Abstract: Nitrogen oxides （NO_x）, generated during the combustion process of coal-fired boilers, constitute a significant source of atmospheric pollutants.A lot of research has focused on achieving stable and effective NO_x emission control technologies, which continue to be refined and improved.The generation of NO_x in coal-fired boiler fluctuates significantly with the complex combustion process while the measurement system exhibits obvious latency. In order to solve the problem above, this article proposes a NO_x prediction model based on combustion and NO_x generation mechanism. This model divides different sections of the boiler according to combustion process. Historical data of coal, air, power and calorific value in different sections of boiler are considered in this model to predict the generation of NO_x in real-time. The results show that this mechanism-based model reduce the amount of input parametersis and it is successfully adapt to complex situation and its operation is uncomplicated which means it could be used in the original system directly. The valve of ammonia operated more effectively with the prediction by mechanism-based model which increase benefit to the power plant.