Sie sind auf Seite 1von 18

Chemical Engineering Science 60 (2005) 7057 7074 www.elsevier.

com/locate/ces

Flexibility and operability analysis of a HEN-integrated natural gas expander plant


Alp Er S. Konukmana, , Ugur Akmanb
a Department of Energy Systems Engineering, Gebze Institute of Technology, Gebze 41400, Kocaeli, Turkey b Department of Chemical Engineering, Bogazici University, Bebek 34342, Istanbul, Turkey

Received 24 July 2004; received in revised form 20 April 2005; accepted 13 May 2005 Available online 15 August 2005

Abstract In the heat-exchanger network (HEN) literature, synthesis, design, and exibility analyses of HENs are done independently from processes to which HENs are integrated. Such analyses are made mostly based on nominal operating conditions at which the HENs source- and target-stream properties are evaluated. However, terminal-stream properties of HENs depend upon temperatures, pressures, and compositions of the process connected to the HEN. In this work, exibility and operability issues of a HEN are investigated with rigorous simulations using the process owsheet simulator HYSYS for a HEN-integrated natural gas turbo-expander plant (TEP) operating under ethane-recovery mode. The contribution of this work is threefold. First, the HEN-plant interactions are exemplied via the process owsheet simulator. Second, exibility and operability issues are tackled using the optimization capability of the owsheet simulator. Third, for highly energy-integrated complex plants like the TEP, the difculties or impossibilities of automated HEN synthesis and exibility analysis with process owsheet simulators are demonstrated. 2005 Elsevier Ltd. All rights reserved.
Keywords: Natural-gas expander plant; Ethane recovery; Heat-exchanger network; Flexibility; Simulation; Product processing; Optimization; Energy

1. Motivation The synthesis of heat-exchanger networks (HENs) is one of the extensively studied problems in chemical engineering (Furman and Sahinidis, 2002). In the HEN literature, synthesis implies the determination of a network structure along with its design and costing. The objective of the HEN synthesis is the generation of a network of heat exchangers with minimum utility consumption, minimum number of heat-transfer units, or minimum total cost. The vast number of works in the literature are concentrated mostly around the standard denition of the HEN-synthesis problem which may be stated as: Given hot/cold streams to be cooled/heated from specied (nominal) supply temperatures to specied target temperatures, and hot/cold
Corresponding author. Tel.: +90 262 754 2360x2402; fax: +90 262 653 8490. E-mail addresses: konukman@gyte.edu.tr (A.E.S. akman@boun.edu.tr (U. Akman).

Konukman),

utility specications, synthesize a minimum-cost HEN. In this denition, the minimum-cost objective may be taken as minimum total-utility consumption (leading to maximum energy-recovery network) or, more completely, as minimum annualized total (utility and area) cost. Most of the heuristic, algorithmic, evolutionary (Nishida et al., 1977; Linnhoff and Flower, 1978a,b; Gundersen and Naess, 1988; Dolan et al., 1990; Mehta et al., 2001; Li, 2002a,b; Lin and Miller, 2004; Serna and Jimenez, 2004; Wei et al., 2004; Ravagnani et al., 2005; Pettersson, 2005), and automated, mathematical-programming-based, sequential or simultaneous HEN-synthesis techniques (Gundersen et al., 1991; Floudas, 1995; Grossmann et al., 1999; Abbass et al., 1999; Cheung and Hui, 2001; Li and Yao, 2001; Mikkelsen and Qvale, 2001; Bjork and Westerlund, 2002; Shivakumar and Narasimhan, 2002; Sorsak and Kravanja, 2002; Frausto-Hernandez et al., 2003; Mizutani et al., 2003; Furman and Sahinidis, 2004; Serna-Gonzalez et al., 2004) originating from the standard denition of the

