Network partition is the basis of distributed state estimation of distribution network. This paper analyzes the network structure characteristics of the power distribution network, and considers calculation quantity equilibrium of each partition, realizes the automatic partitioning of the distribution network by virtue of the rapid search capability of the network in the graph theory, so as to realize parallel computing of state estimates. To enhance the robustness of the state estimation algorithm, using exponential variable weight coefficient to reducing the weight value of the bad data in the iterative process, thus reducing the adverse influence on the target function and estimated results. The distributed and robust state estimation algorithm based on partition can realize automatic reasonable partition of distribution network, effectively improve the computing speed of the state estimation, improve the algorithm to the measurement's robust. Finally, the calculation and analysis based on IEEE-118 standard example have showed the validity of the algorithm.