The　rise　of　dual　carbon　goals　and　Web　3.0　has　spurred　the　rapid　development　of　blockchain-based　microgrids(BBMGs).　Due　to　the　inherent　volatility　in　power　generation　within　microgrids　composed　of　distributed　renewable　energy　sources,　frequent　cross-domain　interactions　between　multiple　BBMGs　are　required　to　ensure　stable　system　operation.　As　an　important　stage　of　cross-domain　interaction,　trusted　authentication　is　crucial　to　its　development.　Currently,　cross-domain　trusted　authentication　is　generally　achieved　by　setting　a　unified　reputation　calculation　method　and　cross-domain　validation　standard　for　all　nodes　in　the　multi-microgrid　system.　However,　this　method　identifies　malicious　nodes　through　periodic　and　static　reputation　calculations　but　cannot　detect　or　prevent　nodes　that　suddenly　become　malicious　in　real　time.　Moreover,　the　reputation　calculations　typically　do　not　consider　factors　such　as　node　importance　and　resource　conditions,　which　is　not　ideal　for　BBMG　systems　with　limited　resources.　Additionally,　cross-domain　authentication　in　the　above　approach　is　usually　performed　by　selecting　the　most　trusted　node　through　polling　or　random　methods　for　single　validation,　or　by　using　multiple　trusted　nodes　for　consensus-based　validation.　These　methods　are　inefficient　and　unsuitable　for　microgrid　systems　that　require　frequent　interactions.　To　address　these　challenges,　we　first　propose　an　abnormal　behavior　real-time　detection　method　based　on　federated　learning　to　monitor　the　abnormal　behavior　of　nodes　in　real　time.　Then,　factors　such　as　abnormal　behavior　detection　and　resource　conditions　are　introduced　to　propose　a　reputation　calculation　method　that　is　more　suitable　for　blockchain　systems.　Based　on　this　method,　we　design　a　multi-factor　trustworthy　user　access　control　mechanism　to　assign　roles　and　permissions　of　nodes,　identifying　multiple　trusted　cross-domain　validation　nodes(CDVNs)　for　efficient　parallel　authentication.　To　assist　the　system　in　quickly　selecting　and　locating　the　optimal　CDVNs　and　further　improve　the　efficiency　of　cross-domain　trusted　authentication,　this　paper　proposes　an　efficient　search　and　location　strategy　for　the　optimal　CDVNs　based　on　the　source　node　to　batch　search　for　the　CDVN　with　the　smallest　delay　for　all　nodes.　Experimental　results　show　that　this　scheme　is　feasible　in　practical　applications.　©　2014　IEEE.