The　working　condition　of　wind　turbine　rolling　bearings　is　always　complex,　therefore　acquired　vibration　signals　are　nonlinear,　non-stationary.　Most　traditional　algorithms　based　on　the　frequency　domain　cannot　fully　extract　intrinsic　information　of　signals.　A　new　method　for　fault　diagnosis　was　proposed　polymerization　based　on　empirical　mode　decomposition(EEMD)　and　kernel　entropy　component　analysis(KECA).　Through　the　EEMD　raw　signal　is　decomposed　into　several　intrinsic　mode　function(IMF),　calculation　of　IMF　energy　and　signal　energy　entropy　to　construct　feature　vectors　as　the　input　of　the　KECA　table,　KECA　classifier　is　built　on　a　fault　monitoring　and　identification.　According　to　the　size　of　the　entropy　feature　extraction　using　KECA,　the　maximum　extent　retained　the　features　of　the　signal　and　strong　ability　of　nonlinear　processing,　which　can　realize　fault　classification　and　recognition　more　effectively.　Finally,　the　results　of　experimental　analysis　showed　that　the　proposed　method　can　effectively　extract　sensitive　features,　also　demonstrated　that　the　diagnosis　accuracy　of　the　proposed　model　based　on　EEMD-KECA　that　is　better　than　that　based　on　neural　network　and　wavelet　energy　entropy　methods.　©　2017,　Editorial　Board　of　Acta　Energiae　Solaris　Sinica.　All　right　reserved.