基于即时学习的改进条件高斯回归软测量

doi:10.11949/0438-1157.20240127

摘要/Abstract

摘要：

基于数据驱动的在线软测量是当前工业智能化感知的重要研究方向。在算法实际部署中，过程模态切换以及数据漂移都会导致软测量性能下降，传统自适应方法又存在模型单一、模态遗忘等不足。针对上述问题提出一种基于即时学习的样本时空加权条件高斯回归（STWCGR）软测量算法。该方法用概率密度估计和条件概率计算实现软测量建模和预测：首先根据即时学习思想通过样本时空混合加权方法筛选局部建模数据，然后结合高斯混合回归思想累积局部单高斯概率密度模型对数据分布进行拟合，最后引入预测动量更新和模态更新策略提高预测稳定性并赋予模型对新工况的学习适应能力。通过仿真实验验证了所提方法在预测精度、稳定性以及新模态适应能力上的有效性。

关键词: 智能感知, 数据驱动软测量, 预测, 即时学习, 高斯混合回归

Abstract:

Data-driven online soft sensing is an important research direction in current industrial intelligent sensing. In the practical use of algorithms, process mode switching and data drift might reduce the performance of soft sensors. Traditional adaptive approaches confront limitations, such as a limited variety of models and a tendency to forget previously acquired modes. A sample temporal and spatial weighted conditional gaussian regression (STWCGR) soft sensor algorithm based on just-in-time learning is proposed to overcome these issues. This algorithm uses probability density and conditional probability for soft sensing modeling and prediction. First, a sample spatiotemporal mixed-weight technique is used to pick local modeling data in accordance with the just-in-time learning principle. Then, the local Gaussian probability density models are accumulated to fit the data distribution by incorporating the concept of Gaussian mixture regression. Finally, momentum updates and mode updates are introduced to enhance prediction stability and endow the model with adaptability to new working conditions. The efficacy of the suggested algorithm is confirmed by simulation studies with respect to forecast precision, stability, and flexibility to accommodate new modes.

Key words: AI perception, data-driven soft sensor, prediction, just-in-time learning, Gaussian mixture regression

中图分类号:

TP 274

黎宏陶, 王振雷, 王昕. 基于即时学习的改进条件高斯回归软测量[J]. 化工学报, 2024, 75(6): 2299-2312.

Hongtao LI, Zhenlei WANG, Xin WANG. Improved conditional Gaussian regression soft sensor based on just-in-time learning[J]. CIESC Journal, 2024, 75(6): 2299-2312.

图/表 20

图1 基于STWCGR的软测量流程图

Fig.1 Flowchart for soft sensing based on STWCGR

表1 高斯成分分布参数

Table 1 Parameters of Gaussian components

高斯模态	$μ$	$Σ$
1	$2, 2 T$	$1 0.5 0.5 1$
2	$- 4, 2 T$	$1 - 0.5 - 0.5 1$
3	$- 2, 6 T$	$1001$
4	$8, - 2 T$	$1001$

表1 高斯成分分布参数

Table 1 Parameters of Gaussian components

高斯模态	$μ$	$Σ$
1	$2, 2 T$	$1 0.5 0.5 1$
2	$- 4, 2 T$	$1 - 0.5 - 0.5 1$
3	$- 2, 6 T$	$1001$
4	$8, - 2 T$	$1001$

图2 数值示例一可视化

Fig.2 Visualization of numerical example Ⅰ

图3 数值示例一预测结果

Fig.3 Prediction results of numerical example Ⅰ

图4 数值示例二STWCGR预测结果

Fig.4 Prediction results of numerical example Ⅱ based on STWCGR

图5 数值示例二WGR预测结果

Fig.5 Prediction results of numerical example Ⅱ based on WGR

表2 数值示例二软测量性能

Table 2 Soft sensing performance in numerical example Ⅱ

项目	WGR $f 1$	STWCGR $f 1$	WGR $f 2$	STWCGR $f 2$
RMSE	0.042	0.190	2.702	0.804
$R 2$	0.998	0.966	-1.937	0.732

表2 数值示例二软测量性能

Table 2 Soft sensing performance in numerical example Ⅱ

项目	WGR $f 1$	STWCGR $f 1$	WGR $f 2$	STWCGR $f 2$
RMSE	0.042	0.190	2.702	0.804
$R 2$	0.998	0.966	-1.937	0.732

