Abstract: In order to solve the problem of "dimensional disaster" in the detailed modeling of wind farms, this paper considers the relationship between each control link of the doubly-fed wind turbine and the system oscillation mode, by which the goal is to keep the dominant oscillation mode consistent before and after the equivalent value. For this reason, an equivalent aggregation method for wind farms considering the multi-time scale cascade characteristics of wind turbines is proposed. Firstly, this paper takes into account the cascade of multiple control links of wind turbines to establish a state space model of wind farms. Secondly, use the oscillation energy level to identify the dominant oscillation mode of the system. Analyze the key influencing factors affecting different modes with participation factor. Then, the key influencing factor of the dominant oscillation mode is taken as the clustering index, and data dimensionality reduction is performed in combination with auto-encoder. The equivalent model is obtained after improving aggregation algorithm clustering and topological transformation to obtain equivalent parameters. Finally, a single-machine model and a wind farm model with 24 doubly-fed wind turbines are built in Matlab to verify the validity of the equivalent method proposed in this paper.