基于分布式贝叶斯隐马尔可夫回归的动态软测量建模方法

doi:10.11949/0438-1157.20230360

摘要/Abstract

摘要：

利用软测量技术实时预测化工过程中的关键参数对生产过程的在线监测、自动控制、实时优化具有十分重要的意义。为此，提出了一种基于隐马尔可夫模型的动态软测量建模方法。首先，针对数据规模大导致模型计算效率低和数据缺失导致数据无法充分利用的问题，提出了一种基于分布式贝叶斯隐马尔可夫回归的预测模型；其次，针对该模型进一步提出了一种能够获得精确后验分布的分布式训练方法。最后，利用蜡油加氢过程对所提方法的有效性进行了验证。

关键词: 化工过程, 动态建模, 贝叶斯隐马尔可夫回归, 软测量, 分布式计算

Abstract:

Real-time prediction of key parameters in the chemical process by using soft sensing technology is of great significance for on-line monitoring, automatic control, and real-time optimization of production process. Therefore, a dynamic soft sensor modeling method based on hidden Markov model is proposed. Firstly, aiming at the problem of low computational efficiency caused by large data scales and insufficient utilization of data due to missing data, a predictive model based on distributed Bayesian hidden Markov regression is proposed. Then, a distributed training method that can obtain accurate posterior distribution is proposed for model training. Finally, the effectiveness of the proposed model is verified by the wax oil hydrogenation process.

Key words: chemical processes, dynamic modeling, Bayesian hidden Markov regression, soft sensors, distributed computing

中图分类号:

TK 314

邵伟明, 韩文学, 宋伟, 杨勇, 陈灿, 赵东亚. 基于分布式贝叶斯隐马尔可夫回归的动态软测量建模方法[J]. 化工学报, 2023, 74(6): 2495-2502.

Weiming SHAO, Wenxue HAN, Wei SONG, Yong YANG, Can CHEN, Dongya ZHAO. Dynamic soft sensor modeling method based on distributed Bayesian hidden Markov regression[J]. CIESC Journal, 2023, 74(6): 2495-2502.

图/表 8

图1 分布式贝叶斯隐马尔可夫回归概率图表示

Fig.1 Probabilistic graphical representation of the Dis-VBHMR

图2 蜡油加氢工艺流程简图

Fig.2 Diagram of the WOHP

表1 WOHP中硫含量软测量模型的输入变量

Table 1 Input variables of soft sensor model for sulfur content in WOHP

输入变量	含义
V1	罐区蜡油流量
V2	减压蜡油流量
V3	焦化蜡油流量
V4	加热炉燃料气流量
V5	反应蜡油流量
V6	混合氢流量
V7	反应器入口温度
V8	反应器床层温度1
V9	反应器床层温度2
V10	反应器床层温度3
V11	急冷氢流量
V12	急冷氢温度
V13	反应器出口温度

图3 WOHP中部分输入变量和输出变量的概率密度图(d) (e) (f)

Fig.3 Probability density plots of partial input and output variables in WOHP

图4 蜡油加氢工艺数据缺失说明与数据集划分

Fig.4 Description of missing data of the WOHP and data set partition

图5 4种软测量模型对硫含量的预测结果

Fig.5 Predictions of the sulfur content by four soft sensor models

表2 4种软测量模型对硫含量的预测精度和模型训练时间

Table 2 Prediction accuracy and model training time of sulfur content by four soft sensor models

软测量模型	数据分组数M	RMSE	R²	MTT/s
P-DPMM	1	0.0442	0.6392	19.9408
ESN	1	0.0426	0.6647	8.4688
BHMR-1	1	0.0455	0.6174	185.5044
BHMR-2	1	0.0424	0.6674	196.9019
Dis-BHMR	5	0.0355	0.7670	100.9233
	10	0.0345	0.7804	54.4537
	15	0.0341	0.7850	35.9074
	20	0.0345	0.7796	28.1423

