Two Step Wind Speed Prediction Based on a U-shaped RBiLSTM Hybrid Model

Accurate wind speed prediction can improve energy utilization, reduce energy waste, and save motor maintenance costs. A novel wind speed prediction model, the BEGWO-URBiLSTM model, is proposed for ultra-short-term 2 step wind speed prediction. Using the Empirical Mode Decomposition (EMD) method, IMF components corresponding to different frequency domains are extracted from the raw wind speed data, and then a U-shaped Reinforced Bidirectional Long ShortTerm Memory(URBiLSTM) neural network model is proposed for time series prediction of these extracted IMF components. The U-shape structure equips the model with encoding and decoding capabilities, which helps capture features with long-term dependencies. The peekhole structure in RBiLSTM enables the model to acquire more information during time series processing, while the bidirectional architecture allows it to simultaneously capture information from both the past and the future. We also propose a population behavior-enhanced Gray Wolf Optimizer (BEGWO) that incorporates metaheuristic adaptive improvements to the standard GWO. This approach dynamically balances the algorithm’s global exploration and local search capabilities, thereby enhancing its convergence speed. BEGWO was used to optimize the hyperparameters in URBiLSTM, and the accurate selection of hyperparameters improved the prediction accuracy of the model. Experimental results show that the proposed model outperforms other models in terms of prediction accuracy and reliability when applied to complex wind speed time series forecasting.