0009-2509/$ - see front matter doi:10.1016/j.ces.2005.05.070

2005 Elsevier Ltd. All rights reserved.

7058

A.E.S. Konukman, U. Akman / Chemical Engineering Science 60 (2005) 7057 7074

HEN-synthesis problem consider a single operating point (nominal operating period) and use nominal parameter values (e.g., constant values of source-stream temperatures, heat-capacity-owrates, overall heat-transfer coefcients), and thus, do not take into account the operability and exibility issues. On the other hand, in the limited number of works, the operability and exibility analysis of HENs draw attention mostly indirectly as convenient but challenging test examples for automated, mathematical-programmingbased synthesis, design, retrot-design, exibility-analysis, and multiperiod-optimization algorithms (Swaney and Grossmann, 1985; Floudas and Grossmann, 1986; Grossmann and Floudas, 1987; Papalexandri and Pistikopoulos, 1993; Aguilera and Nasini, 1995; Dimitriadis and Pistikopoulos, 1995; Varvarezos et al., 1995; Uzturk et al., 1996; Glemmestad et al., 1997; Akman and Uygun, 1999; Bansal et al., 2000; Furman and Sahinidis, 2001; Nagy et al., 2001; Oliveira et al., 2001; Tantimuratha et al., 2001; Akman et al., 2002; Castier and Queiroz, 2002; Heggs and Vizcaino, 2002; Karafyllis and Kokossis, 2002; Konukman et al., 2002; Li, 2002a,b; Li et al., 2004; Tantimuratha and Kokossis, 2004; Pintaric and Kravanja, 2004; Chen and Hung, 2004, 2005). In the works mentioned above, HENs are considered as independent from the main process. The inputs to HEN analyses are the terminal-stream properties (source- and target-stream temperatures, owrates, heat capacities), heattransfer coefcients, heat-transfer areas, utility (steam and cooling water) owrates and temperatures, and area and utility cost parameters. For these properties/parameters, either hypothetical (representative) values are used or, at best, they are evaluated at the nominal operating conditions of the main process (e.g., real plant data in the form of terminal temperatures and heat-capacity-owrates) (Briones and Kokossis, 1999; Ozkan and Dincer, 2001; Pettersson, 2005). Thereafter, the process and HEN-process interactions are ignored although the terminal-stream properties of the HEN are dependent upon the temperatures, pressures, and compositions in the process to which the HEN is connected. The term exibility is sometimes used to describe the ability of a process to operate at a countable number of discrete points (multiperiod operability). However, the accepted standard denition of exibility as introduced by Swaney and Grossmann (1985) is that it is a measure of the size of the region of feasible operation in the space of deviations in process parameters from nominal values. Since, in reality, a HEN does not operate independent of the process, operability and exibility of HENs must be analyzed together with the main process by considering HEN-process interactions caused by the propagation of the deviations in the terminal process- and/or HEN-stream parameters. In many instances, there are complex loops and energy recycles between HEN and process, e.g., source- and target-end of a stream of HEN may belong to a pumparound stream in a distillation column. For instance, a

disturbance in one of the source-stream temperatures of the HEN may diffuse into the main process, alter the phase equilibrium in mass-transfer operations, cause changes in stream compositions/pressures, and these changes may affect the terminal-stream properties of the HEN. Besides, an internal disturbance in the process (e.g., reux ratio disturbance) may also affect the HENs terminal-stream properties. Clearly, HEN synthesis, design, and exibility analysis based on the nominal values of the HEN parameters only may yield erroneous results for energy-integrated plants. The motivation behind this work is the lack of case studies in the massive HEN literature that consider HEN-process interactions with rigorous simulations. In the HEN literature, the neglect of the HEN-process interactions is the major reason of considering overwhelmingly the effects of source-stream temperature and owrate variations and overlooking the consequential effects of stream composition and pressure variations. The objective of this work is to demonstrate the computations with regard to operability and exibility analyses of a HEN interacting with the plant using only the rigorous simulations via a commercial sequentialmodular process owsheet simulator and its optimization capability. In this work, steady-state operability and exibility issues of a HEN are investigated with rigorous simulations using the process owsheet simulator HYSYS for a HENintegrated natural gas turbo-expander plant (TEP) operating under ethane-recovery mode. Bypass streams and stream splits in the HEN, selected after a sensitivity analysis, are used to increase exibility and disturbance-rejection ability of the TEP. The emphasis is given to the assessment of exibility of the process in maintaining desired level of cooling in the HEN, and desired levels of ethane recovery and methane rejection under temperature and pressure disturbances in the natural-gas feed stream. It is demonstrated that for such complex and highly energy-integrated plants, the synthesis, design, and exibility analysis of HENs must be done with rigorous simulations considering the plantHEN interactions, which is not widely encountered in the literature. The contribution of this work is threefold. First, the HEN-process interactions are demonstrated with rigorous simulations via the process owsheet simulator. Thus, temperature, pressure, phase-equilibrium, composition, and physical-property effects on HEN streams are handled rigorously. Second, exibility and operability issues are tackled using the optimization capability of the owsheet simulator only, thus preserving the rigorousness of the computations. Third, the difculties or impossibilities of automated HEN synthesis and exibility analysis for highly integrated complex plants like the TEP are discussed in terms of both the sequentialmodular process owsheet simulators and mathematicalprogramming based HEN synthesis and exibility analysis methodologies.

