事件驱动的深度信念网络软测量模型设计方法

doi:10.11949/0438-1157.20241122

摘要/Abstract

摘要：

针对复杂化工过程非平稳性、事件驱动性导致的关键指标参数难以精确软测量的问题，提出了一种事件驱动的深度信念网络（event-driven deep belief network，EDDBN）软测量模型设计方法。首先，获取化工过程运行数据并搭建深度信念网络（driven deep belief network，DBN）模型，以数据驱动的方式对DBN模型进行训练，获得基于DBN的软测量模型。其次，根据DBN模型的训练误差变化特性定义事件，当积极事件发生时会加速当前模型参数的学习步长，当消极事件发生时会跳过当前数据样本并直接进入下一时刻的数据样本学习。这种事件驱动的选择性学习策略不仅能够有效地优化软测量模型训练过程，而且还能降低计算复杂度。同时，通过构造基于马尔可夫链的动态学习过程，分析任意连续两次事件对应输出性能势之差的有界性，给出了EDDBN训练过程的收敛性分析。最后，将EDDBN软测量模型用于湿法烟气脱硫系统二氧化硫（SO₂）浓度软测量实验，结果表明所提出的EDDBN软测量模型能够在非平稳运行工况下实现对SO₂浓度快速、精确地预测分析，并且计算复杂度在数据集①和数据集②上分别降低约63.83%和63.33%。

关键词: 事件驱动的学习, 深度信念网络, 软测量, 化工过程, 湿法烟气脱硫系统

Abstract:

Aiming at the problem that the key parameters of complex chemical process are difficult to be accurately soft-measured due to the non-stationarity and event-driven characteristics, an event-driven deep belief network (EDDBN) soft-sensing model is proposed. First, the operating data of chemical process is obtained and a deep belief network (DBN) model is built. The DBN model is trained in a data-driven way to obtain a soft sensor model based on DBN. Second, some events are defined based on the training-error characteristics of the DBN model. The learning step of parameters in DBN model will be accelerated when the positive events occur, and skip the current data sample and directly go to the next data sample. This event-driven selective learning strategy not only efficiently optimizes the training process of soft-sensing model, but also reduces the computational complexity. Meanwhile, this paper analyzes the boundedness of difference between performance potentials from two consecutive events by construct Markov chain-based dynamicl earning process, which gives convergence analysis of EDDBN training process. Finally, the EDDBN-based soft-sensing model is used to predict the concentration of SO₂ in wet-flue-gas desulfurization system. The results show that it can efficiently and accurately predict the concentration of SO₂ under such non-stationary operating conditions, and the computational complexities of data set ① and data set ② are nearly reduced by 63.83% and 63.33%, respectively.

Key words: event-driven learning, deep belief network, soft-sensing, chemical process, wet-flue-gas desulfurization system

中图分类号:

TP 183

李征, 庄铠泽, 赵东杰, 宋燕星, 王功明. 事件驱动的深度信念网络软测量模型设计方法[J]. 化工学报, 2025, 76(4): 1693-1701.

Zheng LI, Kaize ZHUANG, Dongjie ZHAO, Yanxing SONG, Gongming WANG. Design method of event-driven deep belief network soft-sensing model[J]. CIESC Journal, 2025, 76(4): 1693-1701.

