Local modeling and optimization of K-means-PSO-SVR for methanol to aromatics

doi:10.16085/j.issn.1000-6613.2021-2466

Abstract

Abstract:

Aiming at the characteristics of sample data convergence, high dimension, nonlinear, strong coupling and large local difference in methanol to aromatics (MTA) process, we proposed a local modeling method of K-means-PSO-SVR to solve the problems of low prediction accuracy and weak robustness of single global model. Firstly, K-means algorithm was used to cluster the data in the sample space and divide in to k regions. Then, the support vector regression algorithm (SVR) optimized by particle swarm optimization (PSO) was used to establish mutually independent local models in the divided sample space. Finally, the k mutually independent local models were combined to form an integrated model covering the whole sample space. Under different noise levels, the performance of K-means-PSO-SVR method was compared with that of single global SVR, BP neural network and linear regression algorithm. The results showed that the K-means-PSO-SVR local modeling method was significantly better than the other three modeling methods at all levels of noise with strong robustness to noise. On the basis of the established data model, the key process parameters of the two-stage fixed-bed MTA were optimized which were verified by five independent repeated experiments. When the first and second stage reactor temperatures were 446.2℃ and 467.3℃ respectively, the volume space velocity of methanol was 0.4h^-1 and the pressure was 0.64MPa, and the highest yield total yield of benzene, toluene and xylene (BTX) was 44.30%.

Key words: methanol to aromatics, local modeling, algorithm, model, optimization

摘要：

针对甲醇制芳烃（MTA）过程数据样本趋同、维度高、非线性、强耦合、局部差异大的特性，提出了一种K-means-PSO-SVR的局部建模方法，用以解决单一全局模型预测精度低，鲁棒性不强的问题。该方法首先用K-means算法对样本空间的数据进行聚类，实现对样本空间k个区域的划分，再用经过粒子群优化算法（PSO）优化过超参数的支持向量回归算法（SVR）在划分好的样本空间上建立相互独立的局部模型，最终将建立的k个相互独立的局部模型组合起来组成覆盖整个样本空间的集成模型。在不同噪声水平下将K-means-PSO-SVR方法的建模效果与单一全局SVR、BP神经网络和线性回归3种算法的建模效果进行了比较分析，结果表明：K-means-PSO-SVR局部建模方法的性能在所有水平的噪声下都明显优于其他3种建模方法，并且该方法对噪声具有很强的鲁棒性。以建立的数据模型为基础优化了两段式固定床甲醇制芳烃的关键工艺参数，并用5次独立重复实验验证了优化结果的可靠性，得出当一段温度为446.2℃、二段温度为467.3℃、甲醇体积空速为0.4h^-1、压力为0.64MPa时反应产物中苯、甲苯和二甲苯（BTX）的总收率最高，最高收率为44.30%。

关键词: 甲醇制芳烃, 局部建模, 算法, 模型, 优化

CLC Number:

TQ018

LIU Penglong, XU Xiongfei, ZHANG Wei, XU Xin, ZHANG Kan, WANG Junwen. Local modeling and optimization of K-means-PSO-SVR for methanol to aromatics[J]. Chemical Industry and Engineering Progress, 2022, 41(9): 4691-4700.

刘鹏龙, 许雄飞, 张玮, 许鑫, 张侃, 王俊文. 甲醇制芳烃K-means-PSO-SVR局部建模及优化[J]. 化工进展, 2022, 41(9): 4691-4700.

