基于人工预测调控策略的复杂化工过程多任务学习方法

doi:10.11949/0438-1157.20221340

摘要/Abstract

摘要：

当前，PID反馈控制依然是化工生产过程的主要控制策略。然而，由于复杂化工过程通常具有大时滞和非线性等特性，使得PID控制对于一些关键过程参数控制的性能不佳。为此，在实际工程中通常是现场操作人员依据自身经验对其实施人工预测调控。为了能够从历史调控数据中学习人工预测调控策略，提出了一种多任务学习级联网络（LSTM multi-task network cascades，LSTM-MNC）。根据过程变量长短期不同趋势建立预测短期变化和长期趋势的过程预测模型，并学习过程预测模型估计信息与操纵变量序列的因果关系，由被控变量偏差预测支持操纵变量序列生成，实现生产过程的智能化调控。在工业换热器过程仿真平台上进行实验，获得了满意的结果，验证了所提方法的有效性。

关键词: 预测调控策略, 多任务学习, 级联结构, 化工过程, 过程控制

Abstract:

PID feedback control has been considered as the primary control strategy for chemical production processes ever since. However, due to the large time-delay and nonlinearity of complex chemical processes, PID control usually suffers poor performances in controlling key process variables. For this reason, in actual engineering, it is usually the on-site operators who implement manual prediction and control based on their own experience. In order to effectively learn manual predictive regulation strategies from historical operating data, the paper proposes a multi-task learning cascade network (LSTM multi-task network cascades, LSTM-MNC). Therein, process prediction models that predict short-term changes and long-term trends are established based on different trends of process variables in the long and short term, and the causal relationship between estimated information of the process prediction model and manipulated variable sequences is extracted. Supported by predictions of controlled variable deviations, manipulated variable sequences are generated, achieving intelligent manipulations of production processes. Experiments were conducted on an industrial heat exchanger process simulation platform, leading to satisfactory results, which verify the effectiveness of the proposed method.

Key words: predictive manipulating strategy, multi-task learning, cascade structure, chemical process, process control

中图分类号:

TQ 028.8

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.

Zhongqiu ZHANG, Hongguang LI, Yilin SHI. A multi-task learning approach for complex chemical processes based on manual predictive manipulating strategies[J]. CIESC Journal, 2023, 74(3): 1195-1204.

图/表 14

图1 常规多任务学习网络图

Fig.1 General diagram of multi-task learning networks

图2 多任务级联网络结构

Fig.2 Multi-tasking cascade network structures

图3 LSTM-MNC网络结构

Fig.3 LSTM-MNC network structures

图4 时序数据线性分割与趋势提取(th=0.03)

Fig.4 Linear segmentations of time series data with trend extractions (th=0.03)

图5 过程预测模型的LSTM-MNC模型

Fig.5 The LSTM-MNC based process prediction model

图6 操纵序列离散化

Fig.6 Discretization of manipulating sequences

图7 调控策略学习LSTM-MNC模型

Fig.7 The LSTM-MNC based manipulating strategy learning model

图8 工业换热器过程流程

Fig.8 Industrial heat exchanger process

表1 过程变量描述

Table 1 Descriptions of process variables

变量	描述	作用
T₃	物料出口温度	被控变量CV
F₁	物料入口流量	扰动变量DV1
T₁	物料入口温度	扰动变量DV2
F₂	冷却液流量	操纵变量MV

图9 LSTM-MNC模型的CVE预测

Fig.9 CVE predictions of the LSTM-MNC model

图10 不同网络模型的CVET预测

Fig.10 CVET predictions with different network models

图11 换热器的调控策略控制结果

Fig.11 Control profiles of the heat exchanger using manipulating strategies

表2 学习的调控策略与人工经验调控性能对比

Table 2 Comparative performances of learned manipulating strategies with manual experienced manipulations

项目	MSE	IAE
人工操纵经验	0.07496	3015
LSTM-MNC策略	0.07414	3017

图12 LSTM-MNC策略与人工经验调控的被控变量曲线

Fig.12 Profiles of LSTM-MNC strategies with manual experienced manipulations for controlled variables

