Abstract: In order to reduce the nitrogen oxide (NO_x) emission of the boiler combustion system, a prediction model of NO_x emission was established with the stacked denoising auto-encoder (SDAE) by using the real historical operation data collected from a 1000MW thermal power unit, and then an optimization strategy for boiler air distribution and coal powder distribution based on an improved sparrow search algorithm (SSA) was proposed. To improve the optimization ability of SSA, a chaotic optimization sparrow search algorithm with firefly perturbation (FCOSSA) was proposed. The FCOSSA used Tent chaotic mapping to make the initial individuals as evenly distributed as possible to increase the diversity of the initial population, and then updated the positions of all sparrows through firefly disturbance. The superiority of FCOSSA was proved by optimization tests with representative benchmark functions. For a given load steady-state operation condition, the presented method was employed to optimize the boiler air distribution and coal powder distribution with the goal of reducing NO_x emission. The results show that the NO_x emission can be effectively reduced after optimization, which verifies the effectiveness of the method.