ORCID
https://orcid.org/0000-0002-0083-4537
Abstract
Early and accurate fault detection in wind turbines is essential for improving operational reliability, reducing maintenance costs, and minimizing unplanned downtime. This study proposes a Hybrid Transformer-BiLSTM deep learning model for early fault detection and multiclass fault classification using Supervisory Control and Data Acquisition (SCADA) data. The proposed architecture combines the Transformer's self-attention mechanism to capture global temporal dependencies with the Bidirectional Long Short-Term Memory (BiLSTM) network's ability to model sequential fault evolution, enabling effective learning of multivariate time-series data. The model was developed and evaluated using the recently introduced CARE SCADA dataset, classifying five operating states: No Fault, Transformer Fault, Gearbox Fault, Generator Bearing Fault, and Hydraulic Fault. Performance was evaluated using the Coverage, Accuracy, Reliability, and Earliness (CARE) score, together with conventional classification metrics, including accuracy, precision, recall, F1-score, receiver operating characteristic area under the curve (ROC-AUC), and confusion matrix analysis. Experimental results demonstrate that the proposed Hybrid Transformer-BiLSTM model achieved an overall CARE score of 0.938, an accuracy of 94.2%, and an ROC-AUC of 0.96, outperforming baseline models, including Random Forest, XGBoost, LSTM, Transformer, and CNN-LSTM. These results demonstrate the effectiveness of the proposed model in providing accurate early fault detection and multiclass fault classification, highlighting its potential for intelligent predictive maintenance and condition monitoring of wind turbine systems.
Recommended Citation
Mkindu, H. Y. (2026). Hybrid Transformer-BiLSTM Model for Early Fault Detection and Multiclass Classification of Wind Turbine Faults Using SCADA Data. Tanzania Journal of Engineering and Technology (TJET), 45(2), 119-131. https://doi.org/10.65085/2619-8789.1073
Publisher Name
University of Dar es Salaam