参考文献 36

1	侯忠生, 许建新. 数据驱动控制理论及方法的回顾和展望[J]. 自动化学报, 2009, 35(6): 650-667.
	Hou Z S, Xu J X. On data-driven control theory: the state of the art and perspective[J]. Acta Automatica Sinica, 2009, 35(6): 650-667.
2	Yu T Y, Xu Z H, Zhao J, et al. A fast, fully distributed nonlinear model predictive control algorithm with parametric sensitivity through Jacobi iteration[J]. Journal of Process Control, 2022, 110: 133-153.
3	Cisneros P S G, Werner H. Nonlinear model predictive control for models in quasi‐linear parameter varying form[J]. International Journal of Robust and Nonlinear Control, 2020, 30(10): 3945-3959.
4	戴文战, 娄海川, 杨爱萍. 非线性系统神经网络预测控制研究进展[J]. 控制理论与应用, 2009, 26(5): 521-530.
	Dai W Z, Lou H C, Yang A P. An overview of neural network predictive control for nonlinear systems[J]. Control Theory ＆ Applications, 2009, 26(5): 521-530.
5	Wang Y M, Qing D D. Model predictive control of nonlinear system based on GA-RBP neural network and improved gradient descent method[J]. Complexity, 2021, 2021: 6622149.
6	姚羽曼, 罗文嘉, 戴一阳. 数据驱动方法在化工过程故障诊断中的研究进展[J]. 化工进展, 2021, 40(4): 1755-1764.
	Yao Y M, Luo W J, Dai Y Y. Research progress of data-driven methods in fault diagnosis of chemical process[J]. Chemical Industry and Engineering Progress, 2021, 40(4): 1755-1764.
7	朱坚, 杨博, 王永健, 等. 一种新型的基于Levenshtein距离层次聚类的时序操作优化方法[J]. 化工学报, 2019, 70(2): 581-589.
	Zhu J, Yang B, Wang Y J, et al. New operation optimization method with time series based on Levenshtein distance hierarchical clustering[J]. CIESC Journal, 2019, 70(2): 581-589.
8	Li H G, Tang X J, Zhao W J, et al. Approaches to deep learning based manipulating strategy reconstructions for complex chemical processes[J]. Journal of Process Control, 2021, 107: 127-140.
9	Lin C W, Li T H S, Chen C. Feedback linearization and feedforward neural network control with application to twin rotor mechanism[J]. Transactions of the Institute of Measurement and Control, 2018, 40: 351-362.
10	Fu X G, Li S H, Wunsch D C, et al. Local stability and convergence analysis of neural network controllers with error integral inputs[J]. IEEE Transactions on Neural Networks and Learning Systems, 2021, DOI: 10.1109/TNNLS.2021.3116189 .
11	Xiang W M, Tran H D, Yang X D, et al. Reachable set estimation for neural network control systems: a simulation-guided approach[J]. IEEE Transactions on Neural Networks and Learning Systems, 2021, 32(5): 1821-1830.
12	Hochreiter S, Schmidhuber J. Long short-term memory[J]. Neural Computation, 1997, 9(8): 1735-1780.
13	Lee Y, Choi Y, Cho H, et al. Prediction of distillation column temperature using machine learning and data preprocessing[J]. Korean Chemical Engineering Reserach, 2021, 59(2) : 191-199.
14	Guo J W, Wang W, Tang Y Y, et al. A CNN-Bi_LSTM parallel network approach for train travel time prediction[J]. Knowledge-Based Systems, 2022, 256: 109796.
15	Tao L, Tian G, Aberer K. Hybrid neural networks for learning the local trend in time series[C]// Proceedings of the Twenty-sixth International Joint Conference on Artificial Intelligence. Palo Alto, CA: AAAI Press, 2017: 2273-2279.
16	Li H G, Zhao W J, Shi Y L, et al. MP-CRJ: multi-parallel cycle reservoirs with jumps for industrial multivariate time series predictions[J]. Transactions of the Institute of Measurement and Control, 2022, 44(11): 2093-2105.
17	Wei D, Ma J M, Jiao H Y, et al. Model predictive control for multi-zone variable air volume systems based on artificial neural networks[J]. Journal of Process Control, 2022, 118: 185-201.
18	Jin U K, Jung J H, Kim M, et al. Multi-task learning of classification and denoising (MLCD) for noise-robust rotor system diagnosis[J]. Computers in Industry, 2021, 125: 103385.
19	Ishihara K, Kanervisto A, Miura J, et al. Multi-task learning with attention for end-to-end autonomous driving[C]// 2021 IEEE/CVF Conference on Computer Vision and Pattern Recognition Workshops (CVPRW). Nashville, TN, USA: IEEE, 2021: 2896-2905.
20	Dai J F, He K M, Sun J. Instance-aware semantic segmentation via multi-task network cascades[C]// 2016 IEEE Conference on Computer Vision and Pattern Recognition. Las Vegas, NV, USA: IEEE, 2016: 3150-3158.
21	Khoshkangini R, Mashhadi P, Tegnered D, et al. Predicting vehicle behavior using multi-task ensemble learning[J]. Expert Systems with Applications, 2023, 212: 118716.
22	Chen J, Yu Y X, Guo Q. Freeway traffic congestion reduction and environment regulation via model predictive control[J]. Algorithms, 2019, 12: 220.
23	Liu P F, Qiu X P, Huang X J. Adversarial multi-task learning for text classification[EB/OL]. 2017, .
24	Liu S, Johns E, Davison A J. End-to-end multi-task learning with attention[C]// 2019 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR). IEEE, 2017.
25	Liu S C, Liang Y Y, Gitter A. Loss-balanced task weighting to reduce negative transfer in multi-task learning[C]// Proceedings of the AAAI Conference on Artificial Intelligence. Palo Alto, California, USA: AAAI Press, 2019.
26	Yang P P, Huang K Z, Hussain A. A review on multi-task metric learning[J]. Big Data Analytics, 2018, 3(1) : 1-23.
27	Farhi L, Abbasi H, Rehman R. Smart identity management system by face detection using multitasking convolution network[J]. Security and Communication Networks, 2021, 2021: 7314823.
28	Sgaard A, Goldberg Y. Deep multi-task learning with low level tasks supervised at lower layers[C]// Proceedings of the 54th Annual Meeting of the Association for Computational Linguistics. Association for Computational Linguistics, 2016: 231-235.
29	李龙. 基于过程挖掘的工业过程运行知识发现方法及应用[D]. 北京: 北京化工大学, 2021.
	Li L. Process mining based industrial process operating knowledge discovery method and application[D]. Beijing: Beijing University of Chemical Technology, 2021.
30	Wen M, Li P, Zhang L F, et al. Stock market trend prediction using high-order information of time series[J]. IEEE Access, 2019, 7: 28299-28308.
31	Li L, Su X N, Zhang Y, et al. Trend modeling for traffic time series analysis: an integrated study[J]. IEEE Transactions on Intelligent Transportation Systems, 2015, 16(6): 3430-3439.
32	Liu J, Li T R, Yuan Z, et al. Symbolic aggregate approximation based data fusion model for dangerous driving behavior detection[J]. Information Sciences, 2022, 609: 626-643.
33	Lin J, Keogh E, Wei L, et al. Experiencing SAX: a novel symbolic representation of time series[J]. Data Mining and Knowledge Discovery, 2007, 15(2): 107-144.
34	Keogh E, Kasetty S. On the need for time series data mining benchmarks: a survey and empirical demonstration[J]. Data Mining and Knowledge Discovery, 2003, 7(4): 349-371.
35	Zhang Y, Gallimore M, Bingham C, et al. Hybrid hierarchical clustering—piecewise aggregate approximation, with applications[J]. International Journal of Computational Intelligence and Applications, 2016, 15(4) : 26.
36	李海林, 郭崇慧. 基于云模型的时间序列分段聚合近似方法[J]. 控制与决策, 2011, 26(10): 1525-1529.
	Li H L, Guo C H. Piecewise aggregate approximation method based on cloud model for time series[J]. Control and Decision, 2011, 26(10): 1525-1529.

