A reconfiguration model for the distribution network with the optimization objective of reducing three -phase disequilibrium is suggested in order to cope with an increasingly serious three -phase unbalance in distribution network. Second, the distribution network reconfiguration problem is transformed into a problem of constructing the spanning tree of the graph by evaluating the distribution network in topology, which is solved by the process of breaking -cycle -basis. Then an improved Binary Particle Swarm Optimization (BPSO) algorithm is suggested to solve the reconfiguration problem by randomly choosing the first-branch and canceling the heuristic value of the network, which will extend the search scope and prevent search stagnation. Relevant examples of verification show that, relative to traditional approaches, the proposed algorithm can achieve the best global solution with less computational time and greater probability. The proposed algorithm is tested in 33 bus system and the results show the load balancing in distribution network.