Considering　the　dynamics　and　randomness　of　passenger　demand,　this　paper　investigates　a　train　timetabling　problem　in　the　stochastic　environment　for　an　urban　rail　transit　system.　With　the　scenario-based　representation　of　passenger　distribution,　an　integer　nonlinear　programming　(INLP)　model　is　first　formulated　to　simultaneously　optimize　the　total　number　of　train　services,　headway　settings　and　speed　profile　selection　decision　during　the　planning　time　horizon,　in　which　the　expected　total　service　cost　is　treated　as　the　objective　function.　Through　an　analysis　of　the　features　of　the　nonlinear　constraints,　a　reformulation　method　is　proposed　to　develop　an　equivalent　integer　linear　programming　(ILP)　model　that　can　be　easily　solved　by　commercial　software.　Moreover,　a　variable　neighborhood　search　algorithm　is　developed　to　find　the　approximate　optimal　solutions　for　large-scale　problems　within　the　tolerable　computing　time.　Finally,　two　sets　of　numerical　experiments,　with　the　operation　environments　of　a　simple　urban　rail　transit　line　and　Fuzhou　Metro　Line　1,　are　implemented　to　verify　the　solution　quality　and　effectiveness　of　the　proposed　methods.　©　2021　Elsevier　Ltd