A more robust CNN-BiGRU-TPA model for wind turbine blade icing prediction

Wind turbine blade icing poses a significant challenge to the reliability and efficiency of wind power generation, especially in cold and harsh climates. Accurately detecting icing conditions is essential for maintaining optimal turbine performance and preventing potential mechanical failures. However, conventional detection methods often face limitations when processing complex multivariate time-series data collected from Supervisory Control and Data Acquisition (SCADA) systems. In this study, we propose a novel hybrid deep learning model, CNN-BiGRU-TPA, which integrates Convolutional Neural Networks (CNNs) for spatial feature extraction, Bidirectional Gated Recurrent Units (BiGRUs) for temporal sequence modeling, and a Temporal Pattern Attention (TPA) mechanism to highlight critical temporal features. The proposed model demonstrates superior performance, achieving a recall of 0.8896, precision of 0.8904, F1 score of 0.8900, and Matthews Correlation Coefficient (MCC) of 0.7800. These results indicate its strong capability in accurately identifying icing events from complex SCADA datasets. This research provides a robust and scalable solution for real-time wind turbine monitoring and offers valuable insights for future applications in industrial big data analytics.