To solve the problems of large lag, uncertainty and gas disturbance of main steam pressure control system object of gas-fired power boiler,a control scheme based on mismatch compensation Smith prediction and RBF neural network is designed. The RBF neural network's online learning ability is used to adjust the parameters of the conventional PID, and the mismatch compensation Smith prediction controller to compensate the pure hysteresis in the system.The improved algorithm effectively solves the problem of mismatch of dynamic characteristic model and pure lag of main steam pressure object in thermal power boiler. The simulation research and practical application show that the control method has good stability and anti-interference ability for the main steam pressure control of thermal power boilers.