The　staged-combustion　burner　design　was　optimized　to　reduce　the　environmental　pollution　due　to　NOx　emitted　by　the　boiler　combustion.　The　staged-combustion　model　was　based　on　artificial　neural　network　and　simulation　evolvement　algorithm　for　a　boiler　with　a　100　MW　turbine-generation　unit.　The　boiler　performance　was　optimized　for　16　design　parameters　and　7　regulating　parameters　that　affect　the　combustion.　The　analysis　considered　boiler　loads　of　100%,　90%,　80%　and　70%　which　required　11523,　14810,　13410　and　19732　neural　network　training　steps　for　the　training　values　to　meet　the　mean　square　deviation　requirement.　The　optimized　design　gave　a　species　amount　of　80,　a　cross　probability　of　0.8,　and　a　variation　probability　of　0.15.　The　results　show　that　the　relative　errors　in　the　boiler　efficiencies　and　NOx　emissions　between　the　calculated　and　measured　results　are　less　than　1%,　and　that　the　average　NOx　output　of　the　boiler　decreases　from　812　mg/m3　to　645　mg/m3.