Based On Artificial Neural Network To Simulate Coal Environment Co < Sub > 2 < / Sub > Catalytic Gasification Of Research

Coal gasification is an important link of clean coal technology, and the traditional coal gasification is in the high temperature and high pressure, the reaction conditions are very harsh. Added catalyst in pulverized coal can not only reduce the gasification reaction temperature, also improve the reaction rate. Because there are many influence factors in coal gasification process with catalytic and the complexity of reaction mechanism, which lead to the result that the laws between reaction conditions and product performance are hard to summarize, and so can not to set up the mechanism of accurate model. In order to solve this problem, we use active carbon instead of coal in CO2catalytic gasification process and artificial neural network method to establish forecasting model, which provide certain method and basis for reaction conditions optimization and catalyst screening in coal catalytic gasification process.In this paper, three forecasting models were established based on experimental data of CO2catalytic gasification in the simulated coal environment.In different particle size AC1catalytic gasification prediction model, the network inputs are catalyst type, catalyst content and activated carbon particle size, network outputs are gasification efficiency, gasification peak and reaction index. The test sample errors both multiple-output and single-output forecast models are less than5%. Contrasted the multiple-output forecast model with single-output forecast model, we find that the performance of single-output forecast model is better than that of multiple-output forecast model whether the maximal relative error or average relative error. Based on the model analysis, it is found that the type and content of catalyst play more important roles constrasted with activated carbon particle size.In different catalysts AC1catalytic gasification prediction model, the network inputs are type and concentration of cation and anionic, and catalyst melting point, network outputs are gasification efficiency, initial gasification temperature and reaction index. By comparing convergence speed and forecasting effect, the training goal of0.005BP neural network has the best effect in four models. Based on the model analysis, it is found that type and concentration of catalyst anion have more important effects, so we should also pay attention to the choice of anion in catalyst screening. In order to reduce the quantity, we can forecast gasification process of different catalyst concentration by using established model and choose the best concentration; And ion combination forecast model was established using the existing experimental data, and the result shows that the maximal relative errors of train and test samples of three single-output forecast models guarantee in5%scope, which meet the practical requirement, so it is feasible that Ion combination forecast mode use for prediction of reaction result.In different catalytic gasification prediction model of activated carbon, we select specific surface area, Micropore surface area, most probable pore size and the total volume of activated carbon, type and content of catalyst as inputs, gasification efficiency, initial gasification temperature and reaction index as outputs. We build multiple-output forecast model and single-output forecast model, the results show that the prediction effect of multiple-output model of BP algorithm is better, and that of single-output forecast model of improved BP algorithm is good. Based on three single-output forecast model, it is found that the physical structure of activated carbon to gasification efficiency, initial gasification temperature and reaction index has a little effect, so it is not the important factor in reaction process and can not taken it into consideration in model design, and simplifying model, improving the network of training speed and the prediction performance.