基于自适应采样算法的芳烃异构化代理模型

doi:10.11949/0438-1157.20190359

摘要/Abstract

摘要：

异构化是芳烃生产中的重要环节，提高异构化环节的建模和优化效率对工业生产有着重要意义。但是，直接使用机理模型的优化过程耗时较长，优化效率低。代理模型可以有效地对机理模型进行近似，而代理模型采样方法对模型精度有很大影响。提出了一种新的基于稀疏度和最邻近期望的自适应采样算法，该方法可以平衡全局搜索和局部搜索，通过求解优化问题找到反映函数关键信息的新采样点，再加入原始样本集中，使得代理模型精度不断提高。多个测试函数结果表明，相比于其他自适应采样算法，该算法能有效提升代理模型精度和建模效率。该算法在芳烃异构化环节代理模型中也得到了有效验证，与本文中其他算法对比，该算法模型误差减少5%以上，建模时间缩短30%以上。

关键词: 模型, 算法, 反应, 稀疏度, 芳烃异构化, 自适应采样

Abstract:

Isomerization is an important part in the production of aromatics. Improving the modeling and optimization efficiency of isomerization is of great significance to industrial production. However, the optimization process using the mechanism model directly takes a long time and the optimization efficiency is low. The surrogate model can effectively approximate the mechanism model and greatly reduce the computation time. The surrogate model sampling method has a great influence on the accuracy of the model. A new adaptive sampling algorithm based on sparsity and nearest neighbor expectation is proposed in this paper, which can balance global search and local search. By solving the optimization problem, we can find new sampling points that reflect the key information of the object function, and then add the new point to the sample set, so that the accuracy of the surrogate model is continuously improved. The results of multiple test functions show that compared with other adaptive sampling algorithms, the algorithm can effectively improve the accuracy and modeling efficiency of the surrogate model. The algorithm is also validated in the aromatic isomerization surrogate model. Compared with other algorithms, our algorithm model error is reduced by more than 5%, and the modeling time is reduced by more than 30%.

Key words: model, algorithm, reaction, sparsity, aromatic isomerization, adaptive sampling

中图分类号:

TQ 241.1

谢雨珩, 李智, 杨明磊, 杜文莉. 基于自适应采样算法的芳烃异构化代理模型[J]. 化工学报, 2020, 71(2): 688-697.

Yuheng XIE, Zhi LI, Minglei YANG, Wenli DU. Surrogate model of aromatic isomerization process based on adaptive sampling algorithm[J]. CIESC Journal, 2020, 71(2): 688-697.

图/表 16

图1 稀疏度计算流程图

Fig.1 Flow chart of sparsity calculation

图2 二维稀疏度示意图

Fig.2 Diagram of 2D sparsity

图3 算法流程图

Fig.3 Flowchart of algorithm

图4 PK函数采样点分布情况

Fig.4 Sampling point distribution of PK function

图5 PK函数RMSE变化情况

Fig.5 Variation curve of RMSE of PK function

表1 PK函数自适应采样与拉丁超立方采样结果对比

Table 1 Comparison of PK function adaptive sampling and Latin hypercube sampling results

采样方法	采样点/个	RMSE	时间/s
自适应采样	20+80	0.078739	8.79
拉丁超立方采样	100	0.628361	0.196203

表2 采样方式

Table 2 Sampling configurations

测试函数	维数	初始采样点	新增采样点
PK	2	20	80
AlpineN.1	2	20	80
AlpineN.2	2	20	80
Hartman3	3	30	120
Shekel5	4	40	160
Shekel7	4	40	160
Hartman6	6	60	240
Ackley	8	80	320

表3 各测试函数的平均RMSE

Table 3 Average RMSE obtained by four sampling approaches for test cases

测试函数	MASA	ASA-NNDM	SSA	ASSA-SNN
PK	0.07506	0.11860	0.09055	0.06428
AlpineN.1	1.96015	1.75482	1.21713	1.21350
AlpineN.2	0.02524	0.09999	0.01950	0.04680
Hartman3	0.01351	0.05956	0.01607	0.01510
Shekel5	0.18186	0.13901	0.08644	0.07058
Shekel7	0.25594	0.20254	0.09568	0.07924
Hartman6	0.12602	0.21485	0.11549	0.10785
Ackley	0.39066	0.41528	0.40963	0.38582

