Optimization-based disjoint and overlapping epsilon decompositions of large-scale dynamical systems via graph theory

• Published in: Journal of parallel and distributed computing Vol. 193; p. 104953
• Main Authors: Maleki, Sahar, Zarabadipour, Hassan, Rahmani, Mehdi
• Format: Journal Article
• Language: English
• Published: Elsevier Inc 01.11.2024
• Subjects: Disjoint and overlapping epsilon decompositions; Graph theory; Large-scale systems; Optimization; Graph theory; Large-scale systems; Disjoint and overlapping epsilon decompositions; Optimization
• ISSN: 0743-7315
• DOI: 10.1016/j.jpdc.2024.104953

To address the complexity challenge of a large-scale system, the decomposition into smaller subsystems is very crucial and demanding for distributed estimation and control purposes. This paper proposes novel optimization-based approaches to decompose a large-scale system into subsystems that are either weakly coupled or weakly coupled with overlapping components. To achieve this goal, first, the epsilon decomposition of large-scale systems is examined. Then, optimization frameworks are presented for disjoint and overlapping decompositions utilizing bipartite graphs. Next, the proposed decomposition algorithms are represented for particular cases of large-scale systems using directed graphs. In contrast to the existing user-based techniques, the proposed optimization-based methods can reach the solution rapidly and systematically. At last, the capability and efficiency of the proposed algorithms are investigated by conducting simulations on three case studies, which include a practical distillation column, a modified benchmark model, and the IEEE 118-bus power system. •Novel optimization-based algorithms are proposed for decompositions of large-scale systems.•The disjoint epsilon decomposition is presented using the bipartite graph of the large-scale system.•The overlapping epsilon decomposition is also investigated via bipartite graph.•The decompositions via directed graphs are presented for particular cases of large-scale systems.