基于机器学习的加氢裂化分子筛催化材料构效关系研究

doi:10.1016/S1872-2067(24)60259-7

催化学报 ›› 2025, Vol. 71: 187-196.DOI: 10.1016/S1872-2067(24)60259-7

基于机器学习的加氢裂化分子筛催化材料构效关系研究

马千里^a, 聂红^a^,^*(), 杨平^a, 刘建强^a, 高鸿毅^b, 王薇^a, 董松涛^a

^a中石化石油化工科学研究院有限公司, 北京 100083
^b北京科技大学材料科学与工程学院, 北京材料基因工程高精尖创新中心, 分子与微结构可控高分子材料技术北京市重点实验室, 北京 100083

收稿日期:2024-10-26 接受日期:2024-11-22 出版日期:2025-04-18 发布日期:2025-04-13
通讯作者: * 电子信箱: niehong.ripp@sinopec.com (聂红).
基金资助:
国家重点研发计划(2021YFA1501204)

Insights into Structure-Activity Relationships between Y Zeolites and their n-C₁₀ Hydrocracking Performances via Machine Learning Approaches

Qianli Ma^a, Hong Nie^a^,^*(), Ping Yang^a, Jianqiang Liu^a, Hongyi Gao^b, Wei Wang^a, Songtao Dong^a

^aSINOPEC Research Institute of Petroleum Processing Co., Ltd., Beijing, 100083, China
^bBeijing Advanced Innovation Center for Materials Genome Engineering, Beijing Key Laboratory of Function Materials for Molecule & Structure Construction, School of Materials Science and Engineering, University of Science and Technology Beijing, Beijing 100083, China

Received:2024-10-26 Accepted:2024-11-22 Online:2025-04-18 Published:2025-04-13
Contact: * E-mail: niehong.ripp@sinopec.com (H. Nie).
Supported by:
National Key R&D Program of China(2021YFA1501204)

摘要/Abstract

摘要：

加氢裂化是重油高效转化和炼油向化工高质量转型的关键支撑技术之一. 在氢气(H₂)、高温和催化剂的协同作用下, 可将劣质重油转化为轻质清洁燃料油和有机化工原料. 催化剂作为加氢裂化技术的核心, 是影响反应活性、选择性和稳定性的关键因素之一. 加氢裂化催化剂是典型的双功能催化剂, 通常由金属活性中心(加氢功能)、酸性载体(裂化功能)协同作用, 并辅以氧化铝和助剂优化性能. 分子筛是裂化反应的核心载体, 但其复杂结构导致孔分布、B酸量和酸强度等性能参数相互耦合, 因此分子筛结构与加氢裂化反应性能的构效关系一直是领域的研究热点.
本文利用机器学习揭示了Y型分子筛与正癸烷(n-C₁₀)加氢裂化反应性能的构效关系. 首先基于23个Y型分子筛样本及对应的92组NiMo/Al₂O₃-Y催化剂性能数据, 构建了物化性质-反应性能关联数据库. 通过特征相关性分析, 筛选晶胞常数、酸含量、酸强度、B酸/L酸、介孔体积和微孔体积作为结构描述符(自变量), 并以n-C₁₀转化率、产物C₃/C₇比 (表征产物二次裂化程度)和i-C₄/n-C₄比例(表征产物异构化程度)为性能指标(因变量). 采用随机森林算法构建预测模型, 并通过训练集与预测集R²和均方根误差(RMSE)验证模型可靠性. 结果表明, 3种因变量预测集和训练集的R²均接近1, 相关性良好, 训练集的n-C₁₀转化率、C₃/C₇和i-C₄/n-C₄的均方根误差(RMSE)均较低. 实验发现, 转化率主要受晶胞参数、酸量和介孔体积影响; C₃/C₇比主要受晶胞参数、酸量、介孔体积和微孔体积的影响; 对i-C₄/n-C₄比影响较大的变量为晶胞参数、酸量和微孔体积. 基于模型预测一种新型Y分子筛的性能, 其n-C₁₀转化率、C₃/C₇比和i-C₄/n-C₄比的预测值与实验值R²分别为0.9866, 0.9845和0.9922, RMSE值分别为0.0163, 0.101和0.0211, 表明预测结果与实验结果吻合良好, 模型对新催化剂设计具有较好的预测能力.
综上所述, 本文通过机器学习揭示了Y型分子筛结构与正癸烷加氢裂化反应性能的构效关系, 不仅验证了机器学习在加氢裂化催化剂研究中的可行性, 同时为高性能加氢裂化催化剂和技术的研发提供了理论依据.

关键词: 加氢裂化, 机器学习, Y型分子筛, 正癸烷, 酸性, 孔结构

Abstract:

Hydrocracking technology represents a crucial position in the conversion of heavy oil and the transformation development from oil refining to the chemical industry. The properties of catalysts are one of the key factors in the hydrocracking process. As the main acidic component of hydrocracking catalyst, the influence of zeolite properties on the reaction performance has been the focus of research. In this study, a series of NiMo/Al₂O₃-Y catalysts were prepared using different Y zeolites as acidic components, and their performances in the hydrocracking of n-C₁₀ were also evaluated. The structure-activity relationship between Y zeolite and the cracking performance of n-C₁₀ was investigated with machine learning. First, a database of the physical and chemical properties of Y zeolite and their performance was established, and the correlation analysis was also conducted. Parameters such as the cell constant, acid content, acid strength, B/L ratio, mesopore volume, micropore volume of Y zeolite, and the reaction temperature were selected as independent variables. The conversion of n-C₁₀ and the ratios of products C₃/C₇ and i-C₄/n-C₄ were selected as dependent variables. A model was established by the random forest algorithm and a new zeolite was predicted based on it. The results of model prediction were in good agreement with the experimental results. The R² of the n-C₁₀ conversion, C₃/C₇ ratio, and i-C₄/n-C₄ ratio were 0.9866, 0.9845, and 0.9922, and the minimum root mean square error values were 0.0163, 0.101, and 0.0211, respectively. These results can provide reference for the development of high performance hydrocracking catalyst and technology.

Key words: Hydrocracking, Machine learning, Y zeolite, n-Decane, Acid, Pore structure

马千里, 聂红, 杨平, 刘建强, 高鸿毅, 王薇, 董松涛. 基于机器学习的加氢裂化分子筛催化材料构效关系研究[J]. 催化学报, 2025, 71: 187-196.

Qianli Ma, Hong Nie, Ping Yang, Jianqiang Liu, Hongyi Gao, Wei Wang, Songtao Dong. Insights into Structure-Activity Relationships between Y Zeolites and their n-C₁₀ Hydrocracking Performances via Machine Learning Approaches[J]. Chinese Journal of Catalysis, 2025, 71: 187-196.

×
扫码分享

导出引用管理器 EndNote|Ris|BibTeX

链接本文: https://www.cjcatal.com/CN/10.1016/S1872-2067(24)60259-7

https://www.cjcatal.com/CN/Y2025/V71/I4/187

图/表 7

Fig. 1. Studying the basic process of catalyst structure-activity relationship through machine learning.

Fig. 2. Catalyst evaluation results. (a) Conversion versus temperature curve. (b) C3/C7 versus temperature curve. (c) i-C4/n-C4 versus temperature curve. (d) isomerization selectivity versus conversion curve.

Fig. 3. Pearson correlation coefficient analysis of zeolite structure variables. (a) The correlation coefficients among all variables. (b) The correlation coefficient between the selected independent variables.

Fig. 4. Machine learning prediction results of n-decane hydrocracking reaction. (a1-a3) Training set; (b1-b3) Prediction set.

Table 1 Thermodynamic calculation of n-butane isomerization reaction. K = i-C4/n-C4.

T/°C	△H/J·mol^-1	△G/J·mol^-1	K
310	-8203.6	686.5	0.87
330	-8177.0	991.0	0.82
350	-8149.3	1294.5	0.78
370	-8120.8	1597.2	0.74

Fig. 5. Univariate prediction of n-decane hydrocracking reaction using random forest model. Standard conditions: cell parameters 2.4600 nm, acid content 1400 μmol·g-1, acid strength 350 °C, B/L 8.0; mesopore volume 0.08 cm3·g-1; micropore volume: 0.300 cm3·g-1. (a1-a6) Conversion; (b1-b6): C3/C7; (c1-c6): i-C4/n-C4.

Fig. 6. Machine learning prediction results of n-decane hydrocracking reaction for Y24 catalyst. (a) Conversion; (b) C3/C7; (c) i-C4/n-C4.

参考文献

[1]	M. Kazakov, K. Nadeina, I. Danilova, P. Dik, O. Klimov, V. Y. Pereyma, E. Y. Gerasimov, I. Dobryakova, E. Knyazeva, I. Ivanova, A. S. Noskov, Catal. Today, 2018, 305, 117-125.
[2]	C. Peng, X. Fang, R. Zeng, Catalysis, 2016, 28, 86-118.
[3]	R. Saab, K. Polychronopoulou, L. Zheng, S. Kumar, A. Schiffer, J. Ind. Eng. Chem., 2020, 89, 83-103.
[4]	H.-B. Park, K.-D. Kim, Y.-K. Lee, Fuel, 2018, 222, 105-113.
[5]	C. Peng, Y. Du, X. Feng, Y. Hu, X. Fang, Front. Chem. Sci. Eng., 2018, 12, 867-877.
[6]	Z. Cao, Z. Chen, J. Yu, J. Mei, S. Peng, Z. Wu, R. Guo, X. Fang, X. Zhang, Chem. Eng. Sci., 2023, 282, 119292.
[7]	J. E. Schmidt, B. Smit, C.-Y. Chen, D. Xie, T. L. Maesen, ACS Catal., 2023, 13, 6710-6720.
[8]	J. Weitkamp, ChemCatChem, 2012, 4, 292-306.
[9]	C. Xian, Y. Mao, X. Long, Z. Wu, X. Li, Z. Cao, Catalysts, 2022, 12, 848.
[10]	M. Zhang, B. Qin, W. Zhang, J. Zheng, J. Ma, Y. Du, R. Li, Catalysts, 2020, 10, 815.
[11]	Y.-J. Lee, E.-S. Kim, J.-R. Kim, J.-W. Kim, T.-W. Kim, H.-J. Chae, C.-U. Kim, C.-H. Lee, S.-Y. Jeong, J. Nanosci. Nanotechnol., 2016, 16, 4335-4341.
[12]	K. Sato, Y. Iwata, T. Yoneda, A. Nishijima, Y. Miki, H. Shimada, Catal. Today, 1998, 45, 367-374.
[13]	J. Qi, Y. Guo, H. Jia, B. Fan, H. Gao, B. Qin, J. Ma, Y. Du, R. Li, Fuel Process. Technol., 2023, 240, 107586.
[14]	J. L. Agudelo, E. J. M. Hensen, S. A. Giraldo, L. J. Hoyos, Fuel Process. Technol., 2015, 133, 89-96.
[15]	J. A. Esterhuizen, B. R. Goldsmith, S. Linic, Nat. Catal., 2022, 5, 175-184.
[16]	T. Zhou, Z. Song, K. Sundmacher, Engineering, 2019, 5, 1017-1026.
[17]	J. R. Kitchin, Nat. Catal., 2018, 1, 230-232.
[18]	Z. Yang, W. Gao, Adv. Sci., 2022, 9, 2106043.
[19]	X. Bokhimi, Front. Chem. Eng., 2021, 3, 740270.
[20]	L. H. Mou, T. Han, P. E. Smith, E. Sharman, J. Jiang, Adv. Sci., 2023, 10, 2301020.
[21]	K. McCullough, T. Williams, K. Mingle, P. Jamshidi, J. Lauterbach, Phys. Chem. Chem. Phys., 2020, 22, 11174-11196.
[22]	K. Takahashi, I. Miyazato, J. Comput. Chem., 2018, 39, 2405-2408.
[23]	B. MacQueen, R. Jayarathna, J. Lauterbach, Curr. Opin. Chem. Eng., 2022, 36, 100781.
[24]	P. Schlexer Lamoureux, K. T. Winther, J. A. Garrido Torres, V. Streibel, M. Zhao, M. Bajdich, F. Abild-Pedersen, T. Bligaard, ChemCatChem, 2019, 11, 3581-3601.
[25]	Z. Rao, P.-Y. Tung, R. Xie, Y. Wei, H. Zhang, A. Ferrari, T. Klaver, F. Körmann, P. T. Sukumar, A. Kwiatkowski da Silva, Science, 2022, 378, 78-85.
[26]	M. Erdem Günay, R. Yıldırım, Catal. Rev. Sci. Eng., 2021, 63, 120-164.
[27]	M. Yi chao, L. Kun hong, N. Hong, Acta Petrol. Sin. Petrol. Process. Sect., 2009, 25, 31-35.
[28]	L. Hongquan, Acta Petrol. Sin. Petrol. Process. Sect., 2011, 27, 419.
[29]	H. Bierenbaum, R. Isaacson, M. Drui, Ind. Eng. Chem. Prod. Res. Dev., 1974, 13, 2-9.
[30]	J. Weitkamp, P. A. Jacobs, J. A. Martens, Appl. Catal., 1983, 8, 123-141.
[31]	P. B. Weisz, E. W. Swegler, Science, 1957, 126, 31-32.

