Hybrid Machine Learning Framework for SCADA-Based Anomaly Detection in Wind Turbine Systems Using Ensemble Learning

Abstract

Supervisory Control and Data Acquisition (SCADA) systems play a crucial role in monitoring the operational health of modern wind turbines by continuously collecting large volumes of sensor data. Efficient analysis of this data is essential for early fault detection, predictive maintenance, and reliable turbine operation. However, anomaly detection in SCADA environments remains challenging due to data imbalance, noise, nonlinear relationships, and dynamic operating conditions. Traditional machine learning approaches often suffer from limited generalization capability and may fail to achieve a balanced trade-off between precision and recall. To address these challenges, this paper proposes a Hybrid Machine Learning Framework for SCADA-based anomaly detection in wind turbine systems using ensemble learning. The proposed framework integrates Isolation Forest, One-Class Support Vector Machine (OCSVM), and Deep Autoencoder models to capture complementary anomaly characteristics from operational data. The outputs of these base models are further combined using an AdaBoost-based stacking architecture to improve classification robustness and anomaly detection performance. Experiments were conducted on a publicly available wind turbine SCADA dataset containing more than 50,000 operational samples and multiple turbine health parameters. The proposed hybrid framework was evaluated using Precision, Recall, F1-score, Area Under Curve (AUC), and confusion matrix analysis. Experimental results demonstrate that the proposed model significantly outperforms standalone approaches, achieving a Recall of 0.8702, F1-score of 0.7118, and AUC of 0.9678. Furthermore, the framework substantially reduces false negative predictions, making it highly suitable for predictive maintenance applications. The findings indicate that integrating machine learning, deep learning, and ensemble learning techniques provides a robust and effective solution for intelligent anomaly detection in industrial SCADA systems. Keywords: machine learning; deep learning; Anomaly detection; AutoEncoder; Ensemble Learning; One-Class SVM; SCADA; Wind Turbine Monitoring; Isolation Forest; Predictive Maintenance