Abstract: Exploring the mechanisms and profit models of various investment entities participating in the market within virtual power plants is of great significance in the construction of new energy systems and new power systems. This article constructs a centralized distributed control system for virtual power plants based on the needs of multiple investors, and proposes a capacity optimization configuration model for virtual power plants based on game theory and value at risk under multi-agent and multi power source combinations. In practical applications, the types of distributed power sources can be added according to the requirements of investors, and the corresponding distributed power capacity configuration results can be obtained by arranging and combining different scenarios in sequence. This article conducts simulations based on distributed power sources and load data in a certain region of China. And the results verify the feasibility and effectiveness of the proposed method.