A.E.S. Konukman, U. Akman / Chemical Engineering Science 60 (2005) 7057 7074

7059

2. Natural-gas expander plants The increasing demand for the natural-gas liquids (NGLs), especially for ethane as a petrochemical feedstock, resulted in the construction of TEPs that recover ethane from lean (dry) natural-gas streams at cryogenic temperatures as low as 95 C. The TEPs involve gas turbo-expansion cycles to attain, possibly without external refrigeration, the low temperatures desired for enhanced ethane recovery as high as 95%. The earliest low-temperature expander plant, built about 40 years ago, is still in use, and it is presently the most efcient design for attaining high ethane recovery. Many TEPs built during the last 25 years use a conventional singlestage turbo-expander scheme for moderately high (7080%) ethane recoveries. There are other alternative TEP designs that can achieve higher ethane or propane recoveries, that can process feed streams with relatively high CO2 content without formation of solid CO2 , and that can show exibility to operate either in ethane- or propane-recovery modes with very high methane rejection (Wilkinson and Hudson, 1982, 1992, 1993; Chiu, 1997; Lee et al., 1999; Mehra and Gaskin, 1999; Finn et al., 2000). Maintaining a high ethane recovery is crucial for the TEPs, since even a less than 1% increase in recovery may amount to a signicant increase in prots due to processing of high volumes of feed. In the last two decades, the determination of the optimal operating conditions and the development of more efcient owsheets have been the subject of research. Wang (1985) has concluded that a combination of turbo-expansion and refrigeration leads to a minimal energy consumption. Different TEP designs to improve ethane recovery without inlet CO2 removal have been proposed by Wilkinson and Hudson (1982). Bandoni et al. (1989) considered the synthesis and optimization of ethane recovery process based on energy analysis. Fernandez et al. (1991) presented the optimization of ethane extraction plants from natural gas containing CO2 . Diaz et al. (1995) studied the debottlenecking of ethane recovery plants. Diaz et al. (1997) proposed different expansion alternatives within superstructure for automatic process conguration and debottlenecking of natural-gas processing plants. Eliceche et al. (1998) determined the feasible operating region of natural-gas plants under feed disturbances with optimization techniques. Chebbi et al. (2004) concluded after simulations using HYSYS that signicantly higher ethane recoveries were possible with gasliquid and gasgas heat exchangers and that additional complexities in the design were not necessarily useful in increasing ethane recovery. The literature lacks studies on energy integration of TEPs, particularly the interactions of the HEN with the expander part of the plant. In this work, we adopt the owsheet of a typical, industry-standard single-stage TEP and concentrate on the HEN part of the plant in maintaining a desired level of cooling to achieve a desired level of ethane recovery and methane rejection. For this purpose, bypass streams

together with the stream splits are used to increase the exibility and disturbance-rejection ability of the HEN. The stream splits, bypasses, and the external cooling load, are used as the manipulated variables in controlling the temperature of the HENs main target stream, the level of which dictates the ethane-recovery/methane-rejection values.

3. Description of the TEP The simplied owsheet of the TEP used in this work is shown in Fig. 1. In a typical TEP, the natural gas is cooled to extremely low temperatures through a HEN after which the cold liquid and vapor are separated in a low-temperature separator (LTS). The liquid stream of the LTS is ashed across a JouleThomson (JT) valve for pressure reduction and additional cooling. The vapor stream of the LTS is fed to the expander side of the TEP where temperature is further reduced and the work produced is utilized for recompression. The liquid generated by the expander is separated in the vaporliquid separator (VLS). This liquid as well as the JT valve outlet are both fed to the top of the Demethanizer Tower (DT) (a reboiled absorber), the bottom product of which is the ethane-rich stream. The vapors from the DT and VLS are combined and fed back to the HEN to help cooling of the incoming natural-gas feed stream. The residue-gas (sales gas) stream (rich in methane) leaving the HEN is partially compressed using the energy released in the expander. Efcient heat integration in the HEN, which is used to pre-cool the feed to the LTS, is of major importance in the design of TEPs. The HEN should maximize the cooling of the feed by matching it with the overhead (methane-rich) stream and with the reboiler of the DT. The HEN also helps to maintain high ethane recovery and high methane rejection in the bottoms of the DT. With respect to TEP, the heatintegration efciency may be dened as the ability of the HEN to achieve chilling of the natural-gas feed before entering the expander part of the plant with minimum external cooling requirement.

