用于间歇过程的拓扑时间卷积网络软测量方法  

Soft Sensor Method Based on Topology-guided Temporal Convolutional Network for Batch Process

作　　者：贾明伟 徐丹雅 杨涛[2] 刘毅[1] JIA Mingwei;XU Danya;YANG Tao;LIU Yi(Engineering Research Center of Process Equipment and Remanufacturing,Ministry of Education,Zhejiang University of Technology,Hangzhou 310023,China;State Key Laboratory of Synthetical Automation for Process Industries,Northeastern University,Shenyang 110819,China)

机构地区：[1]浙江工业大学过程装备及其再制造教育部工程研究中心,浙江杭州310023 [2]东北大学流程工业综合自动化全国重点实验室,辽宁沈阳110819

出　　处：《控制工程》2024年第7期1237-1243,共7页Control Engineering of China

基　　金：国家自然科学基金资助项目(62022073,61873241)。

摘　　要：在间歇过程中,变量的相关性和时序性通常演变为变量间的交叉相关性。在间歇过程的数据驱动软测量建模中,捕捉交叉相关性可以提升模型的透明度,因此提出基于拓扑引导时间卷积网络的软测量方法。首先,使用条件格兰杰因果关系检验构建变量关系的拓扑结构,并使用自注意力机制完善拓扑图。然后,使用图注意力机制以每个节点为中心构建子图,并通过时间卷积捕捉子图中变量间的交叉相关性。最后,使用图解释器对模型预测逻辑的物理一致性做出评估。基于青霉素发酵过程的质量预测进行实验,实验结果验证了所提方法的有效性和优越性,证明了模型的预测逻辑符合过程机理。In batch processes,the correlation and timing of variables usually evolve into cross-correlation between variables.In the data-driven soft sensor modeling of batch processes,the capture of cross-correlation can improve the model transparency,so a soft sensor method based on topology-guided temporal convolutional network is proposed.Firstly,conditional Granger causality test is used to construct the topology of variable relationships,and a self-attention mechanism is used to supplement the topology.Then,a graph attention mechanism is used to establish subgraphs centered on each node and the cross-correlations between variables in the subgraph are captured by temporal convolution.Finally,a graph explainer is used to evaluate the physical consistency of the model prediction logic.Based on the quality prediction of penicillin fermentation process,the effectiveness and superiority of the proposed method are verified by the experimental results,and the prediction logic of the model is consistent with the process mechanism.

关 键 词：软测量 时间卷积网络 图注意力机制 自注意力机制 图解释器 间歇过程 

分 类 号：TP18[自动化与计算机技术—控制理论与控制工程]