Wind Speed Prediction Based on RD-BiLSTM

Due to the characteristics of wind speed such as fluctuation and non-stationality, the accuracy of wind speed prediction is not high. An amazing wind speed prediction method based on RD-BiLSTM deep neural network is established and proposed. The original sequence is decomposed by the RD module, and then input into the BiLSTM model for training and prediction. Finally, the predicted results are superimposed to obtain the final wind speed prediction result of the model. The value of MAE, RMSE, MAPE of RD-BiLSTM model on March data are 0.169, 0.239, 0.056, respectively. It is superior to ARIMA, LSTM, BiLSTM, GRU, Transformer models, and is basically superior to these models when tested on other five data sets.