A distributed feasible-side convergent modifier-adaptation scheme for interconnected systems, with application to gas-compressor stations. (12th July 2018)

• Record Type: Journal Article
• Authors: Milosavljevic, Predrag; Schneider, René; Cortinovis, Andrea; Faulwasser, Timm; Bonvin, Dominique
• Is Part Of: Computers & chemical engineering. Volume 115(2018)
• Page Start: 474
• Page End: 486
• Publication Date: 2018-07-12
• Subjects: Real-time optimization -- Interconnected systems -- Modifier adaptation -- Model uncertainty -- Proximal-gradient algorithm -- Gas compressors; Chemical engineering -- Data processing -- Periodicals; 660.0285
• DOI: 10.1016/j.compchemeng.2018.05.006
• Languages: English
• ISSNs: 0098-1354

Abstract

Highlights: The equivalence of modifier adaptation (MA) scheme and proximal-gradient algorithms is shown. A convergence analysis of centralized MA schemes in the case of inexact plant gradients is given. A distributed MA scheme for the real-time optimization of interconnected systems is proposed. The application of the method on a load-sharing optimization of serial gas compressors is discussed. Abstract: Recently, a feasible-side globally convergent modifier-adaptation scheme has been proposed for the real-time optimization of uncertain systems. We show that this scheme is related to proximal-gradient algorithms in numerical optimization and we exploit this relationship to analyze its convergence in the case of inexact gradient information. We also make use of this relationship to propose a novel distributed modifier-adaptation algorithm for interconnected systems that uses a coordinator and knowledge of the interconnection topology. We then prove its feasible-side convergence to plant optimality. In addition, our distributed algorithm ensures confidentiality of local models and data. We finally demonstrate the applicability and effectiveness of this algorithm on a load-sharing optimization case study for serially connected gas compressors.