To　ensure　the　connectivity　and　service　quality　of　the　power　communication　network　and　reduce　the　hazards　caused　by　node　failures,　a　reliability　prediction　method　of　the　power　communication　network　based　on　node　importance　is　proposed.　First,　to　integrate　the　dynamic　and　static　characteristics　of　nodes,　the　service　importance　calculation　method　is　studied　from　the　perspective　of　dynamic　services,　and　the　GAT-based　node　topology　importance　evaluation　algorithm　is　studied　from　the　perspective　of　static　topology.　Then,　a　comprehensive　node　importance　calculation　method　based　on　FAHP　is　proposed　to　fuse　the　dynamic　and　static　importance　to　achieve　a　more　comprehensive　node　importance　evaluation.　Next,　a　Bi-LSTM-based　reliability　prediction　method　for　the　power　communication　network　is　proposed.　To　predict　the　reliability　of　the　network　more　accurately,　the　method　combines　node　importance,　extracts　valid　fields　and　formulary　mapping　for　real-time　monitoring　data,　and　fuses　multi-dimension　data　based　on　information　entropy　technology　to　enhance　the　generalization　capability　of　the　Bi-LSTM　prediction　model.　The　experimental　results　show　that　the　proposed　node　evaluation　algorithm　can　accurately　identify　the　important　nodes　that　have　a　greater　impact　on　the　network　reliability　compared　with　the　existing　literature.　Meanwhile,　the　proposed　prediction　algorithm　can　accurately　predict　network　reliability　and　keep　abreast　of　the　changing　trend　of　network　reliability.　©　2023　IEEE.