In　recent　years,　with　the　increasing　proportion　of　wind　power　generation,　the　impact　of　wind　power　generation　on　grid　security　is　also　growing.　This　makes　the　prediction　accuracy　of　wind　power　generation　higher　and　higher.　This　paper　utilizes　the　LSTM　model　of　the　deep　learning　domain　to　predict　wind　power　generation.　Besides,　Auto　Encoder　is　employed　to　reduce　the　data　dimension,　improve　the　generalization　ability　of　the　model,　and　shorten　the　training　time.　Simulation　experiments　show　that　the　LSTM　model　has　better　prediction　accuracy　than　other　machine　learning　model　such　as　SVM.　©　2019　Association　for　Computing　Machinery.