This　paper　proposes　an　improved　non-dominated　sorting　genetic　algorithm　(NSGA2)-DNSGA2,　with　the　aim　of　preserving　diversity　of　obtained　optimal　solution　and　avoiding　the　original　NSGA2　algorithm　falling　into　local　optimal.　The　proposed　DNSGA2　algorithm　which　introduces　a　differential　mutation　operator　to　replace　the　original　polynomial　mutation　because　the　method　of　differential　local　search　is　helpful　to　the　uniformity　of　Pareto　optimal　solution　set.　The　performance　of　the　proposed　DNSGA2,　NSGA2　and　W-LRCD-NSGA2　(Based　on　left-right　crowding　distance　non-dominated　sorting　genetic　algorithm)　are　compared　via　four　benchmark　functions.　Simulation　results　indicate　that　the　diversity　and　uniformity　of　Pareto　optimal　solution　obtained　by　DNSGA2　are　better　than　the　other　two　algorithms.　©　2015　Technical　Committee　on　Control　Theory,　Chinese　Association　of　Automation.