烟气脱硫过程含硫量自适应预测与异常状态溯源方法

doi:10.11949/0438-1157.20251260

摘要/Abstract

摘要：

烟气脱硫过程是控制二氧化硫（SO₂）排放的核心举措，对改善空气质量、保护生态环境具有显著作用，其出气口SO₂浓度状态直接决定着脱硫过程的成效及是否达标。针对烟气脱硫系统多工况运行、状态交互耦合、噪音干扰等特点导致的出气口SO₂浓度难以精确预测与异常状态分析困难等问题，本文提出了一种基于增量式深度模糊神经网络（Incremental deep fuzzy neural network， IDFNN）的烟气脱硫过程含硫量自适应预测与异常状态溯源方法。首先，基于烟气脱硫过程实际运行数据，设计数据驱动的增量式深度特征提取器（Incremental deep feature extractor， IDFE），用以自适应提取原始数据中的关键深层特征，克服状态耦合与噪音对数据效率的影响。其次，将IDFE提取到的关键特征作为模糊神经网络（Fuzzy neural network， FNN）的输入，对含硫量动力学特性进行高效地有监督学习，进而实现对未来时间段的预测任务。同时，针对含硫量状态的预测信息，设计基于互信息度量的含硫量状态关联性统计分析方法，实现IDFNN模型对输入项的敏感度分析，进而实现对异常状态的溯源分析。最后，将所提方法用于实际烟气脱硫过程含硫量预测与异常状态溯源分析实验，结果显示IDFNN不仅能在复杂运行环境下实现对含硫量的精确预测，还能实现对异常状态诱因变量的溯源和筛选。

关键词: 烟气脱硫过程, 含硫量预测与状态溯源, 增量式深度特征提取器, 模糊神经网络, 互信息度量

Abstract:

The flue gas desulfurization process (FGDP) is a core measure to control the emission of SO₂, which plays a significant role in improving air quality and protecting the ecological environment. The SO₂concentration at its outlet directly determines the effectiveness of FGDP and whether it meets the standards. In view of the problems such as the difficulties in accurately predicting the SO₂ concentration at the outlet and analyzing abnormal states caused by the multi-condition operation, state interaction and coupling as well as noise interference of FGDP, this paper proposes effective methods of adaptive prediction and abnormal states traceability for sulphur content in FGDP based on incremental deep fuzzy neural network (IDFNN). First, a data-driven incremental deep feature extractor (IDFE) is designed based on real data of FGDP to adaptively extract the key deep features from the raw data. Second, the extracted features are considered as the inputs of fuzzy neural network (FNN), which can efficiently approximate the dynamic characteristics of sulphur content by supervised learning and further realize the prediction task for future time periods. Meanwhile, a statistical analysis method for the correlation of sulfur content based on mutual information measurement is designed to analyze the model sensitivity to inputs, which can realize the abnormal states traceability. Finally, the proposed INFNN is used to predict the sulphur content and analyze the abnormal states traceability in real FGDP, the results show that IDFNN does not only accurately predict the sulfur content in complex operating environments, but also traces and screens the inducing variables of abnormal states.

Key words: flue gas desulfurization process, sulfur content prediction and state traceability, incremental deep feature extractor, fuzzy neural network, mutual information measurement

中图分类号:

TP 183

王功明, 刘起浩, 陈红, 韩红桂, 乔俊飞. 烟气脱硫过程含硫量自适应预测与异常状态溯源方法[J]. 化工学报, DOI: 10.11949/0438-1157.20251260.

Gongming WANG, Qihao LIU, Hong CHEN, Honggui HAN, Junfei QIAO. Methods of adaptive prediction and abnormal states traceability for sulphur content in flue gas desulfurization process[J]. CIESC Journal, DOI: 10.11949/0438-1157.20251260.

图/表 11

图1 湿法烟气脱硫系统流程图

Fig. 1 Flow chart of WFGD process

图2 深度特征提取器

Fig. 2 Deep feature extractor

图3 模糊神经网络拓扑结构

Fig. 3 Topology structureof FNN

图4 IDFE的一步增长过程

Fig. 4 One-step growing process of IDFE

图5 SO2浓度关联变量的原始数据

Fig. 5 Raw data of related variables for SO2 concentration

图6 IDFE从原始输入变量中提取的特征

Fig. 6 Extracted features from raw input data by IDFE

表1 IDFE与普通深度特征提取器针对SO2关联变量特征提取中的对比结果

Table 1 Comparison results of IDFE with DFEs in extracting features from variables related to SO2

IDFE

DFE1

DFE2

DFE3

DFE4

0.0823±0.0037

0.5107±0.0872

0.1894±0.0316

0.1017±0.0072

0.0995±0.0084

12-9-9-9-3 (自适应)

12-6-6-6-3 (固定)

12-10-10-10-3 (固定)

12-16-16-16-6 (固定)

12-20-20-20-6 (固定)

18.26±0.93

16.97±1.28

29.52±1.56

48.41±1.79

106.83±2.62

1.51±0.03

6.75±0.19

4.26±0.37

3.18±0.45

2.25±0.68

图7 IDFNN对SO2浓度的训练结果及误差

Fig. 7 SO2 training results and errors of IDFNN

图8 IDFNN对SO2浓度的预测结果及误差

Fig. 8 SO2 prediction results and errors of IDFNN

表2 IDFNN与其他模型在烟气脱硫过程SO2浓度预测中的对比结果

Table 2 Comparison results of IDFNN with the others in predicting SO2 concentration

预测模型

训练集误差RMSE/(mg/m³)

预测集误差

RMSE/ (mg/m³)