图/表 7

参考文献 30

1	中华人民共和国国务院. 国家重大科技基础设施建设中长期规划(2012—2030年)[Z/OL]. [2013-02-23]. .
	The State Council of the People's Republic of China. National medium and long term plan for major science and technology infrastructure construction (2012—2030)[Z/OL]. [2013-02-23]. .
2	李文俊, 赵中阳, 倪震, 等. 基于气-液传质强化的湿法烟气脱硫CFD模拟研究[J]. 化工学报, 2024, 75(2): 505-519.
	Li W J, Zhao Z Y, Ni Z, et al. CFD numerical simulation of wet flue gas desulfurization: performance improvement based on gas-liquid mass transfer enhancement[J]. CIESC Journal, 2024, 75(2): 505-519.
3	Yi H, Su Z G, Wang P H. Closed-loop shaping-based active disturbance rejection control with delay robustness and application to wet desulfurization process[J]. IEEE Transactions on Industrial Electronics, 2024, 71(6): 6117-6127.
4	Cai Y K, Lv L, Lu X P. The effects of inner electrode diameter on the performance of dielectric barrier discharge reactor for desulfurization and denitrification[J]. IEEE Transactions on Plasma Science, 2021, 49(2): 786-793.
5	Yamamoto H, Kuroki T, Fujishima H, et al. Pilot-scale NO _x and SO _x aftertreatment using a two-phase ozone and chemical injection in glass-melting-furnace exhaust gas[J]. IEEE Transactions on Industry Applications, 2019, 55(6): 6295-6302.
6	Qu J Y, Qi N N, Zhang K, et al. Wet flue gas desulfurization performance of 330 MW coal-fired power unit based on computational fluid dynamics region identification of flow pattern and transfer process[J]. Chinese Journal of Chemical Engineering, 2021, 29: 13-26.
7	Cheng T, Zhou X C, Yang L J, et al. Transformation and removal of ammonium sulfate aerosols and ammonia slip from selective catalytic reduction in wet flue gas desulfurization system[J]. Journal of Environmental Sciences, 2020, 88: 72-80.
8	王功明, 李文静, 乔俊飞. 基于PLSR自适应深度信念网络的出水总磷预测[J]. 化工学报, 2017, 68(5): 1987-1997.
	Wang G M, Li W J, Qiao J F. Prediction of effluent total phosphorus using PLSR-based adaptive deep belief network[J]. CIESC Journal, 2017, 68(5): 1987-1997.
9	Wang G M, Chen H, Han H G, et al. Predicting water quality with nonstationarity: event-triggered deep fuzzy neural network[J]. IEEE Transactions on Fuzzy Systems, 2024, 32(5): 2690-2699.
10	Yuan X F, Qi S B, Wang Y L. Stacked enhanced auto-encoder for data-driven soft sensing of quality variable[J]. IEEE Transactions on Instrumentation and Measurement, 2020, 69(10): 7953-7961.
11	Wang G M, Chen H, Jiang S L, et al. Neurodynamics-driven prediction model for state evolution of coastal water quality[J]. IEEE Transactions on Instrumentation and Measurement, 2024, 73: 2519409.
12	蒙西, 乔俊飞, 韩红桂. 基于类脑模块化神经网络的污水处理过程关键出水参数软测量[J]. 自动化学报, 2019, 45(5): 906-919.
	Meng X, Qiao J F, Han H G. Soft measurement of key effluent parameters in wastewater treatment process using brain-like modular neural networks[J]. Acta Automatica Sinica, 2019, 45(5): 906-919.
13	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.
14	汤健, 乔俊飞. 基于选择性集成核学习算法的固废焚烧过程二英排放浓度软测量[J]. 化工学报, 2019, 70(2): 696-706.
	Tang J, Qiao J F. Dioxin emission concentration soft measuring approach of municipal solid waste incineration based on selective ensemble kernel learning algorithm[J]. CIESC Journal, 2019, 70(2): 696-706.
15	夏恒, 汤健, 崔璨麟, 等. 基于宽度混合森林回归的城市固废焚烧过程二英排放软测量[J]. 自动化学报, 2023, 49(2): 343-365.
