A non-revisited NSGA-II algorithm, which integrates BSP (binary space partitioning)-based non-revisiting mechanism and the non-dominated sorting genetic algorithm-II (NSGA-II) based on the ideology of multi-objective Pareto optimization, is constructed and is then used to solve the problem of distribution networks multi-objective reconfiguration with the consideration of loss minimization and the improvement of reliability. The proposed algorithm achieves a strictly non-revisited research, which avoids the recalculation of power flow and reliability of revisited schemes and saves the computing resources. The test results of IEEE 16-bus and IEEE 33-bus sample systems indicate that the proposed method can obtain the optimal solution in every target direction as well as an optimal Pareto frontier aggregates including a lot of non-dominated solutions in less iterations. According to incidence relations between network losses and reliability goals, along with the topology construction analysis of corresponding reconfiguration schemes, conclusions can be drawn that network losses are in apparent consistency with reliability goals in global scope of solution space, and whether out of the consideration of network loss optimization or reliability optimization, the topology of networks should be close to breadth-first tree rather than depth-first tree. This work is supported by National Natural Science Foundation of China (No. 51467012 and No. 51167012).