In order to improve the utility boilers' economy and reduce NOxemission as much as possible

this paper regards the minimum coal consumption rate and NOxformation as the goals to establish the prediction models of coal consumption rate and NOxformation for a sub-critical 300 MW unit boiler

on the basis of the support vector machine(SVM)theory.Combining with the operation data of this 300 MW unit

the above two models were verified.Moreover

the established coal consumption rate prediction model was coupled with the NOxformation prediction model

thus a combustion optimization model was formed.Both the coal consumption rate and NOxformation were considered as the optimization object to set up the objective function.Then the genetic algorithm(GA)was employed to find out the optimal operation solution for the boiler.It concludes that the SVM-based boiler combustion model has high ability in predicting the unit coal consumption rate and NOxformation.It is often conflicting to reduce NOx generation as well as coal consumption rate

namely increasing the over fire air volume can reduce the NOx generation

but at the same time will increase the coal consumption rate.Aiming at optimizing the boiler operation

the coal consumption rate and NOxgeneration should be integrated as the optimization goal.For this sub-critical 300 MW unit boiler burning bituminous coal

it's better to run at full load for the lowest NOxgeneration

and at other loads

the unit economy and NOxemission can achieve the best situation when the secondary air ratio and the opening degree of each layer of the secondary air baffles are kept at a certain value.