Arc　fault　voltage　signal　of　load　terminal　in　low-voltage　is　not　affected　by　the　singularity　signal　of　power　line　to　bring　about　fault　misjudgment.　An　arc　fault　experimental　platform　was　built　with　references　to　the　United　States　standard-UL1699.The　experiment　was　conducted　to　collect　a　large　number　of　typical　load　arc　fault　signal.　Firstly,　the　characteristics　of　arc　fault　signal　intrinsic　mode　function　(IMF)　components　were　extracted　effectively　by　using　empirical　mode　decomposition　(EMD).　Secondly,　with　analysis　of　the　contribution　rate　of　IMF　variance,　the　front　five　orders　IMF　was　taken　to　reflect　various　load　arc　fault　characteristic　information.　Finally,　an　arc　fault　identification　model　for　different　loads　based　on　extreme　learning　machine　(ELM)　was　put　forward,　whose　input　vectors　is　the　IMF　component　ratio　of　energy　for　front　five　orders.　The　experiment　and　simulation　results　show　that　the　arc　fault　diagnostic　method　with　the　combination　of　EMD　and　ELM　identifies　arc　fault　for　various　loads　effectively.　©　2016,　Harbin　University　of　Science　and　Technology　Publication.　All　right　reserved.