Short-term prediction of wind power based on phase space reconstruction and BiLSTM

Abstract

Aiming at the chaotic characteristics of wind power sequence and combined with meteorological information, a short-term prediction method of wind power based on phase space reconstruction and bidirectional long short-term memory neural network (Re-BiLSTM) is proposed. Firstly, the embedding dimension m and time delay τ of the time series are determined by the C–C method, and the wind power data is reconstructed based on the embedding theorem. The reconstructed data and normalized meteorological data (wind speed, wind direction) are then used as inputs, and bidirectional long short-term memory neural network (BiLSTM) is used to make short-term prediction of wind power. The results show that compared with artificial neural networks, BiLSTM, Random forest, and K-Nearest Neighbor, Re-BiLSTM has lower prediction error, which fully proves the effectiveness of the model.