Abstract: More and more Dispersed Generation has been interconnected with rural distribution networks and rural distribution networks commonly operate with a radial topology. Dispersed Generation is complicating the operation and protection schemes. But a radial distribution network is the basis of operation, protection schemes and dispatch in a rural distribution network. The models of distribution network reconfiguration should consider radial topology in their formulation. So the paper presented radial constraints in a mathematical model for a distribution network Reconfiguration and gave a preliminary analysis of radial constraints. Concerning an interconnecting Distributed Resource with an Electric Power System, mathematical models based on mixed integer nonlinear programming (MINLP) for the rural distribution network reconfiguration were built. The objective function was power losses of the distribution system. The binary nature of variables was proposed to represents the circuit between buses i and j connecting or not. Considering interconnecting with a substation for Dispersed Generation, a new constraint has been added to ensure that a distributed generator is not isolated from the substation. The mathematical model for distribution network reconfiguration problem was one in which the radial constraints were represented explicitly and showed that radial constraints can be considered in a simple and efficient way. With this method, distribution networks interconnecting as a Distributed Resource are still regarded as a passive termination of the distribution network with a radial structure. The paper proposed the use of a branch - bound method and Primal-dual interior point method to solve power dispatch and optimization reconfiguration problem of a distribution power system. The models of distribution network reconfiguration were formulated as a mixed integer non-linear programming (MINLP) problem. The MINLPIS relaxed, resulting in a set of non-linear programming (NLP) problems, which are solved at each node of the branch - bound tree through a Primal-dual interior point algorithm. The non-linear branch - bound algorithm proposed has special fathoming criteria to deal with non-linear and optimization reconfiguration models. The fathoming tests are redefined, adding a safety value to the objective function of each problem before they are fathomed through the objective function criteria, avoiding a convergence to local optimum solutions. From the result of the case study, it can be seen from the 33 nodes radial distribution test system that DG has the effects of a loss reduction improvement over feeders in the case, and the distribution system without DG are different from those with DG. With the increasing of capacity and numbers of DG, there has been a significant loss reduction and high efficiency of a distribution system with distributed generation (DG). The presented method has been compared with a GA algorithm based on the 54 nodes radial distribution test system The results show that solo transfer node can be identified by the presented method whereas by GA algorithm. The last case showed the total CPU time used by the branch - bound and Primal-dual interior point method and the number of binary variables for different test systems. The results showed that the presented method in this paper needs more time to obtain a global solution, but with the development of computer technology, the proposed method is a promising and effective tool for a distribution network with DG.