Aiming at the issue of less fault samples and low diagnosis accuracy for wind turbine bearing, an approach for fault diagnosis of wind turbine bearing based on digital twin is proposed. First, digital twin system framework for wind turbine is constructed, which can provide data support for bearing fault diagnosis. Then, vibration signal of bearing is processed based on HHT (Hilbert-Huang transform) for data augmentation of sample data and noise reduction of vibration signal. Subsequently, wind turbine bearing fault diagnosis model is established based on CNN (convolutional neural network), where data augmentation sample is used as training and testing sample. Finally, experiment is designed to verify the feasibility and effectiveness of the proposed approach, which can realize the data enhancement of one-dimensional vibration signal and improve accuracy and stability of fault diagnosis of wind turbine bearing.