The　primary　problem　of　distribution　system　protection　is　single-line-to-ground　(SLG)　fault,　particularly　in　networks　with　distributed　generators　(DGs),　where　asymmetrical　phase　sequence　components　caused　by　SLG　faults　may　lead　to　system　instability.　Currently,　the　location　and　suppression　methods　of　SLG　faults　are　based　on　reliable　fault　diagnosis.　Therefore,　considering　the　imbalance　of　line　parameters,　this　paper　analyzes　the　fault　characteristics　of　zero-sequence　voltage　(ZSV),　and　proposes　a　threshold-online-calculating-based　SLG　fault　diagnosis　method,　which　includes　fault　detection,　fault　occurrence　moment　capture　and　fault　nature　estimation.　Firstly,　the　ZSV　is　measured　in　real　time　by　the　designed　feeder　terminal　unit　(FTU),　which　is　embedded　with　the　proposed　SLG　fault　diagnosis　method.　Secondly,　the　ZSV　signal　is　decomposed　by　the　Mallat　algorithm,　and　the　fault　diagnosis　thresholds　are　calculated　through　the　maximum　margin　hyperplane,　which　is　used　for　fault　detection　and　fault　nature　estimation.　Finally,　the　fault　occurrence　moment　is　captured　by　variational　mode　decomposition　(VMD)　and　its　Teager　Energy　Operator　(TEO).　The　proposed　SLG　fault　diagnosis　method　has　been　tested　in　simulation　and　physical　experiments　in　a　10kV　system,　the　effectiveness　and　feasibility　have　been　validated.