Non-orthogonal　multiple　access　(NOMA)　has　been　recognized　as　an　efficient　means　to　boost　the　spectral　utilization　and　increase　the　number　of　connections　for　the　next　generation　mobile　communications　systems.　User　fairness　is　a　crucial　performance　metric　in　the　design　of　downlink　NOMA　systems,　but　its　optimum　is　reaped　at　the　cost　of　excessive　power　consumption.　In　this　paper,　we　study　the　performance　tradeoff　between　user　fairness　and　energy　conservation　by　means　of　power　allocation,　and　we　formulate　the　problem　as　a　bi-objective　optimization.　To　obtain　the　complete　Pareto　optimal　solutions,　we　transfer　the　above　bi-objective　optimization　into　a　single-objective　one　using　the　weighted　Tchebycheff　method　and　solve　the　resulting　non-convex　optimization　with　the　generalized　Benders　decomposition.　In　addition,　we　investigate　several　important　properties　of　the　proposed　mechanism,　including　the　convergence,　computational　complexity,　and　applicability　to　the　case　with　imperfect　successive　interference　cancellation.　Based　on　the　Monte　Carlo　simulations,　we　validate　the　necessity　of　conducting　the　tradeoff　optimization　and　the　efficiency　of　the　proposed　mechanism.