In　view　of　the　problem　of　overvoltage　identification　in　the　current　distribution　network　of　power　system,　a　temporary　overvoltage　classification　and　recognition　method　of　distribution　network　based　on　time-frequency　distribution　characteristics　was　proposed.　In　neutral　isolated　power　system,　the　three-phase　voltage　of　bolted　single-phase　ground　overvoltage　has　the　lowest　energy　distribution　uniformity;　the　zero-sequence　voltage　of　intermittent　arc　grounding　overvoltage　has　the　highest　DC　component;　and　the　zero-sequence　voltage　of　different　kinds　of　temporary　overvoltage　has　great　differences　on　amplitude　and　frequency　concentration　band.　The　energy　contribution　rate　of　the　zero-sequence　voltage,　the　mean　value　of　the　zero-sequence　voltage,　and　the　singular　spectrum　entropy　were　calculated　to　extract　the　time-domain　energy　distribution　features.　Local　characteristic-scale　decomposition(LCD),　Hilbert　transform　and　the　band-pass　filter　algorithm　were　adopted　to　calculate　the　center　gravity　frequency　band,　so　as　to　extract　the　frequency-domain　energy　distribution　features.　Finally,　the　types　of　temporary　overvoltage　were　recognized　by　the　threshold　discrimination　method.　The　method　does　not　need　classifier,　and　it　has　the　advantages　of　simple　algorithm　and　little　computation　time.　Under　different　fault　conditions,　higher　recognition　rate　was　obtained　both　in　simulation　and　reality　testing.　©　2018,　Harbin　University　of　Science　and　Technology　Publication.　All　right　reserved.