As　a　complicated　and　combinatorial　optimization　problem,　the　design　of　Water　distribution　systems　(WDSs)　is　difficult　to　be　solved.　In　this　paper,　in　order　to　design　WDSs　better,　three　objectives　are　considered,　including　the　initial　construction　cost,　the　sum　of　node　surplus　head　and　the　variance　of　node　surplus　head.　The　consideration　of　the　variance　of　node　surplus　head　can　reflect　the　distribution　of　the　node　surplus　head　in　WDSs.　This　problem　is　solved　by　the　strength　pareto　evolutionary　algorithm　(SPEA2).　Finally,　the　model　is　applied　into　two　well-known　benchmark　case　studies,　the　two-loop　network　and　the　New　York　Tunnels　network.　By　comparing　with　some　algorithms,　such　as　Tabu　search　algorithm　and　non-dominated　sorting　genetiv　algorithm　(NSGA2),　the　performance　of　SPEA2　shows　the　effectiveness　in　terms　of　reliability,　convergence,　as　well　as　diversity.