Distribution network reconfiguration is essentially a complex multi-objective nonlinear integer combinatorial optimization problem. In order to overcome the influence of the infeasible solution and achieve fast convergence, a fast method for judging whether particles meet topological constraints through chain loops matrix is proposed. A discrete Binary Particle Swarm Optimization (BPSO) algorithm based on the Pareto criterion is proposed to solve the multi-objective optimization problem of a distribution network. The algorithm is improved from three aspects: improving the method for updating particles to increase the feasible probability of the new particles; improving the sigmoid function and proposing a neighborhood search mechanism to enhance the convergence ability of the algorithm in the later stage; proposing a niche sharing mechanism based on the sub-optimal solution retention strategy to improve the method for updating a group's optimal particle, thereby enhancing the algorithm's global search capability. Through the IEEE33 node distribution system, it is verified that the improved BPSO can converge to the Pareto optimal frontier accurately and efficiently when solving the multi-objective optimization problem of a distribution network with Distributed Generation (DG). This work is supported by the Science and Technology Project of the State Grid Corporation of China (No. JSDL- XLFW-SQ-2016-10-092).