Distributed model predictive control‐oriented network decomposition based on full dynamic response. Issue 1 (15th November 2022)

• Record Type: Journal Article
• Authors: Wang, Jiaorao; Song, Chunyue; Zhao, Jun; Mo, Ziyan; Xu, Zuhua
• Is Part Of: AIChE journal. Volume 69:Issue 1(2023)
• Page Start: n/a
• Page End: n/a
• Publication Date: 2022-11-15
• Subjects: community detection -- distributed model predictive control -- input–output interactions -- network decomposition; Chemical engineering -- Periodicals; Génie chimique -- Périodiques; 660.28
• DOI: 10.1002/aic.17951
• Languages: English
• ISSNs: 0001-1541

Abstract

Abstract: This article addresses network decomposition for distributed model predictive control (DMPC), which includes two improvements. First, in the weighted input–output bipartite graph construction of a process network, a new measure called frequency affinity is proposed to characterize the input–output interaction considering the full dynamic response and structural information of a process. Then, in community detection, which is used to decompose the process network, the gap metric is added to quantify stability and the loss of control performance of each subsystem. Through the proposed decomposition, the obtained subsystems can be dynamically well‐decoupled since both transient and steady‐state responses are measured by the frequency affinity. As structural information is considered, the decomposition is consistent with the process physical topology. Furthermore, the utilization of gap metric can facilitate controller design for DMPC. Case studies on a reactor separator process and an air separation process demonstrate the effectiveness of the proposed decomposition method.