In order to solve the optimal power problem of three-phase unbalanced distribution system for military bases, a scalable distributed optimization algorithm based on equivalent network approximation is proposed. The projection values of the power flow variables are exchanged with the adjacent distributed areas through the distributed coordination mechanism, and the shared variables are carefully selected according to the power flow characteristics of the radial distribution system, so that the number of macro iterations is greatly reduced. In the optimization process, the digital twins of the system are used to simulate and verify its robustness under the condition of limited communication. The simulation results show that the proposed method can successfully converge to the optimal solution in a reasonable time, and still perform well under the significant pressure of communication delay and low bandwidth in military bases.