There are a large number of power electronic inverters with weak overload capacity and lack of physical inertia in the microgrid. There is a high risk of instability when the microgrid fails. Therefore, it is urgent to study the transient stability assessment method to ensure the safe and stable operation of the microgrid. In this paper, a time series data-driven stable operation assessment method is proposed. Firstly, the transient interaction mechanism between constant power conversion controller and virtual synchronous machine ( VSG ) parallel grid-connected system is studied. Then a set of feature data sets with strong representation ability is constructed according to the interaction mechanism. Finally, based on long short-term memory neural network, a time series data-driven microgrid stable operation assessment model is established. The simulation results show that the feature selection of the proposed method is reasonable, and the stable operation assessment can be quickly and accurately realized under complex working conditions with good evaluation performance.