Electricity　consumption　analysis　and　prediction　are　of　great　significance　for　power　grid　planning　and　the　management　of　enterprise　resources.　Based　on　the　historical　power　consumption　data　collected　by　the　cloud　platform　of　Fujian　Huatuo　energy　management　system,　this　paper　analyzes　the　power　consumption　of　the　three　floors　of　Fujian　Huatuo　Automation　Technology.　The　results　show　that　the　biggest　influencing　factors　are　weather　conditions　and　working　days.　Observing　the　fact　that　the　pattern　of　power　consumption　has　a　great　difference　when　the　temperature　grows　larger　than　a　threshold,　this　paper　proposes　a　clustering-based　prediction　mechanism　to　predict　the　power　consumption　with　different　models.　The　experimental　results　show　that　the　proposed　mechanism　can　effectively　improve　the　prediction　accuracy　of　power　consumption.　©　2019　IEEE.