According to the traditional distribution network reconfiguration, the fault feeder with distributed generation (DG) will separate from the distribution network immediately while the network goes wrong. In order to improve the system power supply reliability and the utilization rate of DG, the new standard allows the distribution network change into island operation. This paper establishes the mathematics model of the distribution network reconfiguration with DG, the objective function and constraint conditions. The traditional genetic algorithm (GA) has the shortcomings of premature convergence and slow convergence speed to solve this nonlinear optimization problem. This paper applies the cloud genetic algorithm (CGA) to solve the network reconfiguration problem. The Case study on IEEE33 test system shows that the algorithm is reasonable and effective.