The optimal site selection of distributed power stations on a weighted network is studied by combining the topology and electrical characteristics of the network. Firstly, the second-order Kuramoto-like phase oscillator model is used to model the power grid, and the network power flow value under equal coupling strength is used to weight the power grid lines to construct a weighted network model of the power grid; by calculating the transmission efficiency from each load node to each generator node and the power absorbed by each load node from each generator node, which defines the power supply efficiency index of source-load node pair. Then, according to the power supply efficiency value of the source-load nodes of each load node, three grid access methods are defined. The influence of three different grid access methods of distributed power stations on the synchronizability of the grid is studied on the coupling weighted network. Study shows that the access mode of distributed power stations is the best in order to arrange the power supply efficiency value from small to large, followed by the way of randomly selects load nodes, and the way in which the power supply efficiency value is arranged in descending order is the worst. Therefore, the power grid can achieve better synchronizability when the distributed power stations connect to the power grid on the load node with the lower power supply efficiency index of the source-load node pair.