Abstract: Abstract: Recently, as an effective method of using distributed power generation, microgrid has been developed rapidly. The microgrid's safe and economic operation has been paid more and more attentions. Microgrid economic operation cannot goes well without perfect energy optimization scheduling. Microgrid energy optimization scheduling is based on the distributed power generation forecasting and load forecasting, aiming at making full use of energy storage unit, coordinating microgrid internal power supply and the power/energy flow between microgrid and large power grid and maximizing the microgrid economic benefits. In current literature, the dynamic optimization model and static optimization model can't combine the day-ahead scheduling with real-time scheduling and fail to make full use of the storage and removal effect of energy storage unit. In order to meet the project needs for real-time optimal dispatch of microgrid, in this paper, we proposed a real-time optimal dispatch model for microgrid based on the state of charge of day-ahead scheduling, and the network flow model. In this method, based on the real-time super short power generation/consumption prediction data, we developed methods to follow the state of charge of day-ahead scheduling, the future 5-15 minutes state of charge, charge and discharge power, the micro power supply, and power output planning. Firstly, the model took minimum equivalent power supply cost as the goal, it put four parts as constraint conditions: the controllable micro powers output, storage level, the interaction power between microgrid and power grid, and the power balance of microgrid. Then, based on the minimum cost and maximum flow theory, the nonlinear model was converted into a network flow model that can be solved linearly. In this theory, a directed topology network had been constructed. Three kinds of nodes existed in the network: 1) two virtual nodes: an original node and an terminal node; 2) four power output nodes: renewable energy generation output node, discharge power of grid node, discharge power of energy storage node, and controlled micropower supply output node; and 3) three power input nodes: load node, charge power of grid node and charge power of energy storage node. The arcs of directed topology were divided into three subsets. They were 1) four power output arc, from original node to power output nodes; 2) ten mutual power arc, from power output nodes to power input nodes; and 3) three power collection arc, from power input nodes to terminal node. Finally an example of microgrid was given, using the optimization model and algorithm which were proposed in this paper. The optimization scheduling results of grid-connected operation model and island operation model showed that by reasonable control of state of charge in each period, the given optimization scheme can maximize utilization of renewable energy power and minimize power supply cost of microgrid.