In order to achieve high-precision and forward-looking prediction of FeO content in sinter,and support the stable forward operation of blast furnace,charge quality optimization and energy efficiency improvement,a deep learning prediction method that combines temporal and non-temporal information is proposed. Based on the actual production data of a steel mill’s sintering machine,the isolated forest algorithm is used to identify and clean the abnormal samples,and the dimensional differences between different process parameters are eliminated by the normalization method. On this basis,the feature screening method combined with Spearman correlation analysis and maximum information coefficient (MIC) is used to extract 9 key process parameters closely related to FeO content from multiple production variables,and a prediction dataset is constructed. The LSTM-DNN hybrid model integrating long short-term memory network (LSTM) and deep neural network (DNN) is further established to fully explore the time-series-dependent features and high-dimensional nonlinear mapping relationship in the sintering process. The results show that compared with the single LSTM model and DNN model,the proposed LSTM-DNN hybrid model has better prediction performance,with a prediction accuracy of 93. 33% within the ±0. 5 error range,and has high stability and generalization ability. The model has been successfully integrated into the thermal imaging monitoring system of the tail section of the sintering machine,which realizes the online prediction and early warning of FeO content linked with image information and process parameters,and can provide effective technical support for inhibiting the fluctuation of FeO content,improving the quality stability of sinter,and promoting the intelligence of the sintering process.