A distributed algorithm based on super nodes is proposed to iteratively locate the large number of sensor nodes in wireless sensor networks. Based on the overlapped decomposition of the entire network, two steps are involved at each iteration of the algorithm, one is the node localization method within each subgraph and the other is the local consensus among neighboring subgraphs. In specific, for each subgraph, the conjugate gradient method is employed to determine the position of nodes. And in subsequent, the position of each node will be adjusted by using a neighboring consensus strategy. These two steps proceed until the iterative termination condition is satisfied. Simulation results show that the proposed algorithm has an order of magnitude lower positioning error than the existing distributed algorithms and can efficiently locate nodes for large size wireless sensor networks.