According to the highly sparse characteristics of the adjacent matrix used for the networks topology storing, its manifestation is improved, and the improved adjacent matrix is applied to node ordering optimization, determination of retrieval information in advance and the formation of node admittance matrix. During the process of factorization, to achieve the nonzero retrieving in the column direction, the stored information of the column direction is added and the corresponding retrieval method is made. According to the relationships between new added branches in node ordering and the nonzero injections in factorization, in the meantime of ordering optimization, the position of the new added elements is recorded and the storage framework is formed. The proposed sparse technology is used in the network impedance scanning and equivalence program, and testing results for six power systems show that, with the increase of system scale, the calculation efficiency of node equations with the proposed method is greatly improved compared with the traditional methods and the NIMSCAN program, and it is applicable to the analysis and calculation for large scale power system.