In order to achieve optimal operation of distribution network better, a method of dynamic distribution network reconfiguration based on membership partition of time intervals is proposed. This paper establishes the mathematical model of dynamic distribution network reconfiguration and uses the minimum of comprehensive operation cost as the target function. This paper uses the ratio of active network loss and voltage offset each period as the operation index of the system, and calculates the peak and valley membership grade as well as the rate of change of the index proposed to partition time intervals. The new particle acception mechanism based on Mertropolis norm of simulated annealing (SA) is introduced into the binary particle swarm optimization (BPSO) so as to enhance the ability of getting rid of local solution and overcome the limitations of premature convergence of BPSO. Results of calculation example show that the proposed method is effective.