A model considering time variation of distributed generation (DG) is proposed here and successfully applied to solving fault recovery problem in distribution network. According to the actual situation that real-time output power of different types of DGs keeps changing over time, multi-period dynamic service restoration based on tensile time scales is considered. A minimum summation of load loss in the restoration period is selected as the main objective function, and the minimum number of switching operations as the secondary objective function. Firstly, the improved binary particle swarm optimization is applied to obtain radial network structure excluding a large number of infeasible solutions. Secondly, the strategy based on optimal recovery path proposed in this paper is used to perform load shedding operation. Lastly, the minimal switch operation problem is solved by enumeration-combination method, then the optimal service restoration switch scheme is eventually obtained. The example shows that maximized service restoration can be achieved with the dynamic coordination between the main grid and DGs. This work is supported by Science and Technology Project of State Grid Corporation of China (No. KJ[2013]896).