Abstract: The existing ultra-short-term prediction methods for renewable energy are difficult to effectively mine and analyze the inherent fluctuation rules of data. Important information is easy to be lost when the time sequence is too long. Therefore, a power prediction method of wind power and photovoltaic based on the wavelet bidirectional long-short term memory network（WBiLSTM） with Attention mechanism is proposed. Firstly, wavelet decomposition is used to extract the time domain information and frequency domain information of the input time series. Then, considering the bi-directional information flow, the bidirectional long-short term memory（BiLSTM） network is used for prediction. Attention mechanism is introduced, and different weights are given to the hidden state of BiLSTM through mapping weighting and learning parameter matrix, so as to selectively obtain the more effective information. Finally, the experiment is carried out using actual data, and the results show that the proposed model has better prediction performance compared with the existing models.