基于BP神经网络乙醇偶合制备C4烯烃条件分析

Analysis of the Conditions for the Preparation of C4 Olefins Based on BP Neural Network Ethanol Coupling

摘要: 本文建立了固定效应变截距模型以及固定效应变系数模型研究乙醇转化率与温度的关系，并通过F检验选择固定效应变系数模型，同时对C4烯烃选择性与温度分别进行三次多项式拟合和Fourier拟合，通过平方和误差SSE选择三次多项式拟合，借助RadViz模型研究350℃实验不同时间的测试结果，得到该催化剂组合在20分钟到70分钟不稳定，并采用单因素控制变量法分别分析催化剂总质量、Co/SiO2和HAP的质量比、Co负载量、乙醇浓度、温度对乙醇转化率及C4烯烃选择性的影响，通过主成分分析法选择出了较好的十种结果；其次使用BP神经网络进行仿真模拟借助遗传算法寻找最大值，通过多次实验寻找符合条件的最优方案。最后，借助正交试验设计分析催化剂总质量、Co/SiO2和HAP质量比、Co负载量、乙醇浓度、温度，使用方差分析得出五个因素的主次关系，并根据直观图选出最优的五组实验。

Abstract: In this paper, a fixed-effect variable intercept model and a fixed-effect variable-coefficient model were established to study the relationship between ethanol conversion rate and temperature, and the fixed-effect variable-coefficient model was selected by the F test. Fourier fitting, selecting cubic polynomial fitting by sum of squares error SSE, and using RadViz model to study the test results of the 350˚C experiment at different times, it was found that the catalyst combination was unstable from 20 minutes to 70 minutes, and the single factor controlled variable method was used to ana-lyze them respectively. The effects of total catalyst mass, Co/SiO2 and HAP mass ratio, Co loading, ethanol concentration, and temperature on ethanol conversion and C4 olefin selectivity were se-lected by principal component analysis. The BP neural network is used for simulation and the ge-netic algorithm is used to find the maximum value, and the optimal solution that meets the condi-tions is found through many experiments. Finally, the total catalyst mass, Co/SiO2 and HAP mass ra-tio, Co loading, ethanol concentration, and temperature were analyzed with the help of orthogonal experimental design, and the primary and secondary relationship of the five factors was obtained by variance analysis, and choose the best five groups of experiments according to the intuitive dia-gram.