The　waveform　similarity　of　line　zero-sequence　currents　after　the　single-phase　grounding　fault　of　resonance　grounding　system　is　analyzed　based　on　its　zero-sequence　network　and　a　method　of　faulty　line　selection　based　on　the　waveform　feature　clustering　in　the　time　domain　is　proposed　for　it.　The　line　zero-sequence　current　waveform　of　the　first　post-fault　half-period　is　decomposed　in　histograms　and　the　relative　entropy　matrix　is　adopted　to　reflect　the　state　difference　among　the　zero-sequence　currents,　as　well　as　their　polarity　information.　Combined　with　the　amplitude　information,　an　integrated　relative　entropy　matrix　is　established　to　represent　the　time-domain　waveform　feature　of　transient　zero-sequence　currents.　The　fuzzy　KFCM　(Kernel　Fuzzy　C-Means)　without　the　threshold　setting　is　used　to　detect　the　faulty　line.　The　proposed　method　is　verified　by　the　simulations　under　different　working　conditions　and　the　results　show　that,　it　is　immune　to　the　arc　fault,　noise　disturbance,　non-synchronized　sampling　and　compensation　degree.　©　2015,　Electric　Power　Automation　Equipment　Press.　All　right　reserved.