Hybrid Deep Recurrent Neural Networks for Accurate Wind Speed Prediction

The wind power industry commonly uses time series wind speed predictions. Wind farms and electrical systems are significantly impacted by wind speed instability and fluctuation, which can result in system failure in extreme circumstances. Therefore, accurate wind speed predictions are crucial for generating renewable energy. In this study, we proposed three models LSTM-GRU-attention, LSTM-BiLSTM-attention, and GRU-BiGRU-attention, for wind speed prediction based on hybrid deep recurrent neural networks. Data normalization, data prediction, and assessment of the model's performance on actual and expected values were all included in the planned research. In terms of MAE, RMSE, R-squared (r2), and MAPE, the proposed LSTM-BiLSTM-attention model obtained 0.02, 0.111, 0.962, and 0.075 values, respectively. The proposed LSTM-BiLSTM-attention model was cross-validated with time series rolling window validation and time series nested window validation. The experimental findings demonstrate that the proposed LSTM-BiLSTM-attention model works better than the state-of-the-art and other proposed models.