Stroke　tripping　fault　caused　by　lightning　is　the　main　fault　of　high　voltage　transmission　line.　Currently,　the　overvoltage　fault　caused　by　lightning　is　mainly　treated　by　lightning　location　and　data　recording　of　the　traveling　wave　technology.　And　the　effective　intelligent　technology　about　processing　lightning　fault　is　deficient.　Firstly,　we　established　a　simulation　model　for　the　back-flashover　and　lightning　shielding　failure　of　the　over-voltage　of　high-voltage　transmission　line　by　using　ATP-EMTP.　Secondly,　we　put　forward　a　method　about　fault　feature　extraction　for　empirical　mode　decomposition　(EMD)　combined　with　the　correlation　dimension　of　fractal,　which　was　applied　in　the　fault　feature　extraction　for　lightning　overvoltage.　Then　the　first　four　IMFs　were　determined　by　variance　contribution　rates　for　overvoltage　signal　IMF　components　to　reflect　the　main　feature　information　of　fault　overvoltage.　Finally,　we　set　up　the　diagnosis　and　recognition　model　for　the　lightning　overvoltage　fault　of　high-voltage　power　transmission　line　by　the　application　with　extreme　learning　machine　(ELM).　The　simulation　results　show　that　the　various　lightning　overvoltage　faults　such　as　hitting　tower,　lightning　wire,　and　transmission　phase　lines,　are　identified　effectively　by　ELM　on　the　basis　of　the　method　for　fault　feature　extraction　combined　EMD　with　the　correlation　dimension　of　fractal.　©　2016,　High　Voltage　Engineering　Editorial　Department　of　CEPRI.　All　right　reserved.