The invention provides a classification and partitioning extract method of power grid model data and a system. The method includes the steps of sequentially analyzing objects in source CIM XML files; judging whether a classification file which each object corresponds to exists; building the classification file of the type if the classification file does not exist; adding data of the object to the classification file if the classification file exists; analyzing root type files in the classification file, wherein the type of the object is a transformer substation; building a CIM XML partitioning file based on each transformer substation object; writing each transformer substation object into a corresponding transformer file; building a mapping table of object codes and the partitioning files; analyzing all the objects in all the classification files according to type processing sequences; searching codes of an upper-layer object corresponding to the current object; and then adding the current object to the partitioning file corresponding to the upper-layer object in a writing mode. The classification and partitioning extract method and the system can quickly extract large quantities of power grid model data in classifying and partitioning modes.