Accurate　prediction　of　electric　bus　energy　consumption　is　a　key　step　to　realize　the　orderly　planned　charging　of　electric　buses.　Meanwhile,　to　address　the　problem　that　the　current　electric　bus　energy　consumption　prediction　model　is　not　conducive　to　realistic　application,　this　paper　proposes　an　energy　consumption　prediction　model　that　considers　actual　electric　bus　operation　data　to　predict　trip　energy　consumption.　First,　based　on　the　operation　data　of　six　routes　in　Beijing,　the　influencing　factors　of　electric　bus　energy　consumption　are　summarized,　including　route　name,　travel　direction,　weekday　and　nonweekday,　operation　time,　vehicle　number,　and　driver＇s　name.　Secondly,　the　energy　consumption　influencing　factors　were　used　to　extract　trip　energy　consumption　features,　including　departure　moment　features,　vehicle　performance　features,　and　driver　attribute　features.　A　new　simple　method　is　proposed　to　deal　with　un-ordered　characteristic　data　to　solve　the　problem　of　quantifying　the　influencing　factors.　The　energy　consumption　prediction　model　considering　actual　quantifiable　features　utilizes　the　concept　of　distance　to　identify　several　historical　trips　that　have　characteristics　most　similar　to　the　predicted　trip　in　terms　of　energy　consumption.　The　new　prediction　model　is　essentially　a　machine　learning　model　based　on　k-means　clustering　algorithm,　which　leverages　feature　extraction　and　data　analysis　to　make　predictions.　Finally,　the　real　data　are　used　to　predict　the　energy　consumption　of　different　routes　and　different　driving　directions　on　weekdays,　respectively.　The　energy　consumption　prediction　error　is　as　low　as　7.112%,　and　the　prediction　results　are　compared　with　other　traditional　prediction　models,　and　the　model　accuracy　is　high.