Nowadays,　society　pays　much　attention　to　the　problems　of　fuel　consumption.　This　paper　concerns　about　prediction　of　microcosmic　energy　consumption,　and　its　purpose　is　to　realize　fuel　consumptions　of　Beijing　basic　freeway　section.　Based　on　OBD/GPS　terminal　installation　on　taxis,　we　extract　driving　behavior＇s　data　of　taxi　drivers,　select　main　relevant　indexes,　set　up　the　prediction　model　of　fuel　consumption,　and　realize　accurate　prediction　of　fuel　consumption　in　Beijing　basic　freeway　section.　Results　show　that　average　speed,　standard　deviation　of　speed,　max　speed,　rate　of　operating　condition,　average　acceleration　and　deceleration,　distance　and　energy　have　greater　influence　on　fuel　consumption;　PCA　and　neural　network　combination　model　can　realize　energy　consumption　prediction　effectively,　and　the　accuracy　of　prediction　can　reach　92.46%.　This　research　can　provide　strong　supports　on　monitor　and　regulation　of　traffic　energy　consumption.　Copyright　©　2016　by　Science　Press.