[1]	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.
[2]	康飞, 吕伟光, 巨锋, 孙峙. 废锂离子电池放电路径与评价研究[J]. 化工学报, 2023, 74(9): 3903-3911.
[3]	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.
[4]	苏晓丹, 朱干宇, 李会泉, 郑光明, 孟子衡, 李防, 杨云瑞, 习本军, 崔玉. 湿法磷酸半水工艺考察与石膏结晶过程研究[J]. 化工学报, 2023, 74(4): 1805-1817.
[5]	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.
[6]	张江淮, 赵众. 碳三加氢装置鲁棒最小协方差约束控制及应用[J]. 化工学报, 2023, 74(3): 1216-1227.
[7]	赵涛岩, 曹江涛, 李平, 冯琳, 商瑀. 区间二型模糊免疫PID在环己烷无催化氧化温度控制系统中的应用[J]. 化工学报, 2022, 73(7): 3166-3173.
[8]	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.
[9]	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.
[10]	王建松, 许锋, 罗雄麟. 化工过程多回路PID控制系统模式切换参数自整定[J]. 化工学报, 2022, 73(4): 1647-1657.
[11]	张兴硕, 罗雄麟, 许锋. 催化裂化装置反再系统动态模拟精细化与控制系统“工艺优先”配对设计[J]. 化工学报, 2022, 73(2): 747-758.
[12]	张成, 潘立志, 李元. 基于加权统计特征KICA的故障检测与诊断方法[J]. 化工学报, 2022, 73(2): 827-837.
[13]	卢道铭, 唐钊艇, 范怡平, 卢春喜. 大差异颗粒分级再生设备的性能研究[J]. 化工学报, 2021, 72(8): 4184-4195.
[14]	李元, 杨东昇, 赵丽颖, 张成. 层次变分高斯混合模型与主多项式分析的故障检测策略[J]. 化工学报, 2021, 72(3): 1616-1626.
[15]	谢苗苗, 张浪文, 谢巍. 复杂非线性系统的子系统分解方法[J]. 化工学报, 2021, 72(3): 1557-1566.