With　the　development　of　new　energy　technologies　and　business　models,　the　operation　and　monitoring　methods　of　the　Energy　Internet　(EI)　have　changed.　The　Energy　Digital　Twin　Network　(ED　TN),　which　integrates　EI　and　Digital　Twin　(DT),　has　become　an　important　innovation　tool　for　the　energy　industry.　DTs　provide　the　possibility　to　transfer　physical　entities　to　virtual　environments,　as　well　as　the　ability　to　predict　future　states　through　DTs.　Both　service　data　and　synchronous　data　exist　in　EDTN,　for　this　reason,　the　periodic　fluctuation　of　service　data　affects　synchronous　data　transmission,　resulting　in　degraded　synchronization　performance　of　EDTN.　For　better　analysis　and　management,　this　paper　proposes　a　transformer　based　classified　traffic　prediction　(TBCTP)　scheme　for　EDTN.　Firstly,　we　design　the　data　transmission　architecture　of　EDTN　based　on　Software　Defined　Network　(SDN)　and　use　the　global　vision　and　centralized　control　characteristics　of　the　SDN　controller　to　realize　the　intelligent　control　of　ED　TN　traffic.　Then,　we　classify　short-period　traffic　with　different　characteristics　so　that　traffic　of　the　same　class　has　similar　characteristics.　After　that,　we　design　a　long-period　traffic　prediction　model　based　on　transformer　architecture　and　train　corresponding　prediction　models　for　different　types　of　traffic　to　improve　the　prediction　accuracy　of　the　model.　The　experimental　results　show　that　TBCTP　outperforms　other　baseline　models　in　predicting　accuracy.　©　2023　IEEE.