It is an effective approach to improve the stability and economic efficiency of the distribution system by managing large quantities of distributed resources and participating in the electricity market utilizing aggregators. The existing multi-aggregator optimization problems with multiple scenarios and many variables, coupled with the source-load uncertainty and the data privacy protection of each subject, bring great challenges to the efficient solution of the model. In this paper, a two-layer and two-stage collaborative optimization framework is constructed for multi-resources aggregators in the distribution network, in which the upper level is regulated by a regional controller to settle real-time price within the region; and the lower level of each aggregator is constrained by the price adjustment strategy of the upper level to autonomously optimize distributed power sources according to their interests. The data privacy inside the grid is protected by the proposed architecture. Finally, the simulation verifies the validity and superiority of the proposed method.