钴基催化剂F-T合成的人工神经网络模拟

Artificial neural network simulation of cobalt-based catalyst F-T synthesis

反应温度、压力、空速和原料气H_2/CO比等工艺操作条件对F-T合成生成重质烃的选择性影响很大。以上述操作参数为输入变量,CO的转化率、甲烷在产物烃中的质量分数C_1和重质烃的质量分数C_5^+为输出变量,采用LM算法建立了钴基催化剂F-T合成的BP神经网络模型,定量预测工艺操作条件对F-T合成的影响规律。预测结果表明,低温有利于重质烃生成,高温下CO的转化率高,但C_1也高,C_5^+重质烃的选择性较低。压力升高,C_1下降,CO的转化率和C_5^+增加。C_1随空速的提高而增加,CO的转化率和C_5^+随空速的升高而下降。低合成气H_2/CO比CO转化率和C_1较低,C_5^+重质烃高。进一步的实验验证表明,模型具有较高的预测精度,CO转化率和C_5^+的相对误差小于8%,C_1小于9%。

The reaction temperature, pressure, space velocity and H2/CO molar ratio in feed markedly influence the heavier hydrocarbon selectivity of Fiseher-Tropseh synthesis. Therefore, using operation conditions as input variables, the CO conversion, the selectivi- ty of C1 and C5^＋ as output variables, a BP neural network is constructed by Levenberg-Marquardt algorithm, and wish that quantifieationally predict the regulation of operation condition effect for Fiseher-Tropseh synthesis. The predict results indicated that lower reaction temperature facilitated the synthesis of heavier hydrocarbon, the CO conversion was high and the weight fraction of heavy hydrocarbon C5^＋ low under higher temperature; Increasing pressure, the methane selectivity decreased, the CO conversion and C5^＋ increased; With space velocity increasing, the methane selectivity increased, and the CO conversion and C5^＋ dropped down; Lower H2/CO molar ratio in feed caused low CO conversion and methane selectivity, and high C5^＋. Further experimental verification suggested that the construeted model have higher prediction precision, the relative deviation less than 8% for the CO conversion and C5^＋ , less than 9% for methane.