Research on Ultra Short Term Wind Speed Prediction Method Based on CEEMDAN-BiLSTM-AM Model

To improve the accuracy of wind speed prediction and explore the nonlinear characteristics of wind speed sequences, a ultra short term wind speed prediction method based on the CEEMDAN-BiLSTM-AM model is proposed. Firstly, in response to the random fluctuations of wind speed, a complete ensemble empirical mode decomposition with adaptive noise (CEEMDAN) is used to decompose the wind speed sequence into multiple relatively stationary sub modes, to reduce the complexity of prediction. Secondly, a bidirectional long short-term memory network (BiLSTM) with a bidirectional information flow structure is used to mine the changing patterns of their respective patterns, and weights are assigned to the hidden layer states of the neural network through attention mechanisms to highlight key information in long-term sequences. Then, the Bayesian algorithm is employed to optimize the hyperparameters of the model, and the predicted results of each component are superimposed as the final result. Finally, the results of the numerical analysis show that the proposed model exhibits good predictive performance in both single step and multi-step prediction tasks.