Abstarct: The advancement of science, economic problems, environmental pollution and diminishing of fossil fuels are factors that has led human to use of the renewable energies and distributed generation resources (DG). Thus, the experts have tried to study the behavior of power systems in the presence of DGs, because the inappropriate use of these resources in power network not only will not have any advantage, but also it can have negative impact on distribution systems and electricity quality of the customers. A distribution networks is designed as a looped system, but it is operated as a radial system. The connection of DGs to distribution systems can affect greatly on current and line power and voltage magnitude of the customers. The reliability of distribution systems is one of the aims that have attracted the attention of many researchers because the reliability improvement can be economic for distribution companies. Also, it brings the customers satisfaction. System configuration and DG location are the factors that improve reliability and also reduce power losses in distribution networks. System configuration and DG location are related to each other. it means that for different configurations of distribution system, different locations for DGs are to be find and vice versa. Therefore, the distribution planners and operators should consider the impact of configuration and DGs location simultaneously to improve reliability and power loss reduction In this thesis firstly, it is tried to find a method for reliability calculation in the presence of DG in distribution systems is described. Then, by using reconfiguration and DG allocation simultaneously, one of the reliability indices, i.e. ENS (Energy Not Supplied) and power loss reduction are improved. Since, the reliability and loss reduction are related inversely, a multi objective optimization algorithm is needed to improve these two factors simultaneously. The used algorithm in this thesis is non dominated sorting genetic algorithm (NSGA II) that shows solutions of problem in terms of Pareto front. Finally the proposed method is implemented on two test systems for two different scenarios: without DG and with DG, to demonstrate performance of the proposed method. Results of the simulation for the two different scenarios show the impact of DGs on power distribution network as well as effectiveness of the proposed algorithm.