The random fluctuation and uncertainty of renewable energy cause system reliability to decline sharply and the need for system flexibility to increase sharply. This can lead to significant decrease of the renewable energy consumption and serious waste of resources. To solve the above problems, this paper proposes a hierarchical capacity allocation strategy of virtual power plant (VPP) considering reliability, flexibility and economy. An ‘upper’ model formulates a load reduction and transfer scheduling scheme from the perspective of guiding users to respond to the change of renewable energy output by using TOU price. A ‘middle-level’ model formulates the capacity configuration scheme of an energy storage battery from the perspective of optimal reliability. Under the constraint of a flexibility index, the ‘lower’ model further improves the energy storage battery capacity configuration scheme formulated by the middle model from the perspective of optimal and comprehensive operational cost. Finally, taking the load data of a region in Xinjiang as an example, a simulation is carried out. The results show that the proposed method has obvious economic benefits and provides a new idea for carbon peaking and carbon neutralization. This work is supported by the National Natural Science Foundation of China (No. 51767023).