Abstract: Towards the goals of carbon peaking and carbon neutrality,as a thermal power unit with high energy consumption and high emission,its combustion performance directly affects the power generation efficiency and carbon emission. Optimal control of boiler combustionis important for operation optimization and flexible peak shifting. The outlet air powder temperature control of the coal mill in a power station boiler is subject to significant delays and multiple disturbances,which have a profound impact on the safety and stable operation ofthermal power units. This paper proposed a feedback link in the compensation loop based on the Smith prognostic controller to enhance theanti-interference capabilities and performance of the air powder temperature control of the coal mill. The first-order inertia pure hysteresis model was obtained by Taylor series expansion and the method of coefficients to be determined to obtain the model approximation conversion parameters. The anti-disturbance Smith prediction compensation controller was then designed. A comparison with the single-loopPID control experiment reveals that the Smith prediction compensation control algorithm can effectively compensate for the pure delay laglink of the control object. The overshoot of the unit step response is reduced by 36.5% and regulation time is reduced by 65.8%. In comparison to the classical Smith prediction compensation control algorithm,the anti-disturbance Smith prediction compensation control algorithm demonstrates enhanced anti-disturbance performance and notable improvements in the dynamic characteristics of the control object.The overshoot of the unit step response is reduced by 44.7%,and the regulation time is shortened by 15.7%. At the same time,a reductionin the maximum deviation of the perturbation and the convergence time is achieved. The method of this paper was applied to a 660 MW supercritical primary reheat thermal power unit to realize the smooth tracking of the actual value of the air powder temperature at the outlet ofthe coal mill to the set value. The effectiveness of this paper′s method was further verified.