Urban underground cable well, as an important infrastructure, contributes to the performance of urban functions and the coordinated urban development. However, the underground pipeline's complex structure and huge amount of data present great difficulty in displaying, management and analysis. With the development of LiDAR technology characterized by high-speed, non-contact, high-density, and high-precision, point cloud data are gradually introduced into the field of building reverse reconstruction. The threedimensional reconstruction of underground cable wells based on point cloud data has gradually gathered momentum. An algorithm for topological reconstruction of underground cable well model with holes is proposed, which can overcome the problem of model incompleteness and modeling inefficiency when point cloud as data source is used to build triangular network directly. A new hybrid model-CSG-BREP (constructive solid geometry-boundary representation) topological model is adopted to construct the 3D model, which different from the commonly used pipe-lined underground pipe network structure. In this new model, the wells are divided into walls, cables, wellbore, et al, this model can more carefully express the internal topological structure of the underground cable well. Experimental results show that the proposed algorithm can reconstruct many types of walls topologically, and the point-surface relationship between pipe holes and walls and the line-surface relationship between pipes and walls are added into the point-line and line-surface relationship of traditional pipe networks, which allows for a more accurate and detailed description.Through the test of the cable management project, the model shows good results.