基于机器学习的加氢裂化分子筛催化材料构效关系研究

Insights into Structure-Activity Relationships between Y Zeolites and their n-C₁₀ Hydrocracking Performances via Machine Learning Approaches

RichHTML

PDF

Supporting Information

可视化

摘要/Abstract

引用本文

使用本文

扫码分享

图/表 7

参考文献

相关文章 15

编辑推荐

Metrics

[1]	刘浩, 张颖, 刘立扬, 陈天翔, 张兴聪, 胡鹏, 熊超, 周杰, 张浩, 董丽辉, 劳子桓, 南兵, 何晓辉, 纪红兵. 利用硅酸锌沸石稳定铂簇以实现高效丙烷脱氢[J]. 催化学报, 2025, 71(4): 208-219.
[2]	焦自浩, 张成翼, 刘亚, 郭烈锦, 王子运. 基于图神经网络的高性能二氧化碳还原高熵合金催化剂预测[J]. 催化学报, 2025, 71(4): 197-207.
[3]	班涛, 王建伟, 余希阳, 田海阔, 高新, 黄正清, 常春然. 机器学习辅助筛选SA-FLP双活性位催化剂用于甲烷与水共转化制甲醇反应[J]. 催化学报, 2025, 70(3): 311-321.
[4]	孙艺萌, 陈军, 刘琳, 池海波, 韩洪宪. 原子级铱掺杂增强γ-MnO₂析氧活性和稳定性的机理研究[J]. 催化学报, 2025, 69(2): 99-110.
[5]	黄子超, 杨婷惠, 张颖冰, 管超群, 桂文科, 况敏, 杨建平. 提高酸性CO₂电解的选择性:阳离子效应和催化剂创新[J]. 催化学报, 2024, 63(8): 61-80.
[6]	穆楠, 薄婷婷, 胡玉高, 徐瑞欣, 刘彦昱, 周伟. 基于铁电In₂Se₃极化切换的N₂还原单原子催化剂[J]. 催化学报, 2024, 63(8): 244-257.
[7]	刘洁宇, 郭海强, 熊钰林, 陈星, 于一夫, 王长洪. 基于DFT和机器学习的Pt单原子合金电催化剂的理性设计: 应用于NO至NH₃的转化[J]. 催化学报, 2024, 62(7): 243-253.
[8]	蒋兴星, 赵雨昕, 孔艳, 孙建桔, 冯上照, 胡琪, 杨恒攀, 何传新. 充分稳定和暴露的铜基异质结实现CO₂高效电还原制乙醇[J]. 催化学报, 2024, 58(3): 216-225.
[9]	魏世蕾, 华航, 任清璇, 罗景山. 硅酸盐缓冲层促进酸性膜电极体系中电催化CO₂还原[J]. 催化学报, 2024, 66(11): 139-145.
[10]	施兆平, 王子昂, 吴鸿翔, 肖梅玲, 刘长鹏, 邢巍. 快速配体转换制备高密度Ir单原子用于高效催化酸性水氧化[J]. 催化学报, 2024, 66(11): 223-232.
[11]	Hyun Sub Kim, Hwangho Lee, Hongbeom Park, Inhak Song, Do Heui Kim. 低温氨-选择催化还原用氧化锰-沸石杂化催化剂上Brønsted/Lewis酸位的作用[J]. 催化学报, 2024, 65(10): 79-88.
[12]	邹心仪, 顾均. 酸性条件下二氧化碳高效电还原策略[J]. 催化学报, 2023, 52(9): 14-31.
[13]	韩璟怡, 管景奇. 通过表面镧改性和体相锰掺杂协助钴尖晶石酸性下析氧[J]. 催化学报, 2023, 51(8): 1-4.
[14]	陈宇卓, 王浩, 陆冰, 易倪, 曹亮, 王勇, 毛善俊. 用于高精度催化性能预测的精细结构敏感型深度学习框架[J]. 催化学报, 2023, 50(7): 284-296.
[15]	孟帅奇, 李忠玉, 张鹏, Francisca Contreras, 季宇, Ulrich Schwaneberg. 深度学习指导的热裂褐藻角质酶工程用于促进PET解聚[J]. 催化学报, 2023, 50(7): 229-238.

基于机器学习的加氢裂化分子筛催化材料构效关系研究

Insights into Structure-Activity Relationships between Y Zeolites and their n-C10 Hydrocracking Performances via Machine Learning Approaches

RichHTML

PDF

Supporting Information

可视化

摘要/Abstract

引用本文

使用本文

扫码分享

图/表 7

参考文献

相关文章 15

编辑推荐

Metrics

Insights into Structure-Activity Relationships between Y Zeolites and their n-C₁₀ Hydrocracking Performances via Machine Learning Approaches