Abstract: :In order to solve the problem of inaccurate short-term power prediction of wind power caused by stochastic and fluctuating wind power, a short-term wind power prediction method based on bi-directional long short-term memory of variational mode decomposition and sparrow search algorithm is proposed.Firstly, the wind power in the historical data is decomposed into several sub-sequences by using variational mode decomposition, and each element in the sub-sequence corresponds to a historical meteorological data vector at a certain moment, which forms the original data matrixs.Then, the bidirectional long short-term memory of sparrow search algorithm power prediction method is used to model several original data matrices respectively, and the optimal parameters of bidirectional long short-term memory (BiLSTM) are automatically found by using sparrow search algorithm, and finally the prediction results are superimposed to form the final prediction results. Based on the actual operation data of a wind farm in Hunan province, the simulation results show that the root mean square error, mean absolute error, and mean absolute percentage error of the proposed model in this article are reduced by 77.29%, 75.52% and 75.04% respectively compared with the BiLSTM model, which can effectively improve the short-term power prediction accuracy of the wind farm.

Key words: wind power short-term power prediction, variational mode decomposition, sparrow search algorithm, bidirectional long short-term memory network