图6 脱丁烷塔工艺流程示意图

Fig.6 Flowchart of debutanizer column

表3 脱丁烷塔变量含义

Table 3 Description of variables for debutanizer column

变量名	含义
U1	塔顶温度
U2	塔顶压力
U3	回流流量
U4	向下一过程流量
U5	第6隔板温度
U6	塔釜温度A
U7	塔釜温度B
Y	塔釜丁烷含量

图7 脱丁烷塔过程STWCGR预测结果

Fig.7 Prediction results of debutanizer column process based on STWCGR

表4 脱丁烷塔过程软测量性能

Table 4 Soft sensing performance in debutanizer column process

项目	LSTM	GRU	JIT-PLS	WGR	STWCGR
RMSE	0.148	0.132	0.073	0.067	0.062
$R 2$	0.270	0.226	0.787	0.820	0.872

表4 脱丁烷塔过程软测量性能

Table 4 Soft sensing performance in debutanizer column process

项目	LSTM	GRU	JIT-PLS	WGR	STWCGR
RMSE	0.148	0.132	0.073	0.067	0.062
$R 2$	0.270	0.226	0.787	0.820	0.872

图8 脱丁烷塔过程深度学习预测结果

Fig.8 Prediction results of debutanizer column process based on deep learning

图9 消融实验预测结果

Fig.9 Prediction results of ablation experiment

表5 消融实验软测量性能

Table 5 Soft sensing performance of ablation experiment

项目	TWCGR^*	TWCGR	STWCGR^*	STWCGR
RMSE	0.097	0.101	0.063	0.062
$R 2$	0.688	0.666	0.871	0.872

表5 消融实验软测量性能

Table 5 Soft sensing performance of ablation experiment

项目	TWCGR^*	TWCGR	STWCGR^*	STWCGR
RMSE	0.097	0.101	0.063	0.062
$R 2$	0.688	0.666	0.871	0.872

图10 TE过程流程示意图

Fig.10 Flowchart of TE process

表6 TE过程变量含义

Table 6 Description of variables for TE process

变量名	含义	变量名	含义
X1	物料A 流量	X13	分离器
X2	物料D 流量	X14	分离器塔底流量
X3	物料E 流量	X15	汽提塔液位
X4	A、B、C总进料	X16	汽提塔压力
X5	再循环流量	X17	汽提塔塔底流量
X6	反应器进料量	X18	汽提塔温度
X7	反应器压力	X19	蒸汽流量
X8	反应器液位	X20	压缩机功率
X9	反应器温度	X21	反应器冷凝水出口温度
X10	排放速度	X22	冷凝器冷却水出口温度
X11	分离器温度	Y	产物中G组分摩尔分数
X12	分离器液位	Y	产物中G组分摩尔分数

