CBE Seminar - Davood B. Pourkargar

Optimization-based Control of Complex Process Networks: Handling Complexity through Model Reduction and System Decomposition

Abstract

Large-scale process networks are becoming the rule rather than the exception in energy and chemicals production industries due to the increasing demand for integrated process operation, which provides sustainable productivity improvement benefits. Process integration makes the resulting networks complex and thus challenging to operate. The complexity originates from both individual unit operations and the strong interactions between them. The first part of the talk is focused on addressing complexity associated with the control of individual units. A novel supervisory control structure is developed to address the control problem of distributed parameter systems (DPSs) in chemical and advanced materials processing. It circumvents the restrictions of currently used control approaches for DPSs via data-assisted model order reduction, i.e. by designing low-dimensional controllers based on recursively updated reduced order models of the governing equations. In the second part of the talk, a systematic framework is developed for control-oriented optimal decomposition of the large-scale process networks into manageable subsystems with desirable interaction characteristics. From a network theory perspective, this problem can be posed as identifying communities of variables which are strongly connected with each other but weakly connected with variables in other communities. This idea is employed to generate hierarchies of system decompositions with different degrees of decentralization based on input-output connectivity; it is a well-suited approach for distributed control design. A comprehensive study is performed to evaluate the impact of system decomposition on distributed model predictive control (DMPC) of integrated plants. The performance and computational needs of the DMPCs for different decompositions are analyzed and compared to a centralized model predictive control (CMPC) design. The optimal decomposition achieved by the community detection method enables closed-loop performance under the DMPC close to that of the CMPC while reducing the computation effort significantly.

Bio

Davood B. Pourkargar is a Postdoctoral Associate in the Department of Chemical Engineering and Materials Science at University of Minnesota. Prior to joining the University of Minnesota, he was a Postdoctoral Researcher in the Catalysis Center for Energy Innovation (CCEI) and Department of Chemical and Biomolecular Engineering at University of Delaware from 2015 to 2016. He received the B.S. and M.S. in Chemical Engineering and Process Simulation and Control with highest honors from Sharif University of Technology in 2008 and 2010, respectively, and the Ph.D. degree in Chemical Engineering from the Pennsylvania State University in 2015. Davood has won several awards, including the Robert F. Smith School Distinguished Junior Researcher Award from Cornell University in 2017, the O. Hugo Schuck Best Paper Award in the application category from the American Automatic Control Council (AACC) in 2014, the Walter R. and Aura Lee Supina Graduate Fellowship in Chemical Engineering from the Pennsylvania State University in 2011, the Best Session Presentation Award at the American Control Conferences in 2013 and 2015, and Travel Awards from the AIChE Computing and Systems Technology (CAST) Division in 2015, Society of Industrial and Applied Mathematics (SIAM) in 2014 and AACC in 2013-2015. His research interests include process dynamics, optimization and control of complex process networks, computational modeling and applied mathematics, system identification, and uncertainty quantification in multiscale modeling, with application to sustainable energy systems and medicine.