CIESC Journal ›› 2023, Vol. 74 ›› Issue (9): 3841-3854.DOI: 10.11949/0438-1157.20230729

• Process system engineering • Previous Articles     Next Articles

Industrial data driven transition state detection with multi-mode switching of a hydrocracking unit

Yue CAO1(), Chong YU2, Zhi LI1(), Minglei YANG1()   

  1. 1.Key Laboratory of Smart Manufacturing in Energy Chemical Process, Ministry of Education, East China University of Science and Technology, Shanghai 200237, China
    2.Global Institute of Future Technology, Shanghai Jiao Tong University, Shanghai 200240, China
  • Received:2023-07-13 Revised:2023-09-03 Online:2023-11-20 Published:2023-09-25
  • Contact: Zhi LI, Minglei YANG


曹跃1(), 余冲2, 李智1(), 杨明磊1()   

  1. 1.华东理工大学能源化工过程智能制造教育部重点实验室,上海 200237
    2.上海交通大学溥渊未来技术学院,上海 200240
  • 通讯作者: 李智,杨明磊
  • 作者简介:曹跃(1992—),男,博士,助理研究员,
  • 基金资助:


The hydrocracking unit has many operating conditions and frequent switching. There must be a transition state when switching between different stable working conditions, which may cause fluctuations in the operating status of the unit and even cause accidents. Usually, experts will judge that the hydrocracking unit is currently in a stable or transitional state based on expert experience, and adopt corresponding monitoring and adjustment strategies, respectively. However, manual judgment has shortcomings due to individual differences and long experience accumulation period, which may lead to inaccurate transition state judgment. Therefore, a transition state detection method for multi-mode switching of the hydrocracking unit is proposed in this paper. First, combined with industrial big data and device process mechanism, wavelet based noise reduction and smoothing are used for industrial data collection, and then correlation analysis and principal component analysis (PCA) are used to reduce data dimensionality, which extracts the extra highly correlated variables that increase calculation cost and decrease information interference. Combining moving window split and moving variance computation with K-means clustering, transition state detection of the hydrocracking unit is realized. Finally, compared with classical K-means clustering and hierarchical clustering, the proposed method has better performance on transition state detection.

Key words: hydrocracking unit, transition state, principal component analysis, moving variance, integration, algorithm, process control


加氢裂化装置运行工况众多且切换频繁,而不同稳定的工况间切换必然存在过渡状态,可能会引起装置运行状态波动,甚至引发事故。通常,操作员会基于专家经验判断装置当前处于稳定或过渡状态,分别采取相应的监控和调整策略。然而,人工判断具有个体差异、经验积累周期长等不足,可能导致过渡状态判断不够准确,故提出一种加氢裂化装置多工况切换过渡状态检测方法,首先,结合工业大数据和装置过程机理,针对工业采集数据运用了小波降噪和平滑,再利用相关分析法和主元分析(principal component analysis,PCA)进行数据降维,剥离了相关性强的变量所带来的额外计算成本和信息干扰,将滑动窗拆分并计算移动方差,再与K-means(K均值)聚类相结合,实现了加氢裂化装置的过渡态检测。最后,与经典K-means聚类和层次聚类方法进行对比验证,证明了所提方法具有更好检测能力。

关键词: 加氢裂化装置, 过渡状态, 主元分析, 移动方差, 集成, 算法, 过程控制

CLC Number: