浆态床费托合成反应神经网络建模与遗传算法优化  

Slurry F-T synthesis neural network model and genetic algorithms optiaml

作　　者：胡立舜[1] 于广锁[1] 王辅臣[1] 于遵宏[1] 

机构地区：[1]华东理工大学洁净煤技术研究所,272信箱上海200237

出　　处：《计算机与应用化学》2007年第5期633-635,共3页Computers and Applied Chemistry

摘　　要：在实验基础上,利用神经网络对浆态床费托合成反应温度、压力、流量和氢碳比等操作参数及CO转化率、H2转化率和合成气转化率建立三层神经网络模型,通过BP算法对其训练,并将试验值和神经网络仿真值对比,说明了所建神经网络是精确可靠的。同时将神经网络模型嵌入到遗传算法当中,将操作参数编制成21位二进制编码,通过遗传算法对费托合成反应条件进行优化,分别将CO转化率、H2转化率和合成气转化率作为遗传算法的适应度函数,计算出一系列的优化操作参数群体,为研究费托合成操作条件与产物之间的关系提出了一个新的思路。Neural network is used to model three layers neural network with the operation parameters such as temperature, pressure, flow rate and ratio of H2 and CO in slurry column F-T synthesis as input layer and conversions of CO, H2 and syngas as output layer on the basis of experiment. The neural network is trained by BP arithmetic and the neural network is proved exactly and reliable by comparison of the simulation result and the experimental result. Then the F-T synthesis operation parameters is optimized by embedding the neural network into genetic aJgorithms with the conversion of CO, H2 and syngas as adaption function of genetic algorithms and changing the operation parameters into twenty-one bit binary code in order to work out a series of optimal parameters.

关 键 词：浆态床 费托合成 神经网络 遗传算法 

分 类 号：TQ02[化学工程] TQ018