Fig. 1. Simplied owsheet of the turbo-expander plant.

7060

A.E.S. Konukman, U. Akman / Chemical Engineering Science 60 (2005) 7057 7074


Residue (methane) 22 25 24 70 Natural Gas Feed 1 HEN 13 K-100 17 E-100

As can be understood from the above brief description of a typical TEP, the heat-integration efciency, exibility, and operability of the HEN are of crucial importance in reducing the cost of the operation as well as in maintaining high ethane recovery and methane rejection.

MIX-1 19

20

V-101

OptimizerSpreadsheet

15 18 V-100

4. The HYSYS model of the TEP


VLV-100 T-100 16 Q-81 Q-82 Product (ethane) Q-80

We started with the TEP owsheet/model which was available in the application-case library of the old version of HYSYS (v2.4.1, Build 3870) (G-3: Deep-Cut Turbo-Expander Plant) and for the simulations we used AspenTech-HYSYS (v3.2, Build 5029). We did not change the natural-gas feed-stream composition and the options related to the thermodynamics (Peng-Robinson equation of state, binary-interaction parameters, enthalpy-calculation options, etc). The available HYSYS model had been developed mainly for design purposes rather than for performance calculations. For example, in the HEN part of the TEP owsheet, some approach temperatures had been xed and the UA (the product of the overall heat-transfer coefcient and the overall heat-transfer area) of the heat exchangers were being calculated based on the input/output temperatures. Furthermore, in the HEN, the temperatures of the splits, before mixing of the branches, had been set equal to each other. Also, the split ratios in the HEN were being calculated from the specied approach temperatures and from the heat duties related to side-stream reboilers and the main reboiler of the DT. On the other hand, in the DT, the boil-up ratio, the component ratio (methane/ethane molar ratio in the bottoms product), and the ratios of the heat ows (duty ratios) of the side-stream reboilers to the heat ow of the reboiler had all been specied. In the expander part of the TEP, the pressures of the streams leaving the JT valve and the expander had been specied as well. Since these design specications were not suitable for our performance-calculation purposes, all of the specications were removed and the new HYSYS model suitable for direct simulations (performance calculations) was prepared. We did not change the suggested pressure-drop values for various units such as the heat exchangers, DT trays, etc. The major change we did was in the HEN part of the owsheet; we added bypass streams around the two rigorous heat exchangers which matched the combined cold vapors from the DT and VLS with the relatively hot incoming natural-gas feed stream. The HYSYS process owsheet diagrams of the main TEP, the HEN sub-owsheet, and the DT sub-owsheet are shown in Figs. 24, respectively. In the HEN sub-owsheet (Fig. 3), the exchangers E-100 and E-103 are rigorous heat exchangers. In HYSYS, the rigorous heat-exchanger module can perform two-sided energy and material balance calculations, solve for temperatures, pressures, heat and material stream ows, or UA. The methane-rich residue stream (cold uid) entering/leaving the exchangers (Streams 22, 23, 24) is a gas stream, however,

14

21

Fig. 2. HYSYS process owsheet diagram of the main TEP.

the natural-gas feed stream (hot uid) entering/leaving the exchangers (Streams 2, 6, 8, 12) is of two phase (mostly vapor). Therefore, there are phase changes in the hot (tube) sides of exchangers E-100 and E-103. Exchanger E-101 (Fig. 3) is not a rigorous heat exchanger. In HYSYS, the cooler and heater operations are one-sided heat exchangers; the inlet stream is cooled (or heated) to the required outlet conditions, and the energy stream absorbs (or provides) the enthalpy difference between the two streams. These modules are useful when one is interested only in how much energy is required to cool or heat a process stream with a utility, but not interested in the conditions of the utility itself. Exchanger E-101 supplies heat to the reboiler of the DT (Fig. 4) and provides cooling of the natural-gas stream. We set the duty of this exchanger through the DT specication such that the boil-up ratio in the reboiler was always 0.4. Otherwise (if the duty is set independently as a constant), the complete TEP may exhibit multiple steady-state solutions depending on the natural-gas feed-stream temperature. Therefore, the temperature of Stream 4 is calculated through the heat duty of E-101 (Q-80) as dictated by the boil-up ratio in the DTs reboiler. Exchangers E-102 and E-104 (Fig. 3) are not rigorous heat exchangers either. They represent the side-stream reboilers of the DT (Fig. 4). We set the heat duties of exchangers E102 (Q-82) and E-104 (Q-81) as constants (in accordance with the original HYSYS application-case solution). Their xed duties enter the corresponding trays of the DT as heat inputs (Q-81 to stage 9 and Q-82 to stage 6). Similarly, we set the heat duty of exchanger E-105 (Q-83), which represents the additional cooling by an external cold utility, as constant. Exchanger E-102 and E-105 cannot be lumped into a single exchanger to consider it as a side reboiler of the DT, since this causes either insufcient cooling of the natural gas or excessive vapor formation in the corresponding tray of the DT. It should be mentioned that we also tried considering exchangers E-102 and E-104 as rigorous (two-sided) heat exchangers; by drawing out a certain fraction of the liquid