图6 隐状态数量对模型训练时间的影响

Fig.6 Influence of the number of hidden states on the training time of the model

参考文献 31

1	秦美华, 朱红求, 李勇刚, 等. 基于STA-K均值聚类的电化学废水处理过程离子浓度软测量[J]. 化工学报, 2019, 70(9): 3458-3464.
	Qin M H, Zhu H Q, Li Y G, et al. Soft-sensor method for ion concentration of electrochemical wastewater treatment based on STA-K-means clustering[J]. CIESC Journal, 2019, 70(9): 3458-3464.
2	Li Y G, Han W X, Shao W M, et al. Virtual sensing for dynamic industrial process based on localized linear dynamical system models with time-delay optimization[J]. ISA Transactions, 2023, 133: 505-517.
3	闻超垚, 周平. 污水处理过程出水水质稀疏鲁棒建模[J]. 自动化学报, 2022, 48(6): 1469-1481.
	Wen C Y, Zhou P. Sparse robust modeling of effluent quality indices in wastewater treatment process[J]. Acta Automatica Sinica, 2022, 48(6): 1469-1481.
4	Ge Z Q, Song Z H, Ding S X, et al. Data mining and analytics in the process industry: the role of machine learning[J]. IEEE Access, 2017, 5: 20590-20616.
5	周乐, 沈程凯, 吴超, 等. 深度融合特征提取网络及其在化工过程软测量中的应用[J]. 化工学报, 2022, 73(7): 3156-3165.
	Zhou L, Shen C K, Wu C, et al. Deep fusion feature extraction network and its application in chemical process soft sensing[J]. CIESC Journal, 2022, 73(7): 3156-3165.
6	孙凯, 隋璘, 张芳芳, 等. 基于非负绞杀与长短期记忆神经网络的动态软测量算法[J]. 控制理论与应用, 2023, 40(1): 83-93.
	Sun K, Sui L, Zhang F F, et al. Dynamic soft sensor algorithm based on nonnegative garrote and long short-term memory neural network[J]. Control Theory & Applications, 2023, 40(1): 83-93.
7	Khine K L, Nyunt T S. Predictive geospatial analytics using principal component regression[J]. International Journal of Electrical and Computer Engineering (IJECE), 2020, 10(3): 2651-2658.
8	Liu J X, Sun D S, Chen J H. Comparative study on wavelet functional partial least squares soft sensor for complex batch processes[J]. Chemical Engineering Science, 2022, 254: 117601.
9	Shao W M, Ge Z Q, Song Z H. Soft-sensor development for processes with multiple operating modes based on semisupervised Gaussian mixture regression[J]. IEEE Transactions on Control Systems Technology, 2019, 27(5): 2169-2181.
10	Shao W M, Ge Z Q, Song Z H, et al. Nonlinear industrial soft sensor development based on semi-supervised probabilistic mixture of extreme learning machines[J]. Control Engineering Practice, 2019, 91: 104098.
11	Wang J B, Shao W M, Song Z H. Robust inferential sensor development based on variational Bayesian student's-t mixture regression[J]. Neurocomputing, 2019, 369: 11-28.
12	Dong Y N, Qin S J. A novel dynamic PCA algorithm for dynamic data modeling and process monitoring[J]. Journal of Process Control, 2018, 67: 1-11.
13	Shang C, Huang X L, Suykens J A K, et al. Enhancing dynamic soft sensors based on DPLS: a temporal smoothness regularization approach[J]. Journal of Process Control, 2015, 28: 17-26.
14	Shao W M, Han W X, Li Y G, et al. Enhancing the reliability and accuracy of data-driven dynamic soft sensor based on selective dynamic partial least squares models[J]. Control Engineering Practice, 2022, 127: 105292.
15	Liu S T, Gao X W, He H F, et al. Soft sensor modelling of acrolein conversion based on hidden Markov model of principle component analysis and fireworks algorithm[J]. The Canadian Journal of Chemical Engineering, 2019, 97(12): 3052-3062.
16	Ge Z Q, Chen X R. Dynamic probabilistic latent variable model for process data modeling and regression application[J]. IEEE Transactions on Control Systems Technology, 2019, 27(1): 323-331.
17	Zhou L, Zheng J Q, Ge Z Q, et al. Multimode process monitoring based on switching autoregressive dynamic latent variable model[J]. IEEE Transactions on Industrial Electronics, 2018, 65(10): 8184-8194.
18	Shang C, Huang B, Yang F, et al. Probabilistic slow feature analysis-based representation learning from massive process data for soft sensor modeling[J]. AIChE Journal, 2015, 61(12): 4126-4139.
19	Ma Y J, Huang B. Bayesian learning for dynamic feature extraction with application in soft sensing[J]. IEEE Transactions on Industrial Electronics, 2017, 64(9): 7171-7180.
20	邵伟明, 葛志强, 李浩, 等. 基于循环神经网络的半监督动态软测量建模方法[J]. 电子测量与仪器学报, 2019, 33(11): 7-13.
	Shao W M, Ge Z Q, Li H, et al. Semisupervised dynamic soft sensing approaches based on recurrent neural network[J]. Journal of Electronic Measurement and Instrumentation, 2019, 33(11): 7-13.
21	Yuan X F, Li L, Shardt Y A W, et al. Deep learning with spatiotemporal attention-based LSTM for industrial soft sensor model development[J]. IEEE Transactions on Industrial Electronics, 2021, 68(5): 4404-4414.
22	刘国海, 江兴科, 梅从立. 基于连续隐Markov模型的发酵过程软测量方法[J]. 控制与决策, 2011, 26(11): 1753-1756, 1760.
	Liu G H, Jiang X K, Mei C L. Research on soft sensing method based on continuous hidden Markov model in fermentation process[J]. Control and Decision, 2011, 26(11): 1753-1756, 1760.
23	Shao W M, Xiao C F, Wang J B, et al. Real-time estimation of quality-related variable for dynamic and non-Gaussian process based on semisupervised Bayesian HMM[J]. Journal of Process Control, 2022, 111: 59-74.
24	朱熀秋, 樊帅. 基于改进连续隐马尔可夫模型的六极径向主动磁轴承转子位移软测量[J]. 中国电机工程学报, 2021, 41(11): 3933-3943.
	Zhu H Q, Fan S. Soft-sensing modeling for rotor displacements of six-pole radial active magnetic bearing using improved continuous hidden Markov model[J]. Proceedings of the CSEE, 2021, 41(11): 3933-3943.
25	Xiao C F, Han W X, Shao W M, et al. Distributed semisupervised HMM for dynamic inferential sensor development[J]. IEEE Sensors Journal, 2023, 23(3): 2737-2749.
26	Wang J B, Shao W M, Zhang X M, et al. Dynamic variational Bayesian student's t mixture regression with hidden variables propagation for industrial inferential sensor development[J]. IEEE Transactions on Industrial Informatics, 2021, 17(8): 5314-5324.
27	Wang L, Yang C J, Sun Y X. Multimode process monitoring approach based on moving window hidden Markov model[J]. Industrial & Engineering Chemistry Research, 2018, 57(1): 292-301.
28	Bishop C M. Pattern Recognition and Machine Learning[M]. New York: Springer, 2006.
29	Shao W M, Yao L, Ge Z Q, et al. Parallel computing and SGD-based DPMM for soft sensor development with large-scale semisupervised data[J]. IEEE Transactions on Industrial Electronics, 2019, 66(8): 6362-6373.
30	Wang P, Yin Y C, Deng X G, et al. Semi-supervised echo state network with temporal-spatial graph regularization for dynamic soft sensor modeling of industrial processes[J]. ISA Transactions, 2022, 130: 306-315.
31	Yao L, Ge Z Q. Refining data-driven soft sensor modeling framework with variable time reconstruction[J]. Journal of Process Control, 2020, 87: 91-107.

