The predictive control technology based on online model identification is used to replace the traditional “feedforward + PID feedback” control mode to realize AGC control and coordinated control between the boiler and the turbine. At the same time, factors influencing steam temperature like unit load, flue gas volume, coal feed rate, secondary air volume, feed water temperature are considered and identified to form a disturbance model. With this model, lag and stability problems will be solved. Besides, tracking performance of the unit will be improved so that important parameters such as steam pressure and steam temperature will be better controlled.