图11 TE过程STWCGR预测结果

Fig.11 Prediction results of TE process based on STWCGR

图12 TE过程LSTM预测结果

Fig.12 Prediction results of TE process based on LSTM

表7 乙烯精馏塔过程变量

Table 7 Process variables of ethylene distillation tower

变量	含义
X1	塔顶压力
X2	塔顶温度
X3	142^#温度
X4	168^#温度
X5	塔釜采出温度
X6	再沸器出口物料温度
Y	塔顶乙烷浓度

图13 乙烷浓度预测结果

Fig.13 Prediction results of ethane concentration

参考文献 31

1	Kadlec P, Gabrys B, Strandt S. Data-driven soft sensors in the process industry[J]. Computers & Chemical Engineering, 2009, 33(4): 795-814.
2	Sun Q Q, Ge Z Q. A survey on deep learning for data-driven soft sensors[J]. IEEE Transactions on Industrial Informatics, 2021, 17(9): 5853-5866.
3	王晓阳, 王昕, 王振雷, 等. 一种乙烯裂解炉的多工况监控方法[J]. 计算机与应用化学, 2012, 29(9): 1119-1122.
	Wang X Y, Wang X, Wang Z L, et al. A novel method for monitoring ethylene cracking furnace with multiple operation modes[J]. Computers and Applied Chemistry, 2012, 29(9): 1119-1122.
4	赵俊峰. 乙烯裂解炉炉管智能结焦诊断方法研究[D]. 广州: 广东工业大学, 2020.
	Zhao J F. Study on intelligent coking diagnosis method of ethylene cracking furnace tubes[D]. Guangzhou: Guangdong University of Technology, 2020.
5	Gama J, Žliobaitė I, Bifet A, et al. A survey on concept drift adaptation[J]. ACM Computing Surveys, 2014, 46(4): 1-37.
6	Lü Y, Yang H Z. A multi-model approach for soft sensor development based on feature extraction using weighted kernel fisher criterion[J]. Chinese Journal of Chemical Engineering, 2014, 22(2): 146-152.
7	Gholami A R, Shahbazian M. Soft sensor design based on fuzzy C-Means and RFN_SVR for a stripper column[J]. Journal of Natural Gas Science and Engineering, 2015, 25: 23-29.
8	Deng X G, Chen Y X, Wang P, et al. Soft sensor modeling for unobserved multimode nonlinear processes based on modified kernel partial least squares with latent factor clustering[J]. IEEE Access, 2020, 8: 35864-35872.
9	Stržinar Ž, Pregelj B, Škrjanc I. Soft sensor for non-invasive detection of process events based on Eigenresponse Fuzzy Clustering[J]. Applied Soft Computing, 2023, 132: 109859.
10	Yuan X F, Ou C, Wang Y L, et al. A layer-wise data augmentation strategy for deep learning networks and its soft sensor application in an industrial hydrocracking process[J]. IEEE Transactions on Neural Networks and Learning Systems, 2021, 32(8): 3296-3305.
11	Dai Y, Yu Q, Yang T, et al. Enhanced soft sensor with qualified augmented data using centroid measurement criterion[C]//2021 8th International Conference on Information, Cybernetics, and Computational Social Systems (ICCSS). Beijing, China. IEEE, 2021: 435-440.
12	Zhang X R, Song C Y, Zhao J, et al. Gaussian mixture continuously adaptive regression for multimode processes soft sensing under time-varying virtual drift[J]. Journal of Process Control, 2023, 124: 1-13.
13	Yu J, Qin S J. Multimode process monitoring with Bayesian inference-based finite Gaussian mixture models[J]. AIChE Journal, 2008, 54(7): 1811-1829.
14	Yuan X F, Ge Z Q, Song Z H. Soft sensor model development in multiphase/multimode processes based on Gaussian mixture regression[J]. Chemometrics and Intelligent Laboratory Systems, 2014, 138: 97-109.
15	Zhu J L, Ge Z Q, Song Z H. Variational Bayesian Gaussian mixture regression for soft sensing key variables in non-Gaussian industrial processes[J]. IEEE Transactions on Control Systems Technology, 2017, 25(3): 1092-1099.
16	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.
17	Lu J, Liu A J, Dong F, et al. Learning under concept drift: a review[J]. IEEE Transactions on Knowledge and Data Engineering, 2019, 31(12): 2346-2363.
18	Kadlec P, Grbić R, Gabrys B. Review of adaptation mechanisms for data-driven soft sensors[J]. Computers & Chemical Engineering, 2011, 35(1): 1-24.
19	Minku L L, White A P, Yao X. The impact of diversity on online ensemble learning in the presence of concept drift[J]. IEEE Transactions on Knowledge and Data Engineering, 2010, 22(5): 730-742.
20	Shao W M, Tian X M. Semi-supervised selective ensemble learning based on distance to model for nonlinear soft sensor development[J]. Neurocomputing, 2017, 222: 91-104.
21	Krogh A, Vedelsby J. Neural network ensembles, cross validation, and active learning[C]//Advances in Neural Information Processing Systems. MIT Press, 1994.
22	Yuan X F, Ge Z Q, Song Z H. Locally weighted kernel principal component regression model for soft sensing of nonlinear time-variant processes[J]. Industrial & Engineering Chemistry Research, 2014, 53(35): 13736-13749.
23	Hazama K, Kano M. Covariance-based locally weighted partial least squares for high-performance adaptive modeling[J]. Chemometrics and Intelligent Laboratory Systems, 2015, 146: 55-62.
24	Zhang X, Li Y, Kano M. Quality prediction in complex batch processes with just-in-time learning model based on non-gaussian dissimilarity measure[J]. Industrial & Engineering Chemistry Research, 2015, 54(31): 7694-7705.
25	Xiong W L, Zhang W, Xu B G, et al. JITL based MWGPR soft sensor for multi-mode process with dual-updating strategy[J]. Computers & Chemical Engineering, 2016, 90: 260-267.
26	Yeo W S, Saptoro A, Kumar P, et al. Just-in-time based soft sensors for process industries: a status report and recommendations[J]. Journal of Process Control, 2023, 128: 103025.
27	Jin H P, Chen X G, Yang J W, et al. Adaptive soft sensor modeling framework based on just-in-time learning and kernel partial least squares regression for nonlinear multiphase batch processes[J]. Computers & Chemical Engineering, 2014, 71: 77-93.
28	Jiang J S, Jiang Q C. Variational Bayesian probabilistic modeling framework for data-driven distributed process monitoring[J]. Control Engineering Practice, 2021, 110: 104778.
29	Guo F, Wei B, Huang B. A just-in-time modeling approach for multimode soft sensor based on Gaussian mixture variational autoencoder[J]. Computers & Chemical Engineering, 2021, 146: 107230.
30	Yuan X F, Wang Y L, Yang C H, et al. Probabilistic density-based regression model for soft sensing of nonlinear industrial processes[J]. Journal of Process Control, 2017, 57: 15-25.
31	Bishop C M. Pattern Recognition and Machine Learning[M]. New York: Springer, 2006.