预测集运行

时间/s

模型结构

IDFNN

FNN

RBFNN

DBN

ADBN

EDBN

SDFNN

1.5974±0.2104

2.1905±0.4813

3.5147±0.7592

2.4672±0.5819

2.4189±0.4317

1.8962±0.3112

1.9358±0.2762

1.9816±0.2859

2.6287±0.5402

4.3851±0.8739

2.9503±0.7924

2.8544±0.6503

2.3608±0.3912

2.3794±0.3506

29.72±0.93

38.59±1.84

36.85±1.97

49.36±1.62

35.28±1.46

33.59±1.43

31.62±1.25

12-9-9-9-3-8-8-1

12-30-30-1

12-65-1

12-28-28-28-1

12-36-36-1

12-32-32-10-1

12-20-16-10-8-8-1

图9 IDFNN输出SO2浓度与输入变量间的关联度指数

Fig. 9 Correlation index between SO2 and input variables of IDFNN model

参考文献 26

[1]	关于推进实施焦化行业超低排放的意见[Z/OL]. 环大气〔2024〕 5号.
[2]	Wang G M, Li X Y, Gu K, et al. Neurodynamics-driven model predictive control with soft-measurement for desulfurization system[J]. IEEE Transactions on Automation Science and Engineering, 2025, 22: 19546-19554.
[3]	李征, 庄铠泽, 赵东杰, 等. 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.
	Li Z, Zhuang K Z, Zhao D J, et al. Design method of event-driven deep belief network soft-sensing model[J]. CIESC Journal, 2025, 76(4): 1693-1701.
[4]	Yamamoto H, Kuroki T, Fujishima H, et al. Pilot-scale NO _x and SOx 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.
[5]	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.
[6]	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噁英排放软测量[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.
[7]	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噁英排放浓度软测量[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.
[8]	田昊, 汤健, 夏恒, 等. 基于IT2FBLS强化学习PID的MSWI过程炉膛温度控制[J]. 自动化学报, 2025, 51(7): 1626-1641.
	Tian H, Tang J, Xia H, et al. Furnace temperature control using IT2FBLS-based reinforcement learning PID for MSWI process[J]. Acta Automatica Sinica, 2025, 51(7): 1626-1641.
[9]	Wang G M, Zhao Y D, Liu C X, et al. Data-driven robust adaptive control with deep learning for wastewater treatment process[J]. IEEE Transactions on Industrial Informatics, 2024, 20(1): 149-157.
[10]	朱霁霖, 桂卫华, 蒋朝辉, 等. 基于料面视频图像分析的高炉异常状态智能感知与识别[J]. 自动化学报, 2024, 50(7): 1345-1362.
	Zhu J L, Gui W H, Jiang Z H, et al. Intelligent perception and recognition of blast furnace anomalies via burden surface video image analysis[J]. Acta Automatica Sinica, 2024, 50(7): 1345-1362.
[11]	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.
[12]	张璐, 张嘉成, 韩红桂, 等. 基于模糊神经网络的污水处理生化除磷过程控制[J]. 化工学报, 2020, 71(3): 1217-1225.
	Zhang L, Zhang J C, Han H G, et al. FNN-based process control for biochemical phosphorus in WWTP[J]. CIESC Journal, 2020, 71(3): 1217-1225.
[13]	伍小龙, 姜韦宇, 韩红桂, 等. 基于随机事件代理模型的污水处理鼓风机轴承温度智能预测[J]. 中国科学: 技术科学, 2025, 55(5): 879-892.
	Wu X L, Jiang W Y, Han H G, et al. Intelligent prediction of wastewater treatment blower bearing temperature based on random event agent model[J]. Scientia Sinica (Technologica), 2025, 55(5): 879-892.
[14]	Gu K, Liu Y C, Liu H Y, et al. Air pollution monitoring by integrating local and global information in self-adaptive multiscale transform domain[J]. IEEE Transactions on Multimedia, 2025, 27: 3716-3728.
[15]	Said Y, Alassaf Y, Saidani T, et al. Context-aware feature extraction network for high-precision UAV-based vehicle detection in urban environments[J]. Computers, Materials and Continua, 2024, 81(3): 4349-4370.
[16]	Han H G, Sun C X, Wu X L, et al. Dynamic–static model for monitoring wastewater treatment processes[J]. Control Engineering Practice, 2023, 132: 105424.
[17]	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.
[18]	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.
[19]	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.
[20]	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.
[21]	Vuokila N, Cunning C, Zhang J, et al. The application of neural networks to the modeling of magnetic hysteresis[J]. IEEE Transactions on Magnetics, 2024, 60(3): 7300604.
[22]	Bertipaglia A, Alirezaei M, Happee R, et al. An unscented Kalman filter-informed neural network for vehicle sideslip angle estimation[J]. IEEE Transactions on Vehicular Technology, 2024, 73(9): 12731-12746.
[23]	王功明, 乔俊飞, 关丽娜, 等. 深度信念网络研究现状与展望[J].自动化学报, 2021, 47(1): 35-49.
	Wang G M, Qiao J F, Guan L N, et al. Review and prospect on deep belief network[J]. Acta Automatica Sinica, 2021, 47(1): 35-49.
[24]	Han H G, Zhang L, Wu X L, et al. An efficient second-order algorithm for self-organizing fuzzy neural networks[J]. IEEE Transactions on Cybernetics, 2019, 49(1): 14-26.
[25]	Wang G M, Qiao J F, Bi J, et al. An adaptive deep belief network with sparse restricted Boltzmann machines[J]. IEEE Transactions on Neural Networks and Learning Systems, 2020, 31(10): 4217-4228.
[26]	Wang G M, Jia Q S, Qiao J F, et al. A sparse deep belief network with efficient fuzzy learning framework[J]. Neural Networks, 2020, 121: 430-440.