Reduced order modeling of a pressure column of an air separation unit using the Dynamic Edmister Method. (June 2023)

• Record Type: Journal Article
• Authors: Kender, Robert; Stops, Laura; Krespach, Valentin; Wunderlich, Bernd; Pottmann, Martin; Ecker, Anna-Maria; Rehfeldt, Sebastian; Klein, Harald
• Is Part Of: Computers & chemical engineering. Volume 174(2023)
• Publication Date: 2023-06
• Subjects: Air separation unit -- Model reduction -- Pressure-driven simulation -- Dynamic simulation -- Digital twin -- Dynamic Edmister Method; Chemical engineering -- Data processing -- Periodicals; 660.0285
• DOI: 10.1016/j.compchemeng.2023.108250
• Languages: English
• ISSNs: 0098-1354

Abstract

Abstract: Accurate physical dynamic models applied as digital twins are vital for developing advanced control strategies and optimizing dynamic plant operation. First principle models are often large-scale, complex, computationally expensive, and not real-time capable. As a result, they need to be reduced for application in optimization and control. This article introduces the Dynamic Edmister Method, which is a generic numerically robust model reduction approach used for dynamic distillation columns. In illustrative case studies, the Dynamic Edmister Method is applied to a digital twin of the pressure column of an air separation unit. The method is validated by simulations of a start-up procedure and a step test. Both case studies show that the reduced models accurately reproduce the dynamic and steady-state behavior of the underlying rigorous model while reducing the calculation effort by 70%. The model reduction approach presented is a significant first step towards using digital twins for dynamic optimization. Highlights: Development of a novel model reduction approach for distillation columns. Dynamic adaptation of the Edmister short-cut method. Integration of the model reduction approach into the ASU digital twin concept. Up to 70% reduction of the required CPU time in a column start-up and step test scenario. Accurate approximation of the column's process dynamics and steady state profiles.