Accurate wind power prediction is of great importance to balance the power supply and demand, in order to improve the prediction accuracy of wind power, this paper proposes a short-term wind power prediction model based on Convolutional Neural Network (CNN), Bidirectional Long and Short-Term Memory Network (BiLSTM) and Improved Zebra Optimization Algorithm (PKZOA). The model simultaneously considers the spatial characteristics of the sequence data, firstly, the feature extraction capability of the convolutional layer of the CNN is used to capture the spatial distribution pattern of the meteorological factors; secondly, the forward and backward sequences of the data are processed by the BiLSTM layer to improve the performance of the prediction model, and at the same time PKZOA is applied to optimize the parameters of the CNN-BiLSTM model and to optimize the network structure of the model; and finally, the network structure is optimized using the PKZOA-CNN-BiLSTM model for short-term wind power prediction. The results show that the PKZOA-CNN-BiLSTM model can significantly improve the prediction accuracy.