乙醇偶合制备C4烯烃工艺条件的核偏最小二乘建模及其混合优化

Kernel Partial Least Squares Modeling of Process Conditions for the Preparation of C4 Olefins by Ethanol Coupling and Its Hybrid Optimization

摘要: 探究乙醇偶合制备C4烯烃的工艺条件具有重要意义。利用核偏最小二乘回归的方法研究温度、催化剂组合及乙醇浓度对于乙醇转化率，C4烯烃选择性的影响，和偏最小二乘回归方法进行比较，模型的预测精度有明显提高。在此基础上采用基于遗传算法的模拟退火粒子群算法求解最佳工艺条件，该算法相较于遗传算法，模拟退火算法及粒子群算法在收率速度和准确性上都有更好的表现。最终结果得到在温度450度、Co的负载量0.5、Co/SiO₂和HAP装料比2.03、Co/SiO₂和HAP质量和400 mg、乙醇浓度0.3 ml/min的条件下，C4烯烃收率最大值为47.5%。

Abstract: It is important to investigate the process conditions for the preparation of C4 olefins by ethanol coupling. This paper investigates the effect of temperature, catalyst combination and ethanol concentration on ethanol conversion and C4 olefins selectivity using Kernel Partial Least-Squares Regression. Compared with the Partial Least-Squares Regression method, the prediction accuracy of the model is significantly improved. Based on this, Genetic Annealing Particle Algorithms is used to solve the optimal process conditions. The algorithm has better performance in terms of yield speed and accuracy compared to genetic algorithm, simulated annealing algorithm and particle swarm algorithm. The final result was that a maximum C4 olefins yield of 47.5% could be obtained at a temperature of 450 degrees, a Co loading of 0.5, a Co/SiO₂ and HAP loading ratio of 2.03, a Co/SiO₂ and HAP mass sum of 400 mg, and an ethanol concentration of 0.3 ml/min.