K. Q. Al Anazi, M. B. SHAMS and H. AL MASRI, University of Bahrain, Department of Engineering Management, Isa Town, Bahrain
In this research, the integration of an oil refinery and an ethylene-producing facility was examined using a mixed-integer nonlinear programming (MINLP) model that prioritized maximizing ethylene and propylene production and net profit. To achieve the objectives of this study, two plants were modeled and optimized individually, and then optimization was done for the final integrated plant—three mathematical models were employed to perform the simulation using the MINLP framework.
Part 1 of this article—published in the April issue—focused on the mathematical formulations for this study. This article—Part 2—presents two case studies of a multi-objective optimization model that aims to maximize the profitability and production of ethylene and propylene in an integrated plant.
CASE STUDIES
The following presents two case studies of a multi-objective optimization model that aims to maximize the profitability and production of ethylene and propylene in an integrated plant. The first case study is the maximization of both the net profit and the production of ethylene. The second case study is the maximization of both the net profit and the production of propylene. More details about the methodology of these two case studies are presented in the following sub-sections.
Multi-objective optimization. In this study, a conventional oil refinery and an ethylene production plant in Saudi Arabia were modeled individually and integrated as a single system to achieve profit maximization for both plants. The described models and mathematical formulation were implemented in GAMS version 25.1.2 for optimization, with a compilation time of 33.12 sec for the oil refinery model, 2.6 min for the ethylene production plant and 6.7 min for the integration. The compilation time for the oil refinery and the integration was similar to that discussed by Ketabchi et al. due to the simplification of the model.12 Even though the authors’ compilation time was lower than that of Ketabchi, it is within a good range of compilation times reported in previous studies.15
Not all process units were considered for this study. For the oil refinery, the primary units considered in GAMS comprised the gas desulfurization unit, CRU, FCCU, CDU and blenders, as they produce intermediate streams like ethane, propylene, fuel gas, fuel oil, kerosene, diesel and gasoline. For the ethylene production plant, the main units simulated included the separation units and furnaces, as they produced relevant commodities for analysis, such as benzene, butadiene, propylene, ethylene, C4 and C5. All three models were mathematically formulated, and each model was implemented separately to compare and optimize the plants individually. Each scenario was formulated as an MINLP problem and solved with the BARON solver.
Many parameters, factors and assumptions were considered in the MINLP model, including the process yield, product specifications, demand, supply, capacity and price of materials purchased from or sold to the oil refinery or the ethylene production plant. Also considered in the model were the chemical and physical properties of the materials and their intermediates. The values of these parameters were not calculated but were extracted from the work of Ketabchi et al.12 Several assumptions were made to simplify the computation of the modeling. This study assumed that the values of all parameters used in the model were fixed, and that all process units were working satisfactorily. Since this study only investigated mass integration, no utility systems were considered.
Furthermore, insight was drawn from the multi-objective optimization results by comparing performance variables of the oil refinery, ethylene production plant and integrated plant. The performance variables of the single-objective optimization model and the multi-objective optimization model were also compared. The performance variables compared included the quantity of materials produced, the unit capacity of each plant, the operating cost of each plant, the inventory cost of each plant, the material cost of each plant, the amount of materials transferred from one plant to the other and the net profit of each plant. The results of this case study are presented and discussed in the next section.
Results and discussion. The results of the multi-objective scenarios, production optimization and profit maximization of ethylene and propylene are presented and discussed in this section. First, visual representations for the production and profit maximization of ethylene and propylene following plant integration are presented. Next, the multi-objective results of ethylene production maximization and propylene production maximization are presented, followed by a comparison of the two outcomes. Lastly, comparisons between single-objective scenarios and multi-objective scenarios are detailed.
Net profit and production of a propylene-based/ethylene-based optimization model. The amount of ethylene and propylene produced simultaneously in the integrated plant must be optimized to help shareholders maximize profits. First, the authors evaluated a multi-objective scenario in which ethylene’s net profit and production are given optimization preference. The results in FIG. 6 indicate that considering both the net profit and the production volume of ethylene offers better optimization results. The highest net profit for ethylene and propylene obtained from the integrated plant is $840,337/d. At this rate, approximately 174,040 bpd of ethylene and 83,400 bpd of propylene can be produced in contrast to the optimal rate of $821,152.60/d, in which the amount of ethylene and propylene produced are 197,969 bpd and 130,157 bpd, respectively. Overall, the findings suggested that the integrated plant can produce 13.7% more ethylene and 56% more propylene, even though the optimal profit ($821,152.6/d) deviated from the highest profit (840,337/d) by 2.34%.
Next, the authors considered a multi-objective scenario for the integrated plant in which propylene profit and production optimization are prioritized. The results shown in FIG. 7 are similar to the results depicted in FIG. 6. Stated another way, the results of the investigation indicated that ethylene and propylene can be produced more optimally when both production and profit are prioritized (net profit = $840,112.20/d, ethylene production = 197,910 bpd, and propylene production = 130,320 bpd) than when either production or profit is the most important consideration.
