Abstract: It is difficult to control drum water level of supercharged boiler because of the strong load disturbance in voyage process. In order to ensure the stability of drum water level, it is necessary to study the control method of drum water level under strong load disturbance. In this paper we propose a sliding mode controller with improved adaptive genetic algorithm (AGA). We take the velocity and accelerated speed of drum water level error into s-function to design the sliding mode controller. And prove its stability by using Lyapunov stability theorem. On this basis, sliding mode controller parameter was optimized with improved AGA. We compare the control performance of sliding mode controller with improved AGA and PID controller. In response to the ramp disturbance signal and step disturbance signal, sliding mode controller with improved AGA can track the input signal without error. Its stability time is 5s shorter than PID, and overshoot is also less than PID. The simulation results show that sliding mode controller with improved AGA is better than PID controller.