Traditionally, transmission and distribution power flows are studied separately. But lots of large cities have their hybrid transmission and distribution networks. In order to achieve a global unified power flow solution, transmission and distribution networks are studied as a whole in this paper. Global power flow (GPF) equation is built as a mathematical model for the global power system. Based upon the master-slave-typed feature of the global power system, a master-slave-splitting (MSS) iterative method is developed for balancing the boundary mismatch between the transmission and distribution networks. In this method, the GPF problem of large scale is split into a transmission power flow and lots of distribution power flow sub- problems, which supports on-line geographically distributed computation. In order to fit the different features between transmission and distribution networks, each sub-problem can be solved with different algorithms. Several case studies are carried out, and accuracy, efficiency and reliability of the proposed method are verified.