Abstract: Accurate wind power prediction can improve the safety and reliability of grid operation. To further enhance the accuracy of short-term wind power prediction, this paper proposes a CNN-LSTM&GRU multi-model combined prediction method considering the difficulty in obtaining optimal prediction results with a single model. Firstly, a convolutional neural network (CNN) is used to extract local features of data and combined with a long short-term memory (LSTM) network to construct a CNN-LSTM network structure that incorporates local feature pre-extraction modules. Then, the CNN-LSTM network is paralleled with a gated recurrent unit (GRU) network. An adaptive weight learning module is employed to select the best weights for the outputs of the CNN-LSTM module and the GRU module. In this way, the paper constructs a combined short-term prediction model based on CNN-LSTM&GRU. Finally, the model is applied to the power prediction of a wind farm in northwestern China. The experimental results show that the proposed model has a smaller mean absolute error (MAE), a smaller root mean square error (RMSE), and higher prediction accuracy than single models and other combined models.

Key words: short-term wind power prediction, CNN-LSTM, GRU, combined prediction, adaptive weight learning