A.E.S. Konukman, U. Akman / Chemical Engineering Science 60 (2005) 7057 7074

7061

24

29 MIX-3 30 E-100 31 Sales Gas 23

27 MIX-4 28 6 8 TEE-2 9 10 E-102 E-103 22R 26 BP-2 12

RCY-2

22

1 Warm Feed

Tee-1 3

BP-1

7R 7 MIX-1 RCY-1 5

MIX-2 11

13 Low Temperature Product

4 E-101 E-104

E-105

Q-80

Q-81

Q-83

Q-82

Fig. 3. The HYSYS process sub-owsheet diagram of the HEN.

20

18

16 Main DT Q-82

Q-81 Q-80 Boilup 21 To Reboiler Reboiler

Fig. 4. The HYSYS process sub-owsheet diagram of the DT.

leaving stage 5 of the DT and matching it in E-102 with Stream 10 before returning it to stage 6 and by drawing out a certain fraction of the liquid leaving stage 8 and matching it in E-104 with Stream 4 before returning it to stage 9. This more realistic approach did not worked well; it required two additional recycle-convergence blocks which interacted with the other recycle-convergence blocks (RCY-1 and RCY-2 in Fig. 3), created convergence problems, and increased the computation time signicantly. Therefore, for this work, we decided not to consider the side reboilers of the DT as rigorous exchangers and left them as single-sided

coolers with xed duties as explained in the preceding paragraph. Fixing the heat duties of the side reboilers of the DT (Q-81 and Q-82) corresponds to xing the fractions of the liquids leaving stages 5 and 8 for heat exchange in rigorous heat exchangers E-102 and E-104. Fixing either the heat duties or liquid drawal rates of the side reboilers may seem unrealistic. However, the use of heat duties or liquid drawal rates of the side reboilers as manipulated variables should not be preferred since, while attempting to control Stream 13 temperature (hence the ethane recovery) in the HEN before any feed-stream disturbance propagates into the expander part of the plant, their manipulations trigger internal disturbances in the expander part of the plant (which propagate back into the HEN due to energy recycles). The major modication in the HEN part of the owsheet was the addition of two bypass streams (BP-1 and BP-2) around the two rigorous heat exchangers (E-100 and E-103) which match the very cold combined vapors from the DT and VLS with the relatively hot incoming natural-gas feed stream (Fig. 3). We considered non-zero bypass fractions at the nominal (base-case) operating conditions. This gave the exibility to manipulate the bypass ows in either the decreasing or increasing directions as dependent on the direction of the disturbance (e.g., decrease or increase in the natural-gas feed temperature). However, the use of non-zero nominal bypass fractions required higher UA values for exchangers E-100 and E-103 to match the nominal values of the target temperatures of these exchangers with those corresponding to the nominal operating conditions.

5. The base-case solution summary Table 1 lists the major parameters used for the base-case solution. Table 2 shows the conditions and compositions of the natural-gas-feed, ethane-rich product, and methane-rich residue streams at the base-case solution.

7062

A.E.S. Konukman, U. Akman / Chemical Engineering Science 60 (2005) 7057 7074

Table 1 Major parameters used for the base-case solution TEE-1: TEE-2: BP-1: BP-2: Q-80: Q-81: Q-82: Q-83: E-100: E-103: DT Reboiler: Nominal split ratio (Stream 2 to Stream 1) Nominal split ratio (Stream 8 to Stream 7) Nominal bypass fraction (Stream 29 to Stream 6) Nominal bypass fraction (Stream 27 to Stream 22) Nominal heat duty Fixed heat duty Fixed heat duty Fixed heat duty UA UA Boil-up ratio (Boilup over Stream 21) 0.50 0.50 0.20 0.20 4.7322 106 kJ/h 1.5000 106 kJ/h 1.5000 106 kJ/h 5.0000 106 kJ/h 1.5 106 kJ/ C-h 1.0 106 kJ/ C-h 0.40