Figures/Tables 10

References 26

1	华经情报网. 中国对二甲苯行业产量、需求量及进口分析, 国内市场缺口依旧较大[EB/OL]. [2020-12-14]. .
	huaon.com. The analysis of output, demand and import of China’s paraxylene industry, the gap in the domestic market is still large[EB/OL]. [2020-12-14]. .
2	黄格省, 包力庆, 丁文娟, 等. 我国煤制芳烃技术发展现状及产业前景分析[J]. 煤炭加工与综合利用, 2018(2): 6-10.
	HUANG Gesheng, BAO Liqing, DING Wenjuan, et al. An analysis on status quo and prospect of Chinese coal to aromatics technologies[J]. Coal Processing & Comprehensive Utilization, 2018(2): 6-10.
3	代成义, 陈中顺, 杜康, 等. 甲醇制芳烃催化剂及相关工艺研究进展[J]. 化工进展， 2020, 39(12): 5029-5041.
	DAI Chengyi, CHEN Zhongshun, DU Kang, et al. Research progress of catalysts and related technologies for methanol to aromatics[J]. Chemical Industry and Engineering Progress, 2020, 39(12): 5029-5041.
4	朱伟平, 李飞, 薛云鹏, 等. 甲醇制芳烃技术研究进展[J]. 现代化工, 2014, 34(7): 36-40, 42.
	ZHU Weiping, LI Fei, XUE Yunpeng, et al. Advances in methanol to aromatics technology[J]. Modern Chemical Industry, 2014, 34(7): 36-40, 42.
5	张梦轩, 刘洪辰，王敏，等. 化工过程的智能混合建模方法及应用[J]. 化工进展, 2021, 40(4): 1765-1776.
	ZHANG Mengxuan, LIU Hongchen, WANG Min, et al. Intelligence hybrid modelling method and applications in chemical process[J]. Chemical Industry and Engineering Progress, 2021, 40(4): 1765-1776.
6	徐亚荣, 蒋斌波, 冯丽梅, 等. 甲醇制芳烃(MTA)反应动力学的研究[J]. 聚酯工业, 2019, 32(6): 7-12.
	XU Yarong, JIANG Binbo, FENG Limei, et al. Study on reaction dynamic of methanol to aromatic[J]. Polyester Industry, 2019, 32(6): 7-12.
7	施丽丽, 方栩, 刘殿华, 等. Zn改性ZSM-5催化甲醇制芳烃反应动力学[J]. 天然气化工(C1化学与化工), 2017, 42(2): 40-44, 49.
	SHI Lili, FANG Xu, LIU Dianhua, et al. Kinetic model for reaction of methanol to aromatics on Zn modified ZSM-5 catalyst[J]. Natural Gas Chemical Industry, 2017, 42(2): 40-44, 49.
8	KADLEC P, GABRYS B, STRANDT S. Data-driven soft sensors in the process industry[J]. Computers & Chemical Engineering, 2009, 33(4): 795-814.
9	许雄飞, 刘鹏龙, 张玮, 等. 两段法固定床甲醇制芳烃产物预测多元非线性回归模型[J]. 化工学报, 2022, 73(2): 838-846.
	XU Xiongfei, LIU Penglong, ZHANG Wei, et al. Multivariate nonlinear regression model of methanol to aromatics by two-state fixed bed for product prediction[J]. CIESC Journal, 2022, 73(2): 838-846.
10	熊伟丽, 姚乐, 徐保国. 混沌最小二乘支持向量机及其在发酵过程建模中的应用[J]. 化工学报, 2013, 64(12): 4585-4591.
	XIONG Weili, YAO Le, XU Baoguo. Chaos least squares support vector machine and its application on fermentation process modeling[J]. CIESC Journal, 2013, 64(12): 4585-4591.
11	SERRANO D, CASTELLÓ D. Tar prediction in bubbling fluidized bed gasification through artificial neural networks[J]. Chemical Engineering Journal, 2020, 402: 126229.
12	李柠, 李少远, 席裕庚. 基于满意聚类的多模型建模方法[J]. 控制理论与应用, 2003, 20(5): 783-787.
	LI Ning, LI Shaoyuan, XI Yugeng. Multi-model modeling method based on satisfactory clustering[J]. Control Theory & Applications, 2003, 20(5): 783-787.
13	王海宁, 夏陆岳, 周猛飞, 等. 过程工业软测量中的多模型融合建模方法[J]. 化工进展, 2014, 33(12): 3157-3163.
	WANG Haining, XIA Luyue, ZHOU Mengfei, et al. Multi-model fusion modeling method for process industries soft sensor[J]. Chemical Industry and Engineering Progress, 2014, 33(12): 3157-3163.
26	DU Shuxin, WU Tiejun. Support vector machines for regression[J]. Acta Simulata Systematica Sinica, 2003, 15(11): 1580-1585, 1633.
27	邵信光, 杨慧中, 石晨曦. ε不敏感支持向量回归在化工数据建模中的应用[J]. 东南大学学报(自然科学版), 2004, 34(S1): 215-218.
	SHAO Xinguang, YANG Huizhong, SHI Chenxi. Chemical process data modeling based on ε-insensitive support vector regression[J]. Journal of Southeast University (Natural Science Edition), 2004, 34(S1): 215-218.