表4 各测试函数的平均计算时间/s

Table 4 Average running time obtained by four sampling approaches for test cases/s

测试函数	MASA	ASA-NNDM	SSA	ASSA-SNN
PK	15.87	125.68	32.40	8.87
AlpineN.1	16.41	124.21	32.71	8.54
AlpineN.2	15.90	119.78	28.62	8.57
Hartman3	55.14	219.46	162.34	23.12
Shekel5	148.17	369.43	669.46	52.17
Shekel7	150.60	372.50	652.90	53.17
Hartman6	650.93	1127.57	6142.58	189.25
Ackley	2078.48	3028.23	28864.74	490.20

图6 异构化过程流程图1—反应器；2—加热炉：3—热交换器；4—压缩机；5—气液分离罐；6—精馏塔；7—回流槽；8—循环搭；9—回流槽

Fig.6 Flow chart of isomerization process

表5 芳烃异构化重要指标的RMSE

Table 5 RMSE of important indicators for isomerization of aromatics

质量指标	MASA	ASA-NNDM	SSA	ASSA-SNN
轻烃收率	0.001261	0.001115	0.001147	0.001053
C₈₊收率	0.001216	0.001123	0.001061	0.000986

表6 芳烃异构化重要指标的建模时间/s

Table 6 Modeling time for important indicators of aromatic isomerization/s

质量指标	MASA	ASA-NNDM	SSA	ASSA-SNN
轻烃收率	3313.04	4600.36	20404.30	2189.42
C₈₊收率	3313.02	4574.49	20429.36	2122.41

