Aiming at the characteristics of non-linearity and variable model parameters of the main steam pressure control system of gas-fired power generation boiler, an improved implicit generalized predictive control strategy is proposed. First, the forgetting factor recursive least squares method is used to identify the model parameters, and a discrete mathematical model of the main steam pressure is established. An implicit generalized predictive control system for main steam pressure is established on the basis of conventional generalized predictive control theory, which simplifies the control algorithm and improves the robustness of the system by adding a PI structure to the objective function. The simulation results show that: compared with the cascade PID and conventional implicit generalized predictive control, the proposed control strategy reduces the adjustment time by up to 20s when the model is adapted; the overshoot decreases by up to 5.08% and the adjustment time by up to 36s when the model is mismatched Robustness and anti-interference ability are stronger. Engineering application shows that after using the proposed strategy, the main steam pressure control deviation is within ± 0.2Mpa, and the control accuracy is significantly improved.