced machine learning models in Türkiye.

March 26, 2026Open Access

A Comprehensive Case Study of Wind Energy Production Forecasting in Türkiye Using Enhanced Attention BiLSTM

MÇMehmet ÇeçenSelçuk University MRMarco Aurelio RiveraUniversity of Talca MYMehmet YavuzUniversity of Exeter

Key Points

This research aims to enhance wind energy production forecasting using advanced machine learning models in Türkiye.
Utilized EABiLSTM and LSTM models for prediction.
Employed real-time generation and meteorological data.
Conducted data preprocessing and hyperparameter optimization.
Performed performance evaluations using MAE, R^2, and RMSE metrics.
Compared forecasting accuracy between standard LSTM and EABiLSTM models.
EABiLSTM model demonstrated superior accuracy compared to standard LSTM.
Achieved strong performance in forecasting accuracy as evaluated by MAE and RMSE.
Validated the model's reliability through temporal stability and uncertainty quantification tests.

Abstract

ABSTRACT The global demand for electrical energy continues to rise steadily, with Türkiye experiencing particularly significant growth in energy consumption. To address this increasing demand sustainably, renewable energy sources (RES) have become the primary focus, with wind energy (WE) leading the transition. This study presents a comprehensive case study of wind energy production forecasting for a specific region in Türkiye, utilizing advanced machine learning (ML) methodologies. The study employs long short‐term memory (LSTM) and enhanced attention bidirectional long short‐term memory (EABiLSTM) models to predict wind energy production using real‐time generation data and comprehensive meteorological parameters. The methodology encompasses rigorous data preprocessing techniques, hyperparameter optimization, including normalization, temporal feature engineering, and advanced validation strategies. A comprehensive hourly dataset forms the foundation of this analysis, providing robust temporal coverage for training and validation. Meteorological data are sourced from the NASA Power project, while wind power plant production data are obtained from the Energy Markets Operation Corporation of Türkiye (EPIAS) transparency platform, ensuring reliable generation records. The performance evaluation employs multiple metrics, including mean absolute error (MAE), coefficient of determination (), and root mean square error (RMSE), to assess forecasting accuracy. The prediction accuracy and reliability of the proposed EABiLSTM method were validated through temporal stability tests, seasonal robustness evaluation, and uncertainty quantification analysis. The effectiveness of the proposed methodologies is demonstrated through comparative analysis between standard LSTM and EABiLSTM models. The regional case study approach provides practical insights for wind energy operators and grid planners, contributing to renewable energy optimization strategies and supporting the country's sustainable energy transition goals.

Ask AI

Helpful

Bookmark

View Full Paper