28	VAPNIK V N, GOLOWICH S, SMOLA A. Support vector machine for function approximation, regression estimation, and signal processing[C]// MOZER M, PETSCHE T. Advances in Neural Information Processing Systems 9, NTPS 1996. Cambridge, USA: MIT Press, 1997: 281-287.
29	DRUCKER H, BURGES C J C, KAUFMAN L, et al. Support vector regression machines [C]// MOZER M, PETSCHE T. Advances in Neural Information Processing Systems 9, NTPS 1996. Cambridge, USA: MIT Press, 1997: 155-161.
30	张学工. 关于统计学习理论与支持向量机[J]. 自动化学报, 2000, 26(1): 32-42.
	ZHANG Xuegong. Introduction to statistical learning theory and support vector machines[J]. Acta Automatica Sinica, 2000, 26(1): 32-42.
31	杨维, 李歧强. 粒子群优化算法综述[J]. 中国工程科学, 2004, 6(5): 87-94.
	YANG Wei, LI Qiqiang. Survey on particle swarm optimization algorithm[J]. Engineering Science, 2004, 6(5): 87-94.
32	李爱国, 覃征, 鲍复民, 等. 粒子群优化算法[J]. 计算机工程与应用, 2002, 38(21): 1-3, 17.
	LI Aiguo, QIN Zheng, BAO Fumin, et al. Particle swarm optimization algorithms[J]. Computer Engineering and Applications, 2002, 38(21): 1-3, 17.
14	仲蔚, 俞金寿. 基于模糊c均值聚类的多模型软测量建模[J]. 华东理工大学学报, 2000, 26(1): 83-87.
	ZHONG Wei, YU Jinshou. Study on soft sensing modeling via FCM based multiple models[J]. Journal of East China University of Science and Technology, 2000, 26(1): 83-87.
15	李修亮, 苏宏业, 褚健. 基于在线聚类的多模型软测量建模方法[J]. 化工学报, 2007, 58(11): 2834-2839.
	LI Xiuliang, SU Hongye, CHU Jian. Multiple models soft-sensing technique based on online clustering arithmetic[J]. CIESC Journal, 2007, 58(11): 2834-2839.
16	李丽娟, 宋坤, 赵英凯. 基于仿射传播聚类的ARA发酵过程建模[J]. 化工学报, 2011, 62(8): 2116-2121.
	LI Lijuan, SONG Kun, ZHAO Yingkai. Modeling of ARA fermentation based on affinity propagation clustering[J]. CIESC Journal, 2011, 62(8): 2116-2121.
17	李文怀, 李凯旋, 张庆庚, 等. 一种固定床绝热反应器两步法甲醇转化制取烃类混合物的方法: CN106220462B[P]. 2019-01-25.
	LI Wenhuai, LI Kaixuan, ZHANG Qinggeng, et al. The invention relates to a two-step methanol conversion method for producing hydrocarbon mixtures in a fixed bed adiabatic reactor: CN106220462B[P]. 2019-01-25.
18	张宝珠. 甲醇转化制芳烃(MTA)反应的研究[D]. 大连: 大连理工大学, 2013.
	ZHANG Baozhu. Study on methanol to aromatics (MTA) reaction[D]. Dalian: Dalian University of Technology, 2013.
19	解峰, 黎汉生, 赵学良, 等. 甲醇在活性Al₂O₃催化剂表面的吸附与脱水反应[J]. 催化学报, 2004, 25(5): 403-408.
	XIE Feng, LI Hansheng, ZHAO Xueliang, et al. Adsorption and dehydration of methanol on Al₂O₃ catalyst[J]. Chinese Journal of Catalysis, 2004, 25(5): 403-408.
20	SEDIGHI M, BAHRAMI H, TOWFIGHI J. Kinetic modeling formulation of the methanol to olefin process: parameter estimation[J]. Journal of Industrial and Engineering Chemistry, 2014, 20(5): 3108-3114.
21	刘艳, 常琴琴, 杨萌, 等. 甲醇制芳烃工艺研究进展[J]. 化学工程, 2015, 43(9): 74-78.
	LIU Yan, CHANG Qinqin, YANG Meng, et al. Research progress of methanol to aromatics[J]. Chemical Engineering (China), 2015, 43(9): 74-78.
22	OLSBYE U, BJØRGEN M, SVELLE S, et al. Mechanistic insight into the methanol-to-hydrocarbons reaction[J]. Catalysis Today, 2005, 106(1/2/3/4): 108-111.
23	ILIAS S, BHAN A. Tuning the selectivity of methanol-to-hydrocarbons conversion on H-ZSM-5 by co-processing olefin or aromatic compounds[J]. Journal of Catalysis, 2012, 290: 186-192.
24	李航. 统计学习方法[M]. 2版. 北京: 清华大学出版社, 2019: 263-268.
	LI Hang. Statistical learning method[M]. 2nd ed. Beijing: Tsinghua University Press, 2019: 263-268.
25	朱礼涛，欧阳博，张希宝，等. 机器学习在多相反应器中的应用进展[J]. 化工进展, 2021， 40（4）： 1699-1714.
	ZHU Litao, OUYANG Bo, ZHANG Xibao, et al. Progress on application of machine learning to multiphase reactors[J]. Chemical Industry and Engineering Progress, 2021, 40(4): 1699-1714.
26	杜树新, 吴铁军. 用于回归估计的支持向量机方法[J]. 系统仿真学报, 2003, 15(11): 1580-1585, 1633.