[1]	张晗, 张淑宁, 刘珂, 邓冠龙. 基于慢特征分析与最小二乘支持向量回归集成的草酸钴合成过程粒度预报[J]. 化工学报, 2024, 75(6): 2313-2321.
[2]	董益斌, 熊敬超, 王敬宇, 汪守康, 王亚飞, 黄群星. 融合激光雷达料位测算的锅炉燃烧优化模型预测控制[J]. 化工学报, 2024, 75(3): 924-935.
[3]	蒙西, 王岩, 孙子健, 乔俊飞. 基于注意力模块化神经网络的城市固废焚烧过程氮氧化物排放预测[J]. 化工学报, 2024, 75(2): 593-603.
[4]	肖拥君, 时兆翀, 万仁, 宋璠, 彭昌军, 刘洪来. 反向传播神经网络用于预测离子液体的自扩散系数[J]. 化工学报, 2024, 75(2): 429-438.
[5]	王林, 江荣鼎, 张春晓, 李修真, 谈莹莹. 含R1234yf混合工质汽液相平衡的混合规则评估与预测研究[J]. 化工学报, 2024, 75(2): 475-483.
[6]	刘起超, 张世博, 周云龙, 李昱庆, 陈聪, 冉议文. 起伏振动水平管气液两相流型及转变机理[J]. 化工学报, 2024, 75(2): 493-504.
[7]	尹刚, 钱中友, 曹文琦, 全鹏程, 许亨权, 颜非亚, 王民, 向禹, 向冬梅, 卢剑, 左玉海, 何文, 卢润廷. 基于Adaboost-PSO-SVM的铝电解槽健康状态诊断方法研究[J]. 化工学报, 2024, 75(1): 354-365.
[8]	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.
[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]	郭雨莹, 敬加强, 黄婉妮, 张平, 孙杰, 朱宇, 冯君炫, 陆洪江. 稠油管道水润滑减阻及压降预测模型修正[J]. 化工学报, 2023, 74(7): 2898-2907.
[11]	李艳辉, 丁邵明, 白周央, 张一楠, 于智红, 邢利梅, 高鹏飞, 王永贞. 非常规服役超临界锅炉的微纳尺度腐蚀动力学模型建立及应用[J]. 化工学报, 2023, 74(6): 2436-2446.
[12]	于源, 陈薇薇, 付俊杰, 刘家祥, 焦志伟. 几何相似涡流空气分级机环形区流场变化规律研究及预测[J]. 化工学报, 2023, 74(6): 2363-2373.
[13]	吴心远, 刘奇磊, 曹博渊, 张磊, 都健. Group2vec：基于无监督机器学习的基团向量表示及其物性预测应用[J]. 化工学报, 2023, 74(3): 1187-1194.
[14]	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.
[15]	史克年, 郑景元, 钱宇, 杨思宇. 基于马尔可夫链的蒸汽动力系统两阶段随机规划[J]. 化工学报, 2023, 74(2): 807-817.