Abstract: On account of the pressing goal of “carbon peaking and carbon neutrality”, it is extremely urgent to develop wind power and study the broadband oscillation problem of wind farm connected to the grid. With the purpose of establishing a simulation model suitable for wind farm oscillation research, which is able to accurately and efficiently reflect the actual wind farm output characteristics, a fuzzy hierarchical clustering method is proposed that improves the clustering result offset caused by distance measurement in traditional balanced iterative reducing and clustering using hierarchies. Besides, F test is used to evaluate the clustering effect quantitatively, and F score can be used to horizontally compare to determine the optimal cluster number and longitudinal comparison to determine the optimal clustering scheme. Compared with traditional clustering method, fuzzy metric hierarchical clustering method has the highest F score. Finally, a wind farm simulation example consisting of 25 wind turbines with actual parameters in PSCAD is set up, and the conclusion is obtained that the most suitable clustering method for wind farms is fuzzy metric hierarchical clustering method.