The multi-objective results for ethylene and propylene were compared to select the most efficient optimization model. As shown in FIGS. 8 and 9, the net profit of either ethylene or propylene for propylene-prioritized optimization ($840,112.20/d) was significantly greater than that for ethylene-prioritized optimization ($821,152.60/d). However, the production of ethylene or propylene for ethylene-prioritized optimization was 0.03% more than that for propylene-prioritized optimization. Given that this production difference is nearly negligible, the propylene-based outcome is the best option to maximize the net profit of both commodities in an integrated plant.
Comparison of single-objective and multi-objective results. The results shown in TABLE 2 are indicative of the performance of plant integration based on profit maximization (single objective) and that of plant integration based on profit and production maximization (multiple objectives). The findings indicated that the single-objective model results outperformed the multi-objective model results. As shown in TABLE 2, the net profit and production of ethylene and propylene for multi-objective models were significantly greater than those for a single-objective model. For example, in terms of net profit, the integrated plant generated $799,647.60/d for a profit-maximization scenario and either $821,152.60/d (ethylene-prioritized optimization) or $840,112.20/d (propylene-prioritized optimization) for a profit and production maximization scenario. Nonetheless, the increase in propylene production for the multi-objective optimization vs. single-objective optimization is more substantial than the increase in ethylene production for the multi-objective optimization vs. single-objective optimization.
The increases in ethylene and propylene production for the multi-objective model—compared to the single-objective model—can be traced to the output of the oil refinery. As shown in FIG. 10, the amount of intermediate streams (H2, CG and fuel oil) flowing from the petrochemical plant into the refinery is significantly higher for multi-objective optimization (profit and production maximization) than for single-objective optimization (profit maximization). Based on this production disparity, an oil refinery that produces more fuel oil and gasoline earns more net income.
The assessment of the feedstocks of the ethylene production plant also provides further explanation for the increase in the overall production of ethylene and propylene in the multi-objective optimization model of the integrated plant. As presented in FIG. 11, the amount of raw materials [light straight-run naphtha (LSRN), fuel gas, ethane, atmospheric gasoil (AGO) and coker gasoil (CGO)] fed into the petrochemical plant in the multi-objective optimization model is higher than that in the single-objective optimization model. The only exception is propylene transfer, whose production in the single-objective optimization model is greater than its production in the multi-objective optimization model (FIG. 11). Despite this drawback, the transfer of a greater quantity of LSRN, fuel gas, ethane, AGO and CGO from the oil refinery to the petrochemical plant increases the output of the final products (fuel gas, propylene, butane, butadiene, pentane, benzene and ethylene) in the petrochemical plant (FIG. 12).
As shown in FIG. 11, the values of LSRN, fuel gas, ethane, AGO and CGO in multi-objective calculations are higher than those in single-objective calculations for intermediates flowing out of the oil refinery into the petrochemical plant. A significant contribution is made by transferring CGO from the oil refinery into the petrochemical plant in a multi-objective scenario. CGO and other intermediates transferred from the oil refinery to the petrochemical plant can result in higher final production in the petrochemical plant, as shown in FIG. 12. When single-objective calculations are made, propylene transfer is the only material with a higher value.
FIG. 13 illustrates the oil refinery outputs for the single-objective optimization and multi-objective optimization models. As shown, there is a significant increase in gasoline and fuel oil production for the multi-objective optimization model compared to the single-objective optimization model. It should also be noted that the multi-objective optimization model for propylene-based production maximization produces more gasoline (gasoline 90 octane and gasoline 95 octane) than that for ethylene-based production maximization. Nevertheless, the drawback of adopting the multi-objective optimization model is that no diesel-0 is produced.
A breakdown of the economic variables for each optimization model of the oil refinery and petrochemical plant is presented in FIGS. 14 and 15, respectively. For the oil refinery, the unit operating cost, inventory cost and total income obtained from the multi-objective optimization model are significantly greater than those derived from the single-objective optimization model (FIG. 14). For the petrochemical plant, the unit operating cost, inventory cost and total income obtained from the multi-objective optimization model are slightly more than those derived from the single-objective optimization model (FIG. 15). These results suggest that the multi-objective optimization model of an integrated plant offers better economic performance than the single-objective optimization model of an integrated plant.
Takeaways. This study proposed a multi-objective optimization approach to maximize the net profit and production of ethylene and propylene in an integrated plant in the Middle East. This method transformed three production scenarios into MINLP problems using GAMS. The results showed that the multi-objective optimization model offers better economic performance than the single-objective model. The study also found that the multi-objective model for propylene-prioritized optimization produced more ethylene, propylene and other petrochemical plant outputs than the ethylene-prioritized model. However, tradeoffs must be made when adopting a single-objective model, a multi-objective model for propylene-prioritized optimization or a multi-objective model for ethylene-prioritized optimization. The study recommends adopting a multi-objective optimization model—particularly one prioritizing propylene production—for oil and petrochemical companies to boost profits, lower operating costs and reduce natural gas emissions.
Limitations. This research is not immune to limitations. This study focused primarily on optimizing materials produced in the integrated plant, thereby ignoring the optimization of the energy produced in the processing units. The authors hypothesized that the simultaneous optimization of energy and materials in the integrated plant can result in a higher net profit. Future studies should explore this hypothesis. Even though the compilation time is within a good range, a more simplified MINLP algorithm should be developed in future research to obtain computational results in a shorter period. HP
LITERATURE CITED