Considering the fluctuation and uncertainty of wind power generation, a new wind power prediction model based on convolution neural network and LightGBM is proposed in this paper. Firstly, a new feature set is constructed by analyzing the temporal characteristics of the original data of wind farms and adjacent wind farms. Secondly, the information is extracted from the input data by using convolution neural network (CNN) and the network parameters are adjusted by comparing the actual results. Then, considering the limitations of single convolution model in wind power prediction, LightGBM classification algorithm is integrated into the model to improve the prediction. Accuracy and robustness of measurement. Finally, the proposed algorithm is compared with the existing support vector machines, LightGBM and CNN. The results show that the proposed fusion model has better accuracy and efficiency.