A novel two surfaces hybrid approach for Multi-period heat exchanger networks synthesis by combination of imperialist competitive algorithm and linear programming method. (31st August 2022)

• Record Type: Journal Article
• Authors: Feyli, Babak; Soltani, Hadi; Hajimohammadi, Reza; Fallahi-Samberan, Mehrab
• Is Part Of: Chemical engineering science. Volume 258(2022)
• Publication Date: 2022-08-31
• Subjects: Optimization -- Multi-period heat exchanger networks synthesis -- Total annual cost -- Imperialistic competitive algorithm -- Linear programming; Chemical engineering -- Periodicals; Génie chimique -- Périodiques; Chemical engineering; Periodicals; Electronic journals; 660
• DOI: 10.1016/j.ces.2022.117755
• Languages: English
• ISSNs: 0009-2509

Abstract

Highlights: A two-surface hybrid approach was introduced for multi-period HENs synthesis. It was combined linear programming (LP) + imperialist competitive algorithm (ICA). The ICA is used to find the structural variables on the 1st surface. The continuous variables are handled with a combined LP and ICA on the 2nd surface. The relatively linear behavior improved its ability to deal with HENs problems. Abstract: In this paper, an efficient hybrid approach to deal with multi-period heat exchanger networks (HENs) synthesis, which is inherently formulated as a mixed-integer non-linear programming (MINLP) model, is proposed. In this novel two surfaces approach, the creation and determination of optimal HENs structures are accomplished by an imperialist competitive algorithm (ICA) on the first surface. In each HEN structure, there is a sequence of stages containing the addresses of the exchanger(s). The HENs generated by the ICA are sent to the second surface, where their minimum total annual cost (TAC) is calculated as an overall objective function. This surface works on two levels. For all periods of each HEN, the local optimal solutions are determined based on the maximum energy recovery at the outer level, including an external search loop and a linear programming (LP) model. Based on the outer level results, the ICA is re-used, at the inner level, to find the final minimum TAC of the network. As a result, the MINLP model is transformed into a relatively linear LP + ICA hybrid model, which is easily solvable. The results demonstrate that this approach can sometimes reduce the network TAC even by over 7.2% compared to the literature.