At present,the original drawing information of the substation is stored in a single way,and the degree of digital analysis is not high. In order to reduce the huge workload of manually analyzing the original drawing of the substation,a method is proposed to intelligently extract the information of the equipment and connection relationship of the substation based on the portable document format (PDF) drawing for building a physical circuit model,thus implementing the method according to the smart substation physical configuration description (SPCD) file. Firstly,the primitive information is extracted and processed from drawings. Then electrical symbol recognition is realized through string similarity matching running Karp-Rabin greedy string tiling (RKR-GST) algorithm,and substation images are classified by gradient boosting decision tree-logistic regression (GBDT-LR) hybrid algorithm based on features. Finally,the digital description from the original drawing to the physical circuit model is completed according to the SPCD file. Experiments show that the correct rate of electrical symbol matching is 93%,and the correct rate of physical circuit identification is more than 90% when there are errors in primitives.