图7 异构化代理模型

Fig.7 Surrogate model of isomerization

图8 MX含量、OX含量样本分布

Fig.8 Sample distribution of MX content and OX content

图9 异构化的轻烃收率RMSE变化曲线

Fig.9 Variation curves of RMSE of isomerized light hydrocarbon

图10 异构化C8+收率RMSE变化曲线

Fig.10 Variation curves of RMSE of C8+ yield with number of iterations

参考文献 33

1	徐欧官, 傅永峰, 陈祥华. 工业异构化反应器建模及仿真[J]. 化工学报, 2011, 62(8): 2298-2302.
	Xu O G, Fu Y F, Chen X H. Modeling and simulation for an industrial reactor on hydroisomerization of C8-aromatics[J]. CIESC Journal, 2011, 62(8): 2298-2302.
2	Lim D, Jin Y, Ong Y S, et al. Generalizing surrogate-assisted evolutionary computation[J]. IEEE Transactions on Evolutionary Computation, 2010, 14(3): 329-355.
3	Jin Y. A comprehensive survey of fitness approximation in evolutionary computation[J]. Soft Computing, 2005, 9(1): 3-12.
4	Box G E P, Wilson K B. On the experimental attainment of optimum conditions[J]. Journal of the Royal Statistical Society, 1951, 13(1): 1-45.
5	Lian Y S, Liou M S. Multiobjective optimization using coupled response surface model and evolutionary algorithm[J]. AIAA Journal, 2005, 43(6): 1316-1325.
6	Mullur A, Messac A. Extended radial basis functions: more flexible and effective metamodeling[J]. AIAA Journal, 2005, 43(6): 1306-1315.
7	Fang H B, Horstemeyer M F. Global response approximation with radial basis functions[J]. Engineering Optimization, 2006, 38(4): 407-424.
8	Krige D G. A statistical approach to some basic mine valuation problems on the witwatersrand[J]. Journal of the Chemical, Metallurgical and Mining Engineering Society of South Africa, 1951, 52(6): 119-139.
9	Cressie N. Spatial prediction and ordinary kriging[J]. Mathematical Geology, 1989, 21(4): 493-494.
10	Vapnik V, Golowich S E, Smola A. Support vector method for function approximation, estimation regression, and signal processing[C]// Advances in Neural Information Processing Systems 9, 1997, 9: 281-287.
11	Friedman J H, Roosen C B. An introduction to multivariate adaptive regression splines[J]. Statistical Methods in Medical Research, 1995, 4(3): 197-217.
12	Friedman J H. Multivariate adaptive regression spline[J]. The Annals of Statistics, 1991, 19(1): 1-67.
13	Basheer I, Hajmeer M. Artificial neural networks: fundamentals, computing, design, and application[J]. Journal of Microbiological Methods, 2000, 43(1): 3-31.
14	Jin Y, Li J, Du W, et al. Adaptive sampling for surrogate modelling with artificial neural network and its application in an industrial cracking furnace[J]. The Canadian Journal of Chemical Engineering, 2016, 94(2): 262-272.
15	Zhou B D, Yao H L, Shi M H, et al. An new immune genetic algorithm based on uniform design sampling[J]. Knowledge and Information Systems, 2012, 31(2): 389-403.
16	Ye K Q. Orthogonal column Latin hypercubes and their application in computer experiments[J]. Publications of the American Statistical Association, 1998, 93(444): 1430-1439.
17	Wang G. Adaptive response surface method using inherited Latin hypercube design points[J]. Journal of Mechanical Design, 2003, 125(2): 210-220.
18	Jin Y. A framework for evolutionary optimization with approximate fitness functions[J]. IEEE Transactions on Evolutionary Computation, 2002, 6(5): 481-494.
19	Wang H, Jin Y, Doherty J. Committee-based active learning for surrogate-assisted particle swarm optimization of expensive problems[J]. IEEE Trans. Cybern., 2017, 47(9): 2664-2677.
20	Eason J, Cremaschi S. Adaptive sequential sampling for surrogate model generation with artificial neural networks[J]. Computers & Chemical Engineering, 2014, 68: 220-232.
21	Wei X, Wu Y Z, Chen L P. A new sequential optimal sampling method for radial basis functions[J]. Applied Mathematics and Computation, 2012, 218(19): 9635-9646.
22	Garud S S, Karimi I A, Kraft M. Smart sampling algorithm for surrogate model development[J]. Computers & Chemical Engineering, 2017, 96: 103-114.
23	Müller J, Shoemaker C A. Influence of ensemble surrogate models and sampling strategy on the solution quality of algorithms for computationally expensive black-box global optimization problems[J]. Journal of Global Optimization, 2014, 60(2): 123-144.
24	Sacks J, Welch W J, Toby J, et al. Design and analysis of computer experiments[J]. Statistical Science, 1989, 4(4): 409-423.
25	王晓强, 罗娜, 叶贞成, 等. 基于Kriging的差分进化算法及其在苯乙烯流程优化中的应用[J]. 化工学报, 2013, 64(12): 4563-4570.
	Wang X Q, Luo N, Ye Z C, et al. Differential evolution algorithm based on Kriging and its application in styrene plant optimization[J]. CIESC Journal, 2013, 64(12): 4563-4570.
26	段星辰, 杜文莉. 基于bootstrap GEI算法的碳二加氢反应器代理模型[J]. 化工学报, 2015, 66(12): 4904-4909.
	Duan X C, Du W L. Surrogate model of acetylene hydrogenation reactor based on bootstrap GEI algorithm[J]. CIESC Journal, 2015, 66(12): 4904-4909.
27	Xiao N C, Zuo M J, Guo W. Efficient reliability analysis based on adaptive sequential sampling design and cross-validation[J]. Applied Mathematical Modelling, 2018, 58: 404-420.
28	Gong W, Duan Q. An Adaptive Surrogate modeling-based sampling strategy for parameter optimization and distribution estimation (ASMO-PODE)[J]. Environmental Modelling & Software, 2017, 95: 61-75.
29	韩忠华. Kriging模型及代理优化算法研究进展[J]. 航空学报, 2016, 37(11): 3197-3225.
	Han Z H. Kriging surrogate model and its application to design optimization: a review of recent progress [J]. Acta Aeronautica Et Astronautica Sinica, 2016, 37(11): 3197-3225.
30	Jin Y. Surrogate-assisted evolutionary computation: recent advances and future challenges[J]. Swarm & Evolutionary Computation, 2011, 1(2): 61-70.
31	Hooke R, Jeeves T A. “Direct search” solution of numerical and statistical problems[J]. Journal of the ACM, 1961, 8(2): 212-229.
32	张剑超, 杜文莉, 覃水. 基于新型自适应采样算法的催化重整过程代理模型[J]. 华东理工大学学报(自然科学版)， 2018， doi: 10.14135/j.cnki.1006-3080.20180915001. DOI URL
	Zhang J C, Du W L, Qin S. Surrogate model of catalytic reforming process based on a new adaptive sampling algorithm[J]. Journal of East China University of Science and Technology (Natural Science Edition)， 2018， doi: 10.14135/j.cnki.1006-3080.20180915001. DOI URL
33	刘祥荣. 二甲苯异构化反应器的模拟研究[D]. 北京: 北京化工大学, 2010.
	Liu X R. The modeling and dynamic simulation of the xylene isomerization reactor[D]. Beijing: Beijing University of Chemical Technology, 2010.

