The moisture control of sintering compound is one of the very important links in the sinter production process, and there are mainly such control difficulties as large lag, multivariable interference and data fluctuation. By analyzing the water mixing process and characteristics of sintering compound, a multi-stage feed-forward combined periodic gradient optimization control strategy is proposed according to the different working conditions of primary and secondary mixing granulation. Combined with the current ingredient composition and air permeability index and other key parameters, the BP neural network model is used to predict the optimal moisture rate suitable for the current sintering, and the operating states such as shutdown, start-up and mode switching are identified from the perspective of continuous production, and finally the whole process of water mixing is automated instead of manual control. Simulation results and industrial applications show that the performance of the control method proposed is better than that of conventional PID control, and the sintering automation level can be significantly improved after the system is put into operation.