Abstract: To address the issue of imperfect carbon and green certificate trading mechanisms and the impact of source-load-market uncertainty and spatial and temporal correlation on the operation of virtual power plant (VPP), an optimal scheduling model for VPP participating in carbon-green certificate collaborative trading is proposed to consider multiple uncertainties and time-dependence. Firstly, a carbon-green certificate collaborative trading model based on market price is proposed, and an optimal scheduling model to minimize the integrated cost of VPP collaborative trading is constructed in the above market environment. Secondly, considering the uncertainty of the source-load-market, the time-correlation of uncertainties is dynamically learned based on the time generative adversarial network, and the temporal covariance conditions are introduced to get rid of the mismatch distribution of the generative fuzzy set, which is combined with the distributionally robust optimization model to optimize the scheduling of the VPP. Finally, the arithmetic example analysis verifies the advantages of the collaborative trading mechanism proposed in this paper regarding the economy and environmental protection of VPP, as well as the feasibility and validity of the TimeGAN-DRO model proposed in portraying uncertainty and time correlation.