Establishing a Voltage Source Converter (VSC) model plays an important role in the study of wind turbine grid-connected characteristics analysis. In order to solve the problem that the VSC linearization model is cumbersome for modeling high-frequency power electronic switching devices, and the fact that the data-driven model has poor generalization and low precision, this paper proposes a VSC equivalent modeling method based on a model-data hybrid drive method. In the VSC model, fast calculation of a proportional-integral control link is done using a linearization model, and then the data model of the PWM link is constructed using a time convolutional network. At the same time, the signal decomposition synthesis is embedded into the learning process of the data model, and the VSC deep learning modeling framework driven by the model-data hybrid is constructed. The establishment process of a VSC model of a permanent magnet direct driven synchronous generator is given by PSCAD simulation data. The feasibility and superiority of the proposed method are verified by samples under different working conditions. This work is supported by the Fundamental Research Funds for the Central Universities (No. YJ201654).