Abstract: This paper proposes a new method to correct the reservoir inflow. The method is based on the moving average algorithm and the weights of the algorithm are auto-adaptive to reservoir inflow. The correction process includes three steps. Firstly, the moving average algorithm with adaptive weights is used to correct the reservoir inflow and to make it more smooth, then the corrected inflow is properly modified to meet the water balance, and finally the result is modified again according to the inflow variation and other constraints. The method in this paper is compared with the moving average algorithm and the cubical smoothing algorithm with five-point approximation. The results show that the method has a better effect on flood peak correction and is able to ensure water balance, and the correction smoothness is similar to that of the cubical smoothing algorithm with five-point approximation.