基于机器学习的钴基费托催化剂性能探究  

Theoretical Insights of Cobalt-Based Fischer-Tropsch Catalysts Based on Machine Learning

作　　者：黄梦圆 刘冰 刘小浩 Huang Mengyuan;Liu Bing;Liu Xiaohao(School of Chemical and Material Engineering,Jiangnan University,Jiangsu Wuxi 214122)

机构地区：[1]江南大学化学与材料工程学院,江苏无锡214122

出　　处：《化工时刊》2023年第5期1-5,共5页Chemical Industry Times

基　　金：国家自然科学基金(21802054);江苏省自然科学基金(BK20180587)。

摘　　要：本研究使用机器学习方法对钴基费托合成催化剂相关文献数据进行分析,研究催化剂结构及反应条件对费托反应活性的影响。收集了近年钴基费托合成催化剂相关文献,对催化剂组成及其物理性质、制备条件、评价条件进行统计。基于机器学习方法,采用不同回归模型对数据进行分析。结果表明随机森林算法对数据的拟合程度最高,R2值达到0.984。特征重要性分析表明,催化剂中Co3O4颗粒直径对反应选择性影响最高。部分依赖图表明较小的Co3O4粒径有利于C2~C4的选择性,反之则有利于C5+产物的选择性。本研究为进一步理解钴基费托合成催化剂的结构与性能关系提供了理论依据。In this study,machine learning method was introduced to analyze the literature data that related to cobalt-based Fischer-Tropsch catalysts,and to study the relationship between catalysts structure and activity.The relevant literatures of cobalt-based Fischer-Tropsch catalysts in recent years were summarized,and the composition,physical properties,preparation conditions and evaluation conditions of the catalyst were analyzed.Based on machine learning method,different regression models were used to analyze the data.The results show that the random forest algorithm has the highest fitting degree to the data,and its R^(2) value reaches 0.984.The feature importance analysis shows that the diametre of Co_(3)O_(4) particals has the highest importance on the reaction selectivity.The partial dependence plot shows that smaller Co_(3)O_(4) is beneficial to the selectivity of C_(2)-C_(4) product,while the higher content is beneficial to the selectivity of C_(5)+product.This work provides theoretical insights for further understanding the relationship between structure and performance of cobalt-based Fischer-Tropsch catalysts.

关 键 词：机器学习 费托合成反应 钴基催化剂 

分 类 号：O643[理学—物理化学]