Deep learning-based short-term wind speed prediction for offshore wind farms

Wind energy plays a crucial role in the renewable energy sector, particularly within the realm of offshore wind farms. Effective harnessing of offshore wind energy necessitates accurate and dependable wind speed forecasting. This paper proposes a novel deep learning model for short-term wind speed prediction in offshore wind farms, integrating co-decomposition and temporal feature enhancement to improve forecasting precision.The proposed framework begins with Time-Varying Filtering-Based Empirical Mode Decomposition (TVF-EMD) optimized via the Subtraction-Average-Based Optimizer (SABO) to decompose raw wind speed data into intrinsic mode functions (IMFs). Sample Entropy (SE) is then applied to extract meaningful features, followed by multimodal feature selection and co-smoothing hysteresis correction to eliminate noise and transient fluctuations. For prediction, a Convolutional Neural Network (CNN) captures spatial dependencies, while a Gated Recurrent Unit (GRU) network with an attention mechanism enhances temporal modeling, ensuring robust and interpretable forecasts.Extensive experiments demonstrate the model’s superiority over benchmark approaches, achieving an RMSE of 0.1624, MAE of 0.1139, MAPE of 0.83%, and R2 of 0.9962. These results demonstrate the model’s potential for practical application, ensuring reliable wind farm operation.