(112e) An Exchanger-Centric Superstructure for Heat Exchanger Network Synthesis

Heating and cooling requirements form an important component of the energy consumption in industries. There is a potential for energy conservation by synergising these heating and cooling needs through heat integration Although heat integration has been studied since 1970s, there is still opportunity to develop novel superstructures for mathematical programming-based heat exchange network synthesis (HENS). The two most commonly used superstructures for heat integration are the stage-based superstructure of Yee and Grossmann [1] and the match-based superstructure of Floudas, et al. [2]. However, they have some drawbacks in case studies such as the crude-preheating train.

Crude-preheating train has at the most two cold streams but many hot streams. As a result, the network will have many exchangers in series for the cold crude stream. A stage-based superstructure will require many stages to accommodate all possible heat exchange network configurations. This results in a large mixed-integer nonlinear programming (MINLP) model for HEN. Furthermore, the existing stage-based superstructures do not allow certain configurations such as multiple exchangers in a substream. Similarly, the match-based superstructure does not allow repeated matches which may exist in a crude preheating train. Thus, a new superstructure is desirable. In this work, we not only revise but also make successful a novel exchanger-centric superstructure proposed by Huang and Karimi [3], which is neither match-centric nor stage-centric.

An exchanger-centric superstructure assumes a pool of exchangers to which hot and cold streams are assigned via binary variables. It has the potential to allow all possible complex structures as it considers all possible interconnections among the exchangers in the pool. Based on which of this interconnection has non-zero flow, the configurations of the exchangers would be decided. We modified the superstructure of Huang and Karimi [3] and developed an efficient mathematical mixed-integer nonlinear programming (MINLP) model that considers novel configurations such as repeated matches, cross flows, bypasses, etc.

We also developed a heuristic outer-approximation strategy [4] to solve this non-convex MINLP model. Our strategy takes inspiration from the global optimization strategy of Castillo, et al. [5] for solving nonlinear scheduling problems and Viswanathan and Grossmann [4] outer-approximation strategy for non-convex MINLP problems. The basic idea is to create an expanding pool of potential HEN configurations by solving a MILP relaxation of our original model. Then, we solve an NLP for each configuration by fixing the binaries in the original MINLP. We test and compare our model with case studies in the literature and achieve substantial improvements in the total annual costs.

We tested our model on case study from Faria, et al. [6] and Kim and Bagajewicz [7]. Faria, et al. [6] used a partition-based global optimization strategy for stagewise superstructure and Kim and Bagajewicz [7] used the same strategy for an extension of generalized superstructure of Floudas, et al. [2]. This case study has high utility costs and close hot and cold composite curves over a wide range of temperature. Even for such tight curves, our network could reduce utilities owing to the novel configurations that our superstructure allows. These configurations ensured almost parallel temperature profile for hot and cold streams in a counter-current exchanger. As this case study involves unusually high utility costs, the TAC reduction is drastic (67.5% lower). The novel HEN configurations highlight the advantages of the exchanger-centric superstructure. Furthermore, this superstructure can be easily extended for operational optimisation.

[1] T. F. Yee and I. E. Grossmann, "Simultaneous optimization models for heat integration--II. Heat exchanger network synthesis," Computers and Chemical Engineering, vol. 14, no. 10, pp. 1165-1184, 1990.

[2] C. A. Floudas, A. R. Ciric, and I. E. Grossmann, "Automatic synthesis of optimum heat exchanger network configurations," AIChE Journal, vol. 32, no. 2, pp. 276-290, 1986.

[3] K. F. Huang and I. A. Karimi, "Simultaneous synthesis approaches for cost-effective heat exchanger networks," Chemical Engineering Science, Article vol. 98, pp. 231-245, 2013.

[4] J. Viswanathan and I. E. Grossmann, "A combined Penalty Function and Outer Approximation Method for MINLP Optimization," Comput. Chem. Eng., vol. 14, p. 769, 1990.

[5] P. A. C. Castillo, P. M. Castro, and V. Mahalec, "Global Optimization of Nonlinear Blend-Scheduling Problems," Engineering, vol. 3, no. 2, pp. 188-201, 2017.

[6] D. C. Faria, S. Y. Kim, and M. J. Bagajewicz, "Global Optimization of the Stage-wise Superstructure Model for Heat Exchanger Networks," Industrial & Engineering Chemistry Research, vol. 54, no. 5, pp. 1595-1604, 2015/02/11 2015.

[7] S. Y. Kim and M. Bagajewicz, "Global optimization of heat exchanger networks using a new generalized superstructure," Chemical Engineering Science, vol. 147, pp. 30-46, 2016.