Photovoltaic　(PV)　power　generation　systems　are　increasingly　installed　on　the　factory　rooftop.　The　installed　PV　power　generation　system　exits　a　two-dimensional　structure　since　the　factories　are　usually　designed　to　be　square　or　rectangular.　Motivated　by　real　industry　practice,　this　research　proposes　a　two-dimensional　performance　sharing　reliability　model　for　the　factory　rooftop　PV　power　generation　system,　where　a　common　bus　exists　in　each　row　and　across　all　rows.　The　component　availability　is　affected　by　both　internal　failure　and　external　impacts.　The　system　reliability　is　the　probability　that　each　component　can　meet　its　own　demand　and　each　row　can　meet　its　required　demand　after　performance　sharing.　To　mitigate　external　impacts,　the　factory　rooftop　PV　power　generation　system　employs　both　overarching　protection　and　individual　protection　measures.　We　extend　the　universal　generating　function　(UGF)　technique　to　propose　a　reliability　evaluation　algorithm　and　use　the　genetic　algorithm　(GA)　to　determine　optimal　protection　resource　allocation　to　improve　system　reliability.　Further,　the　proposed　two-dimensional　performance　sharing　reliability　model　is　extended　to　consider　multiple　phasedmission.　Numerical　studies　indicate　that　both　expanding　common　bus　capacity　and　optimizing　the　protection　allocation　strategy　are　effective　ways　to　improve　the　factory　rooftop　PV　power　generation　system　reliability.