The extensive participation of thermal power units in peak regulation leads to frequent fluctuations in operating conditions

which increases the difficulty of NO_x emission control at boiler outlet. According to its influence on the combustion

the boiler operation parameters are divided into manipulatable parameters and nonmanipulatable parameters. The data samples are classified according to the non-manipulatable parameter distribution and fuzzy c-means clustering. On this basis

a multi-model method is proposed to predict NO_x emission at boiler outlet. Meanwhile

particle swarm optimization algorithm is used to optimize the manipulatable parameters of the sub-models under each working condition

thus to achieve optimization of the NO_x emission at the boiler outlet. The simulation on operation data of a 1 000 MW coal-fired unit shows that

the proposed multi-model for NO_x emission prediction has higher prediction accuracy and consumes less training time. After optimization by particle swarm algorithm

the NO_x emission at the boiler outlet reduces by 9.98%.