The fundamental way to solve voltage overrun in a high penetration distribution system of distributed generation is to implement reasonable distribution network planning. The existing methods have some problems, such as too large an initial domain search radius and separation of dynamic time division and topology optimization. A semi invariant method is introduced to deal with the uncertainty caused by the dynamic behavior of distributed generation, and then distribution network reconfiguration is realized using an improved NSGA-II algorithm. A probabilistic constrained optimization strategy and solution method for distribution network reconfiguration considering the dynamic behavior of distributed generation is proposed. In particular, based on grid level optimization, the idea of integrating optimization results with optimal voltage parameters at the time level is proposed to construct the dynamic reconfiguration mathematical model of the distribution system. Finally, the NSGA-II algorithm is improved by a nested gene loop search strategy and the above optimization model is solved to solve the problem that the strong correlation random variables are less considered in traditional distribution network planning. The effectiveness of the proposed method is verified on an IEEE-33 bus distribution system. This work is supported by Beijing Natural Science Foundation (No. 3152016) and Science and Technology Project of State Grid Jibei Electric Power Company (No. 5201061801QL).