Intelligent fault detection in wind turbine gearboxes: A machine learning approach using SCADA temperature data

Wind energy plays a pivotal role in the global transition to renewable energy, offering a sustainable solution to reducing greenhouse gas emissions. Gearbox (GB) failures remain one of the most critical issues, often leading to significant downtime and expensive repairs. This study presents a robust methodology for predicting wind turbine GB failures using Supervisory Control and Data Acquisition (SCADA) data. A novel Bagging Ensemble Temperature Prediction (BETP) model is proposed and evaluated alongside other ML models to forecast GB temperature and detect early signs of overheating. The BETP model demonstrated superior performance, achieving the lowest mean squared error (3.0259) and the highest R 2 score (0.9527) among all models. By using its predictive capabilities, the model successfully detected early GB anomalies 59 days before failure and identified critical irregularities 9 days before a complete breakdown. These findings emphasize the effectiveness of the BETP model in enabling predictive maintenance, reducing unplanned downtime, and optimizing O&M costs. The proposed approach not only enhances the reliability and operational efficiency of WTs but also contributes to the long-term sustainability of wind energy systems.