A new method based on graph ant system is proposed for solving dynamic distribution network reconfiguration problem to improve the economy of distribution network. This method divides the solving process into an off-line stage and an on-line stage. The off-line stage analyzes the distribution network’s topology and forms the feasible solution set. The feasible solution set is expressed in structure map and contains both topological structure of the distribution network and the length of time interval. The on-line stage searches the best solution in the way of simulating the behavior that ants search for food in the structure map. Topological analysis is implemented only once during the off-line stage. During the on-line stage, a time extension method is used to reduce the feasible solution set to a much smaller set that contains better solutions. Simulation results of an example show that the proposed method can effectively reduce the searching scope to find the best solution.