This paper focuses on the combustion optimization to cut down NO_x emission with a new strategy.Firstly, orthogonal experimental design(OED) and chaotic sequences are introduced to improve the performance of particle swarm optimization(PSO). Then, a predicting model for NO_x emission is established on support vector machine(SVM) whose parameters are optimized by the improved PSO. Afterwards, a new optimization model considering coal quantity and air quantity along with the traditional optimization variables is established. At last,the operating parameters are optimized by the improved PSO to cut down the NO_x emission. An application on 600 MW unit shows that the new optimization model can cut down NO_x emission effectively and maintain the load balance well. The NO_x emission optimized by the improved PSO is lowest among some state-of-the-art intelligent algorithms. This study can provide important guides for the low NO_x combustion in the power plant. This paper focuses on the combustion optimization to cut down NO_x emission with a new strategy. Firstly, orthogonal experimental design (OED) and chaotic sequences are introduced to improve the performance of particle swarm optimization (PSO). Then, a predicting model for NO_x emission is established on support vector machine (SVM) whose parameters are optimized by the improved PSO. Afterwards, a new optimization model considering coal quantity and air quantity along with the traditional optimization variables is established. At last, the operating parameters are optimized by the improved PSO to cut down the NO_x emission. An application on 600 MW unit shows that the new optimization model can cut down NO_x emission effectively and maintain the load balance well. The NO_x emission optimized by the improved PSO is lower among some state-of-the -art intelligent algorithms. This study can provide important guides for the low NO_x combustion in the power plant.