Employing a detailed model for every wind turbine in large-scale wind farm simulations substantially increases computational complexity. This not only extends simulation time but may also exceed the node limit of the simulation software. Therefore, it is necessary to study the method of multi-machine equivalence for wind farms under different scenarios. To address the problem of subjective and incomplete wind turbine group clustering indicators, variables reflecting the wind turbine operation status are selected as initial data samples, and high-dimensional data samples are downscaled using Kernel Principal Component Analysis (KPCA) to eliminate redundancy and correlation of the indicator data and enhance the validity of the clustering indicators. The optimal number of clusters is determined by the elbow method, and then the K-means algorithm is used to complete the division of the group and compute the equivalence of the machine parameters. The pooling line equivalence is modeled according to the power loss equalization principle. Finally, the simulation result verifies the correctness of the proposed equivalence method.