By making full use of the existing production and operation data of the power plant, and integrating mechanism modeling, experimental research, mathematical statistics, and machine learning methods, calculation models for boiler combustion under variable working conditions and optimal searching of controllable operation parameters, as well as a set of boiler combustion optimization control system, have been developed. This realizes the closed - loop control of boiler combustion, achieves the goal of tapping the equipment performance space and improving the operation level of the power plant, and is of great significance for the digital and intelligent construction of thermal power units.