Real-time optimization of load sharing for gas compressors in the presence of uncertainty. (15th August 2020)

• Record Type: Journal Article
• Authors: Milosavljevic, Predrag; Marchetti, Alejandro G.; Cortinovis, Andrea; Faulwasser, Timm; Mercangöz, Mehmet; Bonvin, Dominique
• Is Part Of: Applied energy. Volume 272(2020)
• Publication Date: 2020-08-15
• Subjects: Gas compressors -- Optimal load sharing -- Interconnected systems -- Real-time optimization -- Adaptive optimization; Power (Mechanics) -- Periodicals; Energy conservation -- Periodicals; Energy conversion -- Periodicals; 621.042
• DOI: 10.1016/j.apenergy.2020.114883
• Languages: English
• ISSNs: 0306-2619

Abstract

Highlights: A novel approach to load-sharing optimization of serial and parallel compressor plants. Consider operation close to surge conditions, uncertainty of compressor models and varying operating conditions. New insights on how the proposed scheme exploits the interconnection structure/configuration of compressors units. Abstract: This paper investigates the problem of load-sharing optimization of gas compressors in the presence of uncertainty. The objective is to operate a set of compressor units in an energy-efficient way, while at the same time meeting a varying load demand. The main challenge is the fact that the available models, and in particular the compressor efficiency maps, carry a significant amount of uncertainty. For this task, real-time optimization (RTO) techniques that rely on plant measurements and correct the model are available in the literature. This paper is tailored to the application of RTO to the compressor load-sharing optimization problem. An adaptive optimization approach that guarantees optimal plant operation upon convergence is used. To this end, we use appropriate measurements to estimate plant gradients and correct the model in such a way that it exhibits the same optimality conditions as the plant. This way, the challenge is shifted from having an accurate model to being able to estimate experimental gradients accurately. We show how the specific problem structure can be exploited for the purpose of efficient estimation of plant gradients. We consider both parallel and serial compressor configurations as well as operation close to surge constraints. The simulation of an industrial case study demonstrates the efficiency of the proposed approach.