工艺条件	最低值	最高值
一段反应温度/℃	390	510
二段反应温度/℃	450	510
甲醇体积空速/h^-1	0.1	0.5
反应压力/MPa	0	0.8

工艺条件	最低值	最高值
一段反应温度/℃	390	510
二段反应温度/℃	450	510
甲醇体积空速/h^-1	0.1	0.5
反应压力/MPa	0	0.8

水平	一段反应温度/℃	二段反应温度/℃	甲醇体积空速/h^-1	反应压力/MPa
1	390	450	0.1	0
2	420	465	0.2	0.2
3	450	480	0.3	0.4
4	480	495	0.4	0.6
5	510	510	0.5	0.8

水平	一段反应温度/℃	二段反应温度/℃	甲醇体积空速/h^-1	反应压力/MPa
1	390	450	0.1	0
2	420	465	0.2	0.2
3	450	480	0.3	0.4
4	480	495	0.4	0.6
5	510	510	0.5	0.8

编号	一段反应温度/℃	二段反应温度/℃	甲醇体积空速/h^-1	反应压力/MPa	BTX 总收率/%
1	389.2	450.0	0.5	0	27.90
2	389.8	509.0	0.5	0	27.06
3	390.2	450.0	0.1	0	30.25
4	389.0	510.0	0.1	0	35.26
5	450.4	480.0	0.3	0	28.10
6	509.3	450.0	0.5	0	23.77
7	510.3	510.0	0.5	0	26.19
8	510.0	450.0	0.1	0	33.92
…	…	…	…	…	…
64	472.3	494.2	0.2	0.23	35.86
65	398.8	477.3	0.3	0.45	27.19
66	461.9	476.0	0.5	0.44	34.57
67	460.3	450.0	0.3	0.40	37.71
68	464.8	508.4	0.3	0.42	40.26
69	442.2	460.0	0.4	0.66	42.62