(397b) Optimization of Nonlinear, Variable-Cost Network Models of the Chemical Manufacturing and Refining Industry | AIChE

(397b) Optimization of Nonlinear, Variable-Cost Network Models of the Chemical Manufacturing and Refining Industry


Skouteris, A. - Presenter, The University of Texas at Austin
Giannikopoulos, I., The University of Texas at Austin
Allen, D., The University of Texas at Austin
Baldea, M., The University of Texas at Austin
Stadtherr, M., The University of Texas at Austin
In recent years, technological advancements in hydraulic fracturing and horizontal drilling have lead to rapid increases in oil and gas production in the United States, particularly from shale formations [1]. Shale gas is typically rich in natural gas liquids (NGLs), which are comprised of light alkanes other than methane (ethane, propane, butanes and, in smaller amounts, C5+ hydrocarbons), molecules that constitute some of the most important blocks of the chemical manufacturing industry. As a result of the shale gas boom, the production of NGLs has also significantly increased. The increase in the availability of shale gas and associated NGLs has thus provided a unique opportunity to expand the U.S. chemical manufacturing industry [2,3].

The overall aim of this work is to develop a framework for applying systems optimization and design principles to model the industry to provide insight on how to strategically achieve this expansion. This is particularly important given the scale of capital investment involved in industry expansions. Assessing the potential for adoption of new technologies that can make optimal use of NGL resources requires a holistic analysis of the industry. Thus, network models are at the core of such decision making frameworks, due to their ability to capture existing and potential pathways between different manufacturing processes [4].

The petrochemical industry itself is a highly complex, interconnected system of chemical manufacturing and refining processes. Network models of the industry can capture these interconnections and are hence employed in this work instead of a conventional techno-economic comparison of different processes, which does not easily capture industry-wide effects [5]. Network models represent the industry as a directed graph, with nodes corresponding to manufacturing processes, connected via edges that correspond to material flows between the nodes. Nodes are characterized by a set of input-output coefficients – which represent the process material and energy balances – and a net production cost, while edges are characterized by a flow rate. Given the linear nature of the input-output treatment of the transformations taking place at each node, assuming the costs to be constant results in a linear network that can be formulated as a linear program (LP) [6].

In recent work [7], we developed and used an optimization-based network superstructure model of the entire United States, seeking to minimize total industry cost and involving several hundred of the highest-volume chemicals and hundreds of potential processing technologies. A key innovation used was that – in contrast to previous work (e.g. [5,6]) – process costs and material prices were considered to be variable and were allowed to respond to external perturbations in the network structure (e.g., adding a new process). As demonstrated in [7], this can lead to more realistic results but also renders the problem nonlinear. Moreover, it is assumed that the price of a material is driven by the production cost at its largest producer. Since the largest producer may change as a function of the network operating configuration, this leads to a potentially discontinuous dependence of material prices and process costs on the network configuration and process utilizations. To solve the problem, we used a successive linear programming (SLP) approach that alternates between solving a constant-cost LP and a cost-propagation algorithm that updates process costs and material prices based on the LP results.

In recent preliminary work [8], we investigated, and successfully tested on a small network, an alternative approach in which the optimization problem is reformulated as a mixed-integer nonlinear program (MINLP) that can be solved using standard software packages. In this paper, we extend this preliminary work, and explore and compare the SLP and MINLP approaches on a more extensive set of test problems. Furthermore, in order to improve the SLP approach, we describe an alternative cost propagation technique that is capable of more effectively capturing the discontinuous behavior of the price changes due to shifts in the largest producer of a material. This new approach is compared, in terms of both reliability and computational efficiency, to the original approach presented in [7] and to the MINLP approach presented in [8]. These comparisons are based on prototype network examples that capture the network structures encountered in industry, which are then scaled up to more practical cases.


[1] EIA, “Natural gas explained: Where our natural gas comes from.” https://www.eia.gov/energyexplained/natural-gas/where-our-natural-gas-comes-from.php, 2022. [Online; accessed 07-April-2022].

[2] M. Yang and F. You, “Comparative techno-economic and environmental analysis of ethylene and propylene manufacturing from wet shale gas and naphtha” Industrial & Engineering Chemistry Research, vol. 56, no. 14, pp. 4038-4051, 2017.

[3] J. J. Siirola, 2014. The Impact of Shale Gas in the Chemical Industry. AIChE Journal 60, 810–819.

[4] M. A. Stadtherr and D. F. Rudd, “Systems study of the petrochemical industry,” Chemical Engineering Science, vol. 31, no. 11, pp. 1019-1028, 1976.

[5] S. E. DeRosa and D. T. Allen, “Impact of natural gas and natural gas liquids supplies on the United States chemical manufacturing industry: Production cost effects and identification of bottleneck intermediates,” ACS Sustainable Chemistry & Engineering, vol. 3, no. 3, pp. 451-459, 2015.

[6] S. E. DeRosa and D. T. Allen, “Impact of new manufacturing technologies on the petrochemical industry in the united states: A methane-to-aromatics case study,” Industrial & Engineering Chemistry Research, vol. 55, no. 18, pp. 5366-5372, 2016.

[7] A. Skouteris, I. Giannikopoulos, T. F. Edgar, M. Baldea, D. T. Allen, M. A. Stadtherr, 2021. Systems analysis of natural gas liquid resources for chemical manufacturing: Strategic utilization of ethane. Industrial & Engineering Chemistry Research 60 (33), 12377–12389.

[8] Skouteris, A.; Giannikopoulos, I.; Allen, D. T.; Baldea, M.; Stadtherr, M. A. MINLP framework for systems analysis of the chemical manufacturing industry using network models. Proceedings of the 32nd European Symposium on Computer Aided Process Engineering (ESCAPE32). Toulouse, France, 2022


This work is supported in part by the National Science Foundation under Cooperative Agreement No. EEC-1647722 (CISTAR – NSF Engineering Research Center for Innovative and Strategic Transformation of Alkane Resources, http://cistar.us). Any opinions, findings and conclusions or recommendations expressed in this material are those of the author(s) and do not necessarily reflect the views of the National Science Foundation.