A　models-3　community　multi-scale　air　quality　(CMAQ)　modeling　system　is　widely　applied　to　air　quality　issues　in　recent　years.　Emission　inventory　is　an　important　input　data　for　the　CMAQ　model.　A　nonlinear　optimizing　system　based　on　genetic　algorithms　(GAs),　which　includes　four　modules:　emission　inventory　adjusting,　population　initializing,　GAs,　and　CMAQ　result　analyzing,　is　developed　under　the　Linux　system　for　optimizing　the　emission　inventory　of　the　CMAQ　model.　The　system　is　used　to　optimize　the　emission　inventory　of　Beijing　in　typical　days.　The　improved　emission　inventory　is　applied　to　simulate　Beijing＇s　PM10　concentrations　of　January,　April,　July,　and　October　in　2002.　The　mean　relative　errors　between　the　monitoring　and　the　simulation　values　are　reduced　by　2.6%,　7.02%,　14.07%　and　2.17%　separately.　This　indicates　that　the　nonlinear　optimizing　system　based　on　genetic　algorithms　is　an　effective　method　to　improve　the　emission　inventory　for　the　CMAQ　modeling　system.