Abstract: Coal-fired units are faced with the dual pressures of flexible operation and ultra-low emissions. The rapid and deep load change of the units puts forward higher requirements for the control of the SCR (Selective Catalytic Reduction) denitrification system. This paper proposed a multi-objective optimization control method that takes into account ultra-low emissions and economic costs. The cost of denitration was added to the optimization objective function, and a predictive control structure was adopted. The neural network and genetic algorithm were also separately used to establish the model and optimize the control value to achieve the ammonia injection optimization. The simulation results show that the method reduces the cost of denitration while ensuring the emission standards, and can adapt to the large-scale variable operating conditions of the boiler.