As an important factor affecting the quality of sinter and ironmaking cost,a combined forecasting model of return fine amount constructed by deep learning algorithm is proposed to solve the problem that its data is difficult to obtain in a short time based on few related prediction studies. Firstly,the big data of the actual sintering production process is combined with the data preprocessing technology to establish the basic data set of the model. Then,the hysteretic nature in the sintering process is combined with the time-domain neural network to realize the early prediction function of the model. At the same time,the Pearson correlation analysis and the sintering accessory rules formulated by PCA technology are combined with deep forest to realize the real-time monitoring function of the model. The analysis of prediction results shows that the overall error range of the model(return fine amount within 15 t/h)can reach more than 90%,and show good forecast in advance and real-time monitor effect,which can achieve the goal of tredicting the trend and amount of return fine.