	Xia H, Tang J, Cui C L, et al. Soft sensing method of dioxin emission in municipal solid waste incineration process based on broad hybrid forest regression[J]. Acta Automatica Sinica, 2023, 49(2): 343-365.
16	韩红桂, 陈治远, 乔俊飞, 等. 基于区间二型模糊神经网络的出水氨氮软测量[J]. 化工学报, 2017, 68(3): 1032-1040.
	Han H G, Chen Z Y, Qiao J F, et al. Soft-sensor method for effluent ammonia nitrogen based on interval type-2 fuzzy neural networks[J]. CIESC Journal, 2017, 68(3): 1032-1040.
17	曹跃, 陈志文, 袁小锋, 等. 部分子块通讯的分布式PCA厂级工业过程监测方法[J]. 控制与决策, 2020, 35(6): 1281-1290.
	Cao Y, Chen Z W, Yuan X F, et al. Distributed PCA for plant-wide processes monitoring with partial block communication[J]. Control and Decision, 2020, 35(6): 1281-1290.
18	Liu S, Zhong W Q, Chen X, et al. Multiobjective economic model predictive control using utopia-tracking for the wet flue gas desulphurization system[J]. Chinese Journal of Chemical Engineering, 2023, 54: 343-352.
19	Wang Q Y, Zhao H, Zhao Q L, et al. Prediction of SO₂ emission concentration in industrial flue gas based on deep learning: the ammonia desulfurization system of the Yunnan aluminum carbon plant as the research object[J]. Process Safety and Environmental Protection, 2024, 185: 340-349.
20	Dabadghao V, Biegler L T, Bhattacharyya D. Multiscale modeling and nonlinear model predictive control for flue gas desulfurization[J]. Chemical Engineering Science, 2022, 252: 117451.
21	Yin X H, Sun K, Li S Y, et al. Enhancing deep learning for the comprehensive forecast model in flue gas desulfurization systems[J]. Control Engineering Practice, 2023, 138: 105587.
22	王康成, 尚超, 柯文思, 等. 化工过程深度神经网络软测量的结构与参数自动调整方法[J]. 化工学报, 2018, 69(3): 900-906.
	Wang K C, Shang C, Ke W S, et al. Automatic structure and parameters tuning method for deep neural network soft sensor in chemical industries[J]. CIESC Journal, 2018, 69(3): 900-906.
23	耿志强, 徐猛, 朱群雄, 等. 基于深度学习的复杂化工过程软测量模型研究与应用[J]. 化工学报, 2019, 70(2): 564-571.
	Geng Z Q, Xu M, Zhu Q X, et al. Research and application of soft measurement model for complex chemical processes based on deep learning[J]. CIESC Journal, 2019, 70(2): 564-571.
24	Wang K, Yuan X F, Chen J, et al. Supervised and semi-supervised probabilistic learning with deep neural networks for concurrent process-quality monitoring[J]. Neural Networks, 2021, 136: 54-62.
25	Wang G M, Jia Q S, Zhou M C, et al. Artificial neural networks for water quality soft-sensing in wastewater treatment: a review[J]. Artificial Intelligence Review, 2022, 55(1): 565-587.
26	Wang G M, Bi J, Jia Q S, et al. Event-driven model predictive control with deep learning for wastewater treatment process[J]. IEEE Transactions on Industrial Informatics, 2023, 19(5): 6398-6407.
27	Wang G M, Jia Q S, Zhou M C, et al. Soft-sensing of wastewater treatment process via deep belief network with event-triggered learning[J]. Neurocomputing, 2021, 436: 103-113.
28	贾庆山, 杨玉, 夏俐, 等. 基于事件的优化方法简介及其在能源互联网中的应用[J]. 控制理论与应用, 2018, 35(1): 32-40.
	Jia Q S, Yang Y, Xia L, et al. A tutorial on event-based optimization with application in energy Internet[J]. Control Theory & Applications, 2018, 35(1): 32-40.
29	Hinton G E, Osindero S, Teh Y W. A fast learning algorithm for deep belief nets[J]. Neural Computation, 2006, 18(7): 1527-1554.
30	Cao X R. Stochastic learning and optimization—a sensitivity-based approach[J]. Annual Reviews in Control, 2009, 33(1): 11-24.

