Study On Mixing Model Of Coal Water Slurry Gasifier Based On Mechanism And Data Driven

The mathematical modeling of gasifiers is an important link in the design and optimization of the gasification process.In this paper,taking the SE gas-bed gasifier of coal water slurry in ZhenHai Refining and Chemical as the object,using object-oriented programming method,the mechanism model,data-driven model and mixed model of gasifier are developed,and the applicable scenarios and limitations of each type of model are explored.Material,reaction and reactor objects related to the coal gasification process were established.The SE coal-water slurry gas equilibrium model and the n-CSTRs model are constructed by the free combination module.After considering the influence of transformation reaction in the process of chilling,The prediction errors of equilibrium model and n-CSTRs model are 4.78%,5.75%,8.13%and 3.47%,and 1.29%,2.44%,2.02%and 1.17%,respectively,for the gasification temperature and the contents of CO,H2 and CO2 at the outlet of washing tower.Although the prediction performance of the n-CSTRs model is better,the complexity and solution time of the model also increase correspondingly.Performance fluctuations may occur in the prediction of BPNN,SVM and ELM single data-driven models.Therefore,the combined information entropy weight distribution method and Stacking fusion method are proposed,and the information entropy Stacking fusion(SFM)model of gasifier is established.When the SFM model predicts the pre-processed sample set,The predicted average relative error values(MRE)of gasification temperature,syngas temperature,flow rate,CO and H2 content at the outlet of the washing tower were 1.89%,0.17%,0.78%,0.95%,0.71%,superior to the single data-driven model,and the fitting speed was improved by nearly 20%.This model can be better applied to guide the real-time optimization of production processes with stable operating states.When the SFM model was used to predict the samples screened in the pretreatment process,the MRE of gasification temperature and the contents of CO,H2 and CO2 of syngas were 2.43%、49.49%、16.94%、64.38%,which were abnormal.Therefore,attempts have been made to combine it with mechanism model with better stability.Since the solution time of the n-CSTRs model is longer than 1h,the equilibrium model with a solution time of less than 10s is selected and combined with the SFM model to build a parallel mixing model for the gasifier.When the model is applied to the prediction of the pre-processed sample points,the MRE values for gasification temperature,syngas CO,H2 and CO2 content are 0.35%、0.18%、0.08%、0.46%;When the model is applied to the prediction of sample points removed by pretreatment,the MRE values are:3.82%、3.83%、2.78%and 1.91%,overcoming the limitations of the datadriven model for application scenarios,the prediction accuracy is close to that of the n-CSTRs model,while the solution speed is faster.