This paper combines the characteristics of a distribution network grid, analyzes the topology and application scenarios of a typical grid, and puts forward the design requirements for the optimization of radial DC distribution network grid structure. Based on the constraints of grid optimization, a Pareto multi-objective planning method is selected, and a multi-objective genetic algorithm is designed to improve the grid reliability index of the distribution network grid planning steps. Analyze the parameters of the example, encode the initial 110kV distribution network frame, carry out the optimal solution of the multi-objective genetic algorithm, compare the iteration number and convergence of different genetic algorithms, and obtain the best economic effect of the optimization of the distribution network frame structure. Analyze the power supply reliability index of distribution networks with different grid switching schemes and explore the degree of reliability influence on grid switching and power supply capacity. The optimal solution is obtained in 10 iterations of the algorithm, that is, the objective function is 65.2421 million yuan, and the optimal solution is reached in the interval of the number of iterations [0,100], and the fitness value is in the range of 2 × 104 million yuan.