Abstract: To address the problems of high carbon emissions and severe environmental pollution in traditional coal-fired power units, the potential of coal ammonia mixed combustion technology in emission reduction and pollution mitigation was deeply studied. The mechanism of NO_x(nitrogen oxides) generation was explored, and a prediction model for the NO_x concentration at the inlet of the SCR denitration system was established. Based on the model, a control method for SCR denitrification system was proposed. Firstly, by analyzing the mechanism of NO_x generation during ammonia doped combustion, the characteristic variables influencing NO_x concentration variations were determined. Principal component analysis method was used to optimize the characteristic variables, and a long short-term memory (LSTM) network prediction model was established. Furthermore, The particle swarm optimization algorithm combined with annealing idea (SA-PSO algorithm) was used to optimize the model's structure parameters, thereby improving the accuracy of predicting the SCR inlet NO_x concentration. Secondly, a feedforward cascade control method for the SCR denitrification system was designed based on the prediction model and compared with other traditional control methods. Finally, simulation verification was conducted using the operation data of a 600 MW coal-fired unit under two conditions: 100% load with 10% ammonia doped ratio and 50% load with 20% ammonia doped ratio. The results indicate that the constructed prediction model can effectively predict the NO_x concentration at the SCR denitrification inlet, providing reliable data support for the control of the denitration system and enhancing both the denitration efficiency and stability of the SCR denitrification system.