An improved genetic algorithm is adopted to operate distribution network structure layout automation. From the point of view of removing the collection of to-be-selected lines and rationally utilizing planning information for optimization， this paper presents a new real-number encoding based on to-be-planned load nodes number and a new decoding based on short distance feeder line layout to carry out the feeder line automatic layout. Such improved genetic operations as fragment reordering crossover， displacement mutation operator and theμ+λ selector are used to search for the optimal radiative structure. Moreover， according to the principle of closed-loop design and open-loop distribution network operation，the planning model takes into account the heuristic tie-line wiring scheme on the basis of the optimal distribution network to obtainhand in hand” distribution network that can enhance power supply reliability. It is shown by a case study that the algorithm can greatly reduce the workload of a planner and yield a satisfactory solution in a relatively small number of iterations. Meanwhile， it is shown by calculation results that the algorithm is feasible and of practical value.