The　effluent　quality　of　a　wastewater　treatment　plant　(WWTP)　is　restricted　by　many　factors.　However,　the　activated　sludge　mathematical　model　(ASM),　which　is　only　based　on　biological　reaction　mechanism,　does　not　involve　the　factors　other　than　biological　reactions.　Therefore,　It　will　bring　inaccuracy　in　real　forecast.　To　solve　this　problem,　the　author　put　forward　a　“black　box　model”　which　is　based　on　data　mining　technology　to　simulate　and　predict　a　WWTP　operation.　First,　the　actual　WWTP　data　and　process　parameters　were　fitted　by　using　neural　network　model.　Second,　the　markov　chain　was　applied　to　fit　the　results　and　error　state　division　to　further　improve　the　prediction　accuracy.　Finally,　BP　neural　network　and　Markov　chain　were　combined　to　analyze　and　forecast　a　large　WWTP　nitrogen　removal　efficiency.　The　results　show　that　the　BP　neural　network　can　be　used　to　simulate　the　WWTP　nitrogen　removal　process,　and　the　accuracy　and　reliability　of　the　simulation　can　be　improved　when　combined　with　the　Markov　chain.　©　2016,　Science　Press.　All　right　reserved.