A secondary apparatus fault diagnosis method based on data excavation, comprises the steps of (1) reading power grid fault event information and the history information about the impact to the apparatus; (2) deducing and summing up diagnosis rule for filling a rule base through a data excavation inference module; (3) associating and classifying the information like remote measurement and remote signaling of a supervision control and data acquisition system (SCADA) read by a data analysis processing module at real time, fault information system (FIS) protection and switch motion and filtering the information irrelevant to the fault; (4) gathering the related fault information for analysis and positioning the apparatus that has a fault preliminarily; (5) diagnosing the processing information corresponding to the fault apparatus positioned preliminarily by an integrated diagnosis module; (6) gathering and analyzing the matched knowledge base rule and the related fault information to position the fault apparatus; and (7) providing the fault reason and the strategy for eliminating the fault in combination with the matched knowledge base rule in dependence on the positioned fault apparatus.