With the frequent occurrence of extreme natural disasters, in order to make the supporting effect of distributed generation (DG) work, such as gas turbine and fuel cell, a microgrid formation strategy considering dynamic reconstruction under extreme natural disasters is proposed, which increases the value of controllable DG to support critical loads. Firstly, Wasserstein distance is used to calculate the similarity of load probability distribution. And each load is divided into different time scenarios. Then, the current state information of the line is read. And the constraints satisfying the formation problem of radial microgrid are established. Taking maximizing the load recovery value as the objective function, the method of mixed integer linear programming is used to solve the static optimal solution for the formation of microgrid. At last, combined with the state change of loads, the dynamic optimal solution is solved without the limit of reconstruction times to find out the moment of structure change. After that, the limit of the number of reconstruction times is taken into account, and the optimal plan is determined to maximize value of the load in the dynamic process. The results show that the dynamic reconfiguration based on Wasserstein distance improves the electric quantity value of the loads supported by distribution network in extreme environment.