With the gradual promotion of the "unit based" planning method for distribution networks, regional distribution networks have been divided into several relatively independent power supply units. However, there are still a number of interconnecting lines within the power supply units, which still has a certain degree of complexity and makes the mapping of power supply units more difficult. Heuristic automated mapping methods based on rules and force orientation are inefficient and relies on manual intervention, which cannot adapt to the complex and changing distribution network application scenarios. For this reason, this paper proposes an automatic mapping method for medium-voltage distribution network power supply units based on the mean square error condition generation adversarial network. The layout generator and genetic mutation algorithm in this method can generate and optimize the layout and connection of distribution network nodes at a fine-grained level, achieving automatic mapping at various node scales. At the same time, this paper designs an evaluation function for node layout generators, which takes topology visualization effects such as node clustering degree, line crossing, and inflection points as key evaluation indicators. It can be used to iteratively optimize the layout generator and thereby improve the graphical effect. The experimental results show that the proposed method outperforms other heuristic layout methods in terms of mapping performance.