With the in-depth implementation of the energy Internet strategy, the participation of renewable energy and microgrid continues to rise．The uncertainty factors in the system have increased significantly, the cooperation and competition among the participating entities have become more and more complicated．A two-layer electric energy trading system with mixed game among power grid, service provider and multi-microgrid is established from vertical and horizontal levels．At the vertical level，the idea of master-slave game is proposed，with service provider as the leader and microgrid as the subordinate．The uncertainty problem is unified in the second stage optimization of the rough and error-free，the segmented robust optimization model of uncertainty stage optimization is constructed，the differential scheduling of uncertainties is realized，and the flexibility of robust optimization is improved．Using Bool-Column and constraint generation(B-C&CG) algorithm to solve the model，and the whole model is divided into the main and sub problems. The main problem optimizes the electricity price uncertainty problem，and the sub-problem optimizes the source load uncertainty problem．In a horizontal level, Nash negotiation model is structured，using the alternating direction method of multipliers(ADMM) algorithm to solve the horizontal level power interaction model between the microgrid．Using the distributed solution method, the trading price strategy is obtained．Then combined with Lagrange multiplier method，each part is optimized alternately and the multiplier is updated．The optimal transaction price of each microgrid is obtained．The simulation results show that the proposed scheme takes into account the robustness，economy and flexibility of the system，reduces the cost of each microgrid and fully protects the privacy of each microgrid．