In　this　paper,　we　propose　an　adaptive　genetic　algorithm　based　on　a　new　entropy　measurement,　and　deduce　the　limit　of　the　selection　probabilities　of　individuals　under　the　entropy　measurement.　The　theoretical　analysis　and　a　comparative　experiment　show　that　the　new　selection　strategy　based　on　the　new　entropy　measurement　can　adjust　dynamically　the　selection　intensity　according　to　the　population　state.　The　proposed　method　shifts　dynamically　the　balance　between　the　exploitation　and　exploration　performance　of　genetic　algorithms　to　enhance　global　optimal　performance　of　algorithm.　©　2014　IEEE.