Super short-term wind speed prediction based on CEEMD decomposition and BILSTM-Transformer model

Wind speed has the characteristics of non-linearity, non-stationarity, intermittence and randomness. In order to improve the accuracy of wind speed prediction, this paper proposes a wind speed prediction model based on Complete Ensemble Empirical Mode Decomposition(CEEMD) and improved transformer. Firstly, CEEMD decomposition is used to reduce the instability of wind speed series and improve its predictability. Then, the BILSTM-Transformer prediction model is established for each sub-sequence obtained by decomposition. The bidirectional long-term and short-term memory artificial neural network BILSTM is used to replace the positional encoding in the transformer to extract the position characteristics of the sequence, and then the encoders in the transformer are used for prediction. Finally, the prediction results of all sub-sequences are superimposed to obtain the wind speed prediction results. Through the experiment of a domestic wind farm data, the prediction accuracy of the model in this paper is improved by 0.65 % ~ 1.47 %. The results show that the combined wind speed forecasting model has good forecasting ability and is feasible and effective in short-term wind speed forecasting.