In order to control NOxemissions and enhance boiler efficiency in coal-fired boilers

the thermal operating data from an ultra-supercritical 1000 MW unit boiler were analyzed.On the basis of the support vector regression machine(SVM)

the fruit fly optimization algorithm(FOA)was applied to optimize the penalty parameter C

kernel parameter g and insensitive loss coefficient of the model.Then

the FOA-SVM model was established to predict the NOxemissions and boiler efficiency

and the performance of this model was compared with that of the GA-SVM model optimized by genetic algorithm(GA).The results show the FOA-SVM model has better prediction accuracy and generalization capability

of which the maximum average relative error of testing set lies in the NOxemissions model

which is only 3.59%.The above models can predict the NOxemissions and boiler efficiency accurately

so they are very suitable for on-line modeling prediction

which provides a good model foundation for further optimization operation of large capacity boilers.