The massive access of renewable energy, electric vehicles makes it more difficult to estimate the real-time state of distribution network. The existing distribution network measurement system is difficult to meet the situational awareness requirements of distribution systems that contain a lot of uncertainty of distributed power and electric vehicle. For this reason, an optimal configuration method of phasor measurement unit in distribution systems for interval state estimation is proposed. First, a mathematical model of the minimum interval state estimation average interval width target is constructed based on the minimum path between each node and other nodes. Then, this paper takes the minimum number of PMU configurations, maximum redundancy and minimum interval state estimation average interval width as the goal and the configuration quantity as the constraint to establish a multi-objective optimization configuration model, and adopts the non-dominant ranking genetic algorithm based on non-dominated sorting genetic algorithm (NSGA-Ⅱ) to obtain the optimal solution set of the model. Finally, through NSGA-Ⅱ, the optimal solution obtained multiple times is solved, and a suitable configuration scheme is selected. A variety of scenario simulation verifications are performed on IEEE 30 power distribution system. The results show that the optimal configuration scheme obtained by the model can effectively reduce the interval width of the interval state estimation results, which proves the practicability, correctness and superiority of the configuration method.