The invention discloses a GIS partial discharge severity determination method and system based on principal component dimension reduction and clustering analysis and a medium. The method comprises steps: acquiring a partial discharge spectrum of a to-be-detected GIS partial discharge signal; extracting a key characteristic quantity for the partial discharge spectrum; performing dimension reductionprocessing on the key characteristic quantity to obtain a dimension-reduced partial discharge characteristic fingerprint; respectively calculating distances between the partial discharge characteristic fingerprint and characteristic fingerprint clustering centers, obtained by clustering on partial discharge characteristic fingerprints corresponding to the discharge stages, at corresponding different discharge stages; and determining a discharge stage corresponding to the to-be-detected GIS partial discharge signal according to a shortest distance. With the disclosed method, the severity of defect discharge can be determined effectively while the discharge type identification is completed. The method has advantages of high calculation efficiency and calculation accuracy; and the referencefor maintenance decision and construction period arrangement can be provided.