[1]	宋嘉豪, 王文. 斯特林发动机与高温热管耦合运行特性研究[J]. 化工学报, 2023, 74(S1): 287-294.
[2]	连梦雅, 谈莹莹, 王林, 陈枫, 曹艺飞. 地下水预热新风一体化热泵空调系统制热性能研究[J]. 化工学报, 2023, 74(S1): 311-319.
[3]	金正浩, 封立杰, 李舒宏. 氨水溶液交叉型再吸收式热泵的能量及分析[J]. 化工学报, 2023, 74(S1): 53-63.
[4]	程成, 段钟弟, 孙浩然, 胡海涛, 薛鸿祥. 表面微结构对析晶沉积特性影响的格子Boltzmann模拟[J]. 化工学报, 2023, 74(S1): 74-86.
[5]	王浩, 王振雷. 基于自适应谱方法的裂解炉烧焦模型化简策略[J]. 化工学报, 2023, 74(9): 3855-3864.
[6]	曹跃, 余冲, 李智, 杨明磊. 工业数据驱动的加氢裂化装置多工况切换过渡状态检测[J]. 化工学报, 2023, 74(9): 3841-3854.
[7]	李科, 文键, 忻碧平. 耦合蒸气冷却屏的真空多层绝热结构对液氢储罐自增压过程的影响机制研究[J]. 化工学报, 2023, 74(9): 3786-3796.
[8]	陈杰, 林永胜, 肖恺, 杨臣, 邱挺. 胆碱基碱性离子液体催化合成仲丁醇性能研究[J]. 化工学报, 2023, 74(9): 3716-3730.
[9]	尹刚, 李伊惠, 何飞, 曹文琦, 王民, 颜非亚, 向禹, 卢剑, 罗斌, 卢润廷. 基于KPCA和SVM的铝电解槽漏槽事故预警方法[J]. 化工学报, 2023, 74(8): 3419-3428.
[10]	李锦潼, 邱顺, 孙文寿. 煤浆法烟气脱硫中草酸和紫外线强化煤砷浸出过程[J]. 化工学报, 2023, 74(8): 3522-3532.
[11]	郑玉圆, 葛志伟, 韩翔宇, 王亮, 陈海生. 中高温钙基材料热化学储热的研究进展与展望[J]. 化工学报, 2023, 74(8): 3171-3192.
[12]	于旭东, 李琪, 陈念粗, 杜理, 任思颖, 曾英. 三元体系KCl + CaCl₂ + H₂O 298.2、323.2及348.2 K相平衡研究及计算[J]. 化工学报, 2023, 74(8): 3256-3265.
[13]	诸程瑛, 王振雷. 基于改进深度强化学习的乙烯裂解炉操作优化[J]. 化工学报, 2023, 74(8): 3429-3437.
[14]	闫琳琦, 王振雷. 基于STA-BiLSTM-LightGBM组合模型的多步预测软测量建模[J]. 化工学报, 2023, 74(8): 3407-3418.
[15]	汪林正, 陆俞冰, 张睿智, 罗永浩. 基于分子动力学模拟的VOCs热氧化特性分析[J]. 化工学报, 2023, 74(8): 3242-3255.