[1]	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.
[2]	贠程, 王倩琳, 陈锋, 张鑫, 窦站, 颜廷俊. 基于社团结构的化工过程风险演化路径深度挖掘[J]. 化工学报, 2023, 74(4): 1639-1650.
[3]	张中秋, 李宏光, 石逸林. 基于人工预测调控策略的复杂化工过程多任务学习方法[J]. 化工学报, 2023, 74(3): 1195-1204.
[4]	周乐, 沈程凯, 吴超, 侯北平, 宋执环. 深度融合特征提取网络及其在化工过程软测量中的应用[J]. 化工学报, 2022, 73(7): 3156-3165.
[5]	王琨, 侍洪波, 谭帅, 宋冰, 陶阳. 局部时差约束邻域保持嵌入算法在故障检测中的应用[J]. 化工学报, 2022, 73(7): 3109-3119.
[6]	高欢, 丁国良, 周发贤, 庄大伟. R410A制冷剂在润滑油中的动态析出特性的研究[J]. 化工学报, 2022, 73(3): 1054-1062.
[7]	罗顺桦, 王振雷, 王昕. 基于二子空间协同训练算法的半监督软测量建模[J]. 化工学报, 2022, 73(3): 1270-1279.
[8]	石珊珊, 魏爱博, 张小斌. 液氮超声空化CFD模拟及实验研究[J]. 化工学报, 2021, 72(4): 1930-1938.
[9]	陈忠圣, 朱梅玉, 贺彦林, 徐圆, 朱群雄. 基于分位数回归CGAN的虚拟样本生成方法及其过程建模应用[J]. 化工学报, 2021, 72(3): 1529-1538.
[10]	刘聪, 谢莉, 杨慧中. 基于改进DPC的青霉素发酵过程多模型软测量建模[J]. 化工学报, 2021, 72(3): 1606-1615.
[11]	王晓慧, 王延江, 邓晓刚, 张政. 基于加权深度支持向量数据描述的工业过程故障检测[J]. 化工学报, 2021, 72(11): 5707-5716.
[12]	薛峰, 李欣铜, 周琨, 魏志强, 葛晓霞, 葛志强, 宋凯. 基于GLSAFIS的氟化工过程操作单元可靠性监测[J]. 化工学报, 2021, 72(11): 5696-5706.
[13]	冯思琦, 罗雄麟. 模式切换类经济预测控制切换时间在线估计[J]. 化工学报, 2020, 71(S2): 225-240.
[14]	孟繁鑫, 孙佳宁, 周月, 高赞军, 程定斌. 飞机环控系统空气循环机仿真建模及试验校核[J]. 化工学报, 2020, 71(S1): 328-334.
[15]	尹林子, 关羽吟, 蒋朝辉, 许雪梅. 基于k-means++的高炉铁水硅含量数据优选方法[J]. 化工学报, 2020, 71(8): 3661-3670.