Abstract: Flexible resources are usually aggregated into virtual power plants to participate in power grid scheduling. A multi-form flexible resource aggregation model for virtual power plant is therefore proposed based on grey target theory and spectral clustering. Firstly, based on the response characteristics of new energy generation, distributed energy storage and flexible loads, the common frequency modulation performance indicators and peak shaving performance indicators are established for each flexible resource from three aspects, including response time, response capacity and daily load fluctuation rate. Secondly, based on the grey target theory, objective weighting method and spectral clustering, the flexible resources in the virtual power plant are classified into frequency modulation resources and peak shaving resources. Finally, a frequency modulation-type virtual power plant aggregation model and a peak shaving-type virtual power plant aggregation model are established respectively, and their post-aggregation response characteristics are studied. The numerical simulation shows that the response time and the daily load fluctuation rate of the post-aggregation virtual power plant are reduced. The peak shaving-type virtual power plants should be used preferentially under peak shaving scenarios. When the peak shaving-type virtual power plants can not meet the peak shaving requirements, the frequency-modulation virtual power plants can participate in peak shaving to improve the utilization efficiency in different scenarios.