The existing store and forward(SnF) scheduling methods suffer from high computational complexity and heavy state maintenance, which seriously restricts its application in the field of big data transfer. Moreover, the traditional global modeling results in a lot of redundant states in the scheduling problem. Aiming at this problem, this paper proposes a state fusion scheduling method based on time-shifting multi-layer graph(TS-MLG). The method reduces redundant states and improves the scheduling efficiency of the algorithm by fusing the states of preselected paths. The simulation results show that compared with the traditional scheduling method, the proposed method has the advantages of short computing time, good scheduling performance and less state maintenance, which suggests that it can provide real-time and efficient scheduling services for large-scale networks.