Abstract: In the seismic fragility analysis of substation equipment, it is necessary to calculate the structural responses under the inputs of a large number of ground motions, and a same fragility curve is often used for different ground motions. In this paper, a fragility analysis method based on ground motions clustering was proposed to reduce the inputs of ground motion samples and to ensure the accuracy of the results. And the ground motions were classified, the fragility curves of multiple ground motions were obtained. Firstly, the principal component indicators combining multiple ground motion parameters were obtained by principal component analysis. The principal component indicators were used as the clustering indicators to classify the time history records of ground motions through k-means clustering algorithm. Based on the clustering results, a small number of typical ground motions were selected instead of all ground motions to calculate the fragility curve of a 1100kV transformer bushing. After that, the fragility curves of each type of ground motions were calculated and compared. The results show that the small number of typical ground motion samples selected based on clustering results can represent all samples well. Using this method, a small number of typical samples can be input to get a relatively accurate fragility curve. In addition, the fragility curves of various kinds of ground motions differ greatly. When the type of ground motion has been determined, it is more accurate to use the fragility curve corresponding to the type of ground motion to assess the failure probability of substation equipment.