Federated　graph　learning　combines　the　power　of　graph　learning　and　the　data　privacy-preserving　properties　of　federated　learning　to　provide　a　new　solution　for　processing　complex　graph　data　scattered　across　different　data　sources.　However,　data　heterogeneity　may　not　only　lead　to　model　performance　degradation　but　also　affect　the　stability　and　convergence　of　model　training.　In　addition,　the　local　model　has　to　consider　the　data　security　problem　in　the　process　of　delivery　and　aggregation.　Therefore,　this　paper　proposes　a　Dual-branch　Federated　Graph　Learning　model　with　Global　Graph　Structure　information　(DFGL2GS)　to　cope　with　the　statistical　heterogeneity　of　graph　data　distribution　and　the　secure　interaction　of　graph　structure　information.　Specifically,　the　local　model　contains　two　branches,　Local　and　Global,　which　are　used　to　capture　personalized　local　features　and　generalized　global　features,　respectively.　The　server　further　assists　global　federated　graph　learning　modeling　by　capturing　global　graph　structure　information　from　a　client　generated　graph　that　incorporates　node　and　graph　structure　information.　Finally,　we　demonstrate　the　effectiveness　of　the　proposed　method　by　conducting　extensive　experiments　on　four　datasets　of　different　domain.　©　2025　Elsevier　B.V.