Abstract: The factors such as the current limiting strategy, the source-source interaction, the variable fault and load level make it very difficult to quickly and accurately assess the transient stability of the microgrid with multiple virtual synchronous generators（VSGs）. Aiming at the existing problems, this paper proposes an online transient stability assessment method for the microgrid with multiple VSGs based on the deep learning. First, by analyzing the influence of VSG control characteristics, current limiter,fault level, and load level on the system stability, a set of original features with the abilities of strong characterization and avoiding dimensionality disasters is constructed with the principle of system dynamic variables as the mainstay and steady-state parameters as the supplement. Based on this, a transient stability nonlinear assessment model for the microgrid with multiple VSGs is proposed with the application of deep feedforward neural network and Levenberg-Marquardt algorithm. The verification results in the microgrid with multiple VSGs show that, compared with the existing methods, the proposed method greatly improves the accuracy of the online transient stability assessment, and can quickly realize the accurate stability judgment of the microgrid with multiple VSGs under complex working conditions, which prove that the proposed method has a good assessment performance.