The maximum flow problem is an important research field in graph theory, and it has a wide range of applications in urban traffic flow, network planning and other fields. There are many approaches to solve the maximum flow problem, but there are still some shortcomings. To solve the problem of high computational complexity of the maximum flow algorithm, a parallel maximum flow algorithm is proposed which divides the original graph into overlay graph and subgraphs. The experiments in TIGER/Line show that the maximum flow result calculated by the algorithm in this paper is consistent with that of the Edmonds-Karp algorithm, the proposed algorithm significantly reduces the calculation time and the time complexity.