Fundamental Study On Influence Factors And Prediction Model Of Sintering NO_x And SO₂ Emission

Due to the rapid development of China’s iron and steel industry,environmental problems have become increasingly serious.However,the sintering process is an important part of the iron and steel industry,and it is also the main source of air pollutant emissions from iron and steel enterprises.In this paper,the effects of fuel characteristics on the sintering process and flue gas emissions were explored through sintering pot experiments.The results show that NO_xin the sintering flue gas mainly comes from fuel.During the process of increasing the proportion of pulverized coal in the fuel structure,the generation of CO and NO_xin the flue gas gradually increased,and the generation of O₂decreased.At the same time,the combustion layer is narrowed and the duration of high temperature is shortened.These factors lead to a gradual decline in the cold strength index of sinters.When the proportion of pulverized coal in the fuel structure is increased from 75%to 100%,the NO_xemission in the flue gas did not show a significant upward trend.At the same time,a large amount of CO is generated,which promotes the reduction of NO to N₂and inhibits the generation of NO_x.Under the condition that the proportion of pulverized coal in the fuel structure is 100%,as the proportion of fuel particle size less than 1 mm gradually increases,the concentration of NO_xin the flue gas gradually increases.However,with the shortening of sintering time,the total NO_xemission decreased.During the sintering process,SO_xis generated by the decomposition and oxidation reaction of sulfate and sulfide in the combustion zone.In the emission process,SO_xis gradually absorbed by material layer to form Ca SO₃,and the unabsorbed part is discharged with the flue gas.In the drying preheat zone,the Ca SO₃is re-decomposed,and the pyrolyzed sulfur oxides are also absorbed by the lower material layer.The thicker the lower material layer,the better the adsorption effect.The change mechanism is controlled by the migration mechanism of S element in the material layer:First,sulfide and sulfate are thermally decomposed to generate SO₂,then the material layer adsorbs SO₂,and finally re-decomposition and SO₂desorption occur.These processes continue to cycle until the sintering end point is approached.Due to the large hysteresis,high coupling and strong nonlinearity of the sintering process,this paper developed a deep neural network and a least squares support vector machine to predict the SO₂and NO_xemissions during the sintering process.The model was trained based on the actual production data of sintering plant,and the influence of fuel characteristics,including industrial analysis,elemental analysis,and particle size distribution,on the predictive ability of the model was investigated.The results show that the DNN has better generalization ability than the LSSVM,and the prediction accuracy is improved by more than 30%.The data set with fuel characteristics can effectively improve model prediction accuracy,and RMSE is reduced by 4.73mg/m³（26.5%）in the NO_xprediction,MAE decreased by 2.84 mg/m³（52.1%）.In addition,after extracting and optimizing features by feature technology,the computation time of NO_xprediction is reduced by 38.07%.The model prediction results can handle the online operation and optimization of sintering plant more accurately,and relieve the pressure of the subsequent treatment of flue gas.