Decarbonization of power systems is the key to zero carbon development for the whole of society. In recent years, distributed microenterprise agents have developed rapidly and will become an important part of China's energy structure. First, starting from the virtual power plant (VPP), and based on the analysis of VPP's participation in China's current carbon trading market, the operation mechanism for VPP participation in the electricity carbon joint market is proposed. Second, a two-stage and two-tier bidding strategy model of a multi-agent interactive game under the joint market operation mechanism is constructed. The first stage is the VPP internal distributed micro agent pre scheduling model. In the second stage, the inner layer is a non-cooperative game bidding model between VPP and multi market players. With the goal of maximizing the total revenue of VPP participation in the day ahead spot market and carbon trading market, it uses multiple scenarios to describe the uncertainty of competitor quotations. The outer layer is a multi-agent interactive decision-making clearing model with the goal of maximizing the whole of social welfare. Finally, a Q-learning algorithm and path tracking interior point method are used to analyze the model. The feasibility and effectiveness of the proposed model are verified by an example.