方法	训练集① RMSE/(mg/m³)	训练集参数有效更新次数	预测集① RMSE/(mg/m³)	预测集① 运行时间/s
EDDBN	1.2865±0.2043	217±6(63.83%±1%)↓	1.4689±0.2716	11.47±0.96
FNN	3.5482±0.4637	600	4.2716±0.5163	15.89±1.35
DBN	3.1692±0.3407	600	3.8649±0.3729	19.54±1.12
DFNN	2.3914±0.3127	600	3.1208±0.3451	18.12±1.05
LSTMNN	4.8724±0.5841	600	6.1520±0.6913	16.37±1.48
RBFNN	4.2319±0.5164	600	5.7628±0.6228	17.35±1.39
BPNN	6.3514±0.7236	600	7.8267±0.8132	23.56±1.87
ESN	3.4226±0.5437	600	4.1692±0.6557	14.93±1.43
DESN	2.7163±0.3916	600	3.1376±0.4158	17.86±1.26

方法	训练集① RMSE/(mg/m³)	训练集参数有效更新次数	预测集① RMSE/(mg/m³)	预测集① 运行时间/s
EDDBN	1.2865±0.2043	217±6(63.83%±1%)↓	1.4689±0.2716	11.47±0.96
FNN	3.5482±0.4637	600	4.2716±0.5163	15.89±1.35
DBN	3.1692±0.3407	600	3.8649±0.3729	19.54±1.12
DFNN	2.3914±0.3127	600	3.1208±0.3451	18.12±1.05
LSTMNN	4.8724±0.5841	600	6.1520±0.6913	16.37±1.48
RBFNN	4.2319±0.5164	600	5.7628±0.6228	17.35±1.39
BPNN	6.3514±0.7236	600	7.8267±0.8132	23.56±1.87
ESN	3.4226±0.5437	600	4.1692±0.6557	14.93±1.43
DESN	2.7163±0.3916	600	3.1376±0.4158	17.86±1.26

方法	训练集② RMSE/(mg/m³)	训练集参数有效更新次数	预测集② RMSE/(mg/m³)	预测集② 运行时间/s
EDDBN	1.3098±0.2371	220±5(63.33%±1%)↓	1.5215±0.2833	11.32±0.94
FNN	3.5203±0.4529	600	4.2518±0.5015	15.97±1.37
DBN	3.1829±0.3561	600	3.8973±0.3869	19.67±1.13
DFNN	2.6718±0.3371	600	3.3824±0.3652	18.75±1.09
LSTMNN	4.8892±0.5974	600	6.1745±0.7108	16.49±1.50
RBFNN	4.2031±0.4972	600	5.7403±0.6102	17.21±1.37
BPNN	6.3764±0.7398	600	7.8429±0.8357	23.82±1.90
ESN	3.4972±0.5564	600	4.1875±0.6794	15.13±1.47
DESN	2.9258±0.4107	600	3.4507±0.4378	18.35±1.29

方法	训练集② RMSE/(mg/m³)	训练集参数有效更新次数	预测集② RMSE/(mg/m³)	预测集② 运行时间/s
EDDBN	1.3098±0.2371	220±5(63.33%±1%)↓	1.5215±0.2833	11.32±0.94
FNN	3.5203±0.4529	600	4.2518±0.5015	15.97±1.37
DBN	3.1829±0.3561	600	3.8973±0.3869	19.67±1.13
DFNN	2.6718±0.3371	600	3.3824±0.3652	18.75±1.09
LSTMNN	4.8892±0.5974	600	6.1745±0.7108	16.49±1.50
RBFNN	4.2031±0.4972	600	5.7403±0.6102	17.21±1.37
BPNN	6.3764±0.7398	600	7.8429±0.8357	23.82±1.90
ESN	3.4972±0.5564	600	4.1875±0.6794	15.13±1.47
DESN	2.9258±0.4107	600	3.4507±0.4378	18.35±1.29

[1]	王法正, 隋璘, 熊伟丽. 面向多采样率数据的TTPA-LSTM软测量建模[J]. 化工学报, 2025, 76(4): 1635-1646.
[2]	王大芬, 唐莉丽, 张鑫焱, 聂春雨, 李明珠, 吴菁. 基于时差的多输出tri-training异构软测量建模[J]. 化工学报, 2025, 76(3): 1143-1155.
[3]	马君霞, 李林涛, 熊伟丽. 基于Tri-training GPR的半监督软测量建模方法[J]. 化工学报, 2024, 75(7): 2613-2623.
[4]	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.
[5]	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.
[6]	李诗浩, 吴振华, 赵展烽, 吴洪, 杨冬, 石家福, 姜忠义. 化工过程中的电子传递、质子传递和分子传递[J]. 化工学报, 2024, 75(3): 1052-1064.
[7]	李文华, 叶洪涛, 罗文广, 刘乙奇. 基于MHSA-LSTM的软测量建模及其在化工过程中的应用[J]. 化工学报, 2024, 75(12): 4654-4665.
[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]	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.
[10]	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.
[11]	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.
[12]	李祥宇, 隋璘, 马君霞, 熊伟丽. 基于时序迁移与双流加权的ONLSTM软测量建模[J]. 化工学报, 2023, 74(11): 4622-4633.
[13]	叶凌箭. 化工过程的自优化控制：原理、发展与应用展望[J]. 化工学报, 2023, 74(11): 4445-4465.
[14]	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.
[15]	王龙洋, 蒙西, 乔俊飞. 基于改进集合经验模态分解和深度信念网络的出水总磷预测[J]. 化工学报, 2021, 72(5): 2745-2753.