Abstract: With a large number of distributed generations (DGs) based on renewable energy connected to the distribution network, the penetration of DGs in distribution network is increasing gradually. However, influenced by natural factors, some DGs such as the photovoltaic (PV) and wind power are obviously time-varying and fluctuant. When the penetration of such DGs is at a higher level, its power fluctuation will have a great impact on the operation of the network and bring the challenges to restoration process after failure isolation. Thus, it’s of great significance to research the dynamic restoration problem with multiple uncertain factors in distribution network to improve the reliability and economy of system operation. The traditional restoration strategies keep the DG and load power unchanged in the calculation process, which only present the static restoration model for a certain time breakpoint. Considering the distribution network with DGs is a time-varying and uncertain system, such research methods cannot meet the real-time demand of distribution network restoration. Faced with the disadvantages of the traditional scheme, most of researches will establish the description for the uncertain factors in the restoration process and convert the uncertain problems into several certain problems to solve. For example, some researches consider the worst case of DG and load power fluctuation in the restoration, which makes the restoration scheme feasible at any periods. However, this restoration scheme is too conservative and not globally optimal. To realize the optimization and dynamic adaptability of long time restoration process in distribution network, some researches divide the restoration period into several small periods and get the static schemes based on power predicted value in every period. And the dynamic restoration scheme is obtained according to the time sequence. However, the feasibility of full time optimal restoration scheme depends on the accurate DG and load forecasting, and the scheme may not be applicable to the PV and wind power with large power fluctuation. In view of the demand for long-time restoration in distribution network and considering the power fluctuation of DG and time-variance of load, this paper proposes an interval number method based dynamic restoration strategy considering DG and load power prediction in the fault distribution network, and the interval number method is used to describe the uncertainty of power fluctuation. Then the dynamic restoration model for fault distribution network is proposed. According to the principle of whether main network and DGs can meet all load power supply during the recovery period, the first objective function is set to maximum restored loads and the second objective function is set to minimize network operation cost considering switch motion cost and network loss in the restoration strategy. To make the dynamic restoration scheme meet the constraint of switch times, a dynamic time division method based on dynamic restoration parameters and loss parameters is proposed. The improved decimal particle swarm optimization algorithm and load cutting strategy based on optimal recovery path is proposed to solved restoration math model. Finally, simulation results of IEEE-33 nodes show that the proposed dynamic restoration strategy can automatically divide the optimal segmented restoration scheme, which ensures the safety and reliability of the distribution network after failure and effectively reduces the network operation cost.