Study On Petroleum Coke Extraction Desulfurization And Prediction The Calcined Smoke Based On Neural Network

The use of high-sulfur coke to produce aluminum will have a serious impact on equipment,atmosphere,and product quality.Therefore,it is of practical significance to reduce the sulfur content of petroleum coke and control the sulfur emissions of flue gas.In view of the fact that the sulfur in petroleum coke is organic sulphur,this paper adopts organic solvent extraction desulfurization.The o-chlorophenol and furfural with relatively good desulfurization effect were selected as the extractant.The effects of extraction temperature,extraction time,particle size and solid-liquid ratio on the desulfurization rate were investigated.The results showed that the best desulfurization rates of ortho-chlorophenol and furfural were 20.1%and 14.8%,respectively.The optimum desulfurization rate of the composite solvent consisting of o-chlorophenol and furfural was 23.2%,which was higher than the desulfurization rate of a single solvent.After pretreatment of petroleum coke with microwave,the desulfurization rate increased.Among them,composite solvent was the best for desulfurization.The desulfurization rate for composite extraction was 35.3%,which was 12.1%higher than that without microwaves.In order to study the correlation between petroleum coke calcining flue gas parameters and SO₂ emission concentration,the statistical analysis method was used.The results showed that the oxygen content,flow rate and NO_x concentration of flue gas are negatively correlated with the SO₂ emission concentration,while there is no significant correlation between flue gas temperature,pressure and dust concentration of flue gas and SO₂ emission concentration.Establishing the traditional double hidden layer BP neural network model and hidden layer combination BP neural network model to predict the desulphurization rate change of petroleum coke calcined flue gas desulfurization system.Both the two BP neural network models have good prediction results.After comparing the two evaluation indexes,it is found that the combined hidden layer BP neural network was better than the traditional BP neural network to some extent.This shows that by changing the desulfurization system parameters can be quantitatively controlled petroleum coke calcined flue gas desulfurization efficiency,thereby effectively reducing the calcined flue gas sulfur emissions.