Table 2 The conditions and compositions of the natural-gas-feed, ethane-rich product, and methane-rich residue streams at the base-case solution Streams & conditions Natural gas feed stream (Stream 1) 0.9975 35 60 5000 0.00550 0.00910 0.84570 0.08200 0.03400 0.00580 0.00860 0.00280 0.00210 0.00180 0.00120 0.00050 0.00040 0.00050 Methane-rich residue (Stream 25) 1 40.8 22.1 4342.9 Mole fractions 0.00633 0.00598 0.96497 0.02195 0.00074 0.00002 0.00001 0.00000 0.00000 0.00000 0.00000 0.00000 0.00000 0.00000 Methane rejection = 99.11% Ethane-rich product (Stream 21) 0 7.7 19.1 657.1 0.00000 0.02973 0.05741 0.47886 0.25382 0.04403 0.06538 0.02130 0.01598 0.01370 0.00913 0.00381 0.00304 0.00381

Vapor fraction Temperature [ C] Pressure [atm] Molar ow [kgmol/h] Components Nitrogen CO2 Methane Ethane Propane i-Butane n-Butane i-Pentane n-Pentane n-Hexane n-Heptane n-Octane n-Nonane n-Decane Ethane recovery = 76.75%

The natural-gas feed is rich in methane (84.57%) and lean in ethane (8.20%). At the base-case operating conditions, the TEP is capable of concentrating the ethane in its product stream (47.89%) with a small ethane loss in the residue stream (2.20%). The product stream contains low methane (5.74%); all of the methane is basically concentrated in the residue stream (96.50%). The output streams of the plant are not balanced; the molar ow of the ethane-rich product stream (657.1 kgmol/h) is only 13.14% of the feed, and the molar ow of the methane-rich residue stream (4342.9 kgmol/h) is 86.86% of the feed. In Table 2, the ethane recovery (76.75%) is calculated as the ratio of the molar ow of ethane in Stream 21 (product) to the molar ow of ethane in Stream 1 (natural-gas feed), and the methane rejection (99.11%) is calculated as the ratio of the molar ow of methane in Stream 25 (residue) to the molar ow of methane in Stream 1 (natural-gas feed).

6. HEN synthesis Although we adopted the HEN/TEP owsheet from the application-case library of HYSYS, we carried out the synthesis of minimum-utility HEN structures as a check. For this task we applied the superstructure-based simultaneous optimization formulation which was proposed by Yee and Grossmann (1990a,b) and used extensively in the literature since then. The linear energy-balance constraints in the superstructure formulation, together with the linear minimum total utility objective function, make the synthesis problem convex MILP (mixed-integer linear programming) and thus guarantee the global solution. The formulation bases on a stagewise superstructure representation that considers potential exchanges between any pair of hot and cold streams within a superstructure stage via binary decision variables. Overall heat balances for each stream, heat

A.E.S. Konukman, U. Akman / Chemical Engineering Science 60 (2005) 7057 7074

7063
BP-2

balances at each stage of the superstructure, assignment of superstructure inlet temperatures, equations for the feasibility of temperatures, hot- and cold-utility load equations, logical constraints, and approach-temperature equations are the set of linear constraints in the formulation. The explicit MILP formulation, superstructure representation and other information can be found elsewhere (Yee and Grossmann, 1990a,b; Konukman et al., 2002). The objective of HEN synthesis was selected as the minimization of the total (hot and cold) utility consumption at the base-case operating conditions. In the TEP, there is one hot stream which is the natural-gas feed (Stream 1) and four cold streams that are the methane-rich residue gas (Stream 22 in Fig. 3), liquid streams drawn from stages 5 and 8 of the DT (Fig. 4), and the process-uid of the reboiler labeled as To Reboiler and Boilup in Fig. 4. Hypothetical hot and cold utilities were also specied. The source and target temperatures of these hot and cold streams were obtained from the base-case solution of the plant. The heat-capacity-owrates were obtained by using the heat duties and inputoutput temperatures of the exchangers in the base-case solution of the plant. It is known that this MILP model yields multiple HEN structures (determined from the optimal values of binary variables) depending on initial values of decision variables and MILP solver parameters. However, all of these alternative structures have a unique objective function value (minimum total utility) (Konukman et al., 2002). The MILP optimization formulation was solved with GAMS/XA (Brooke et al., 1992). All alternative structures obtained for the TEP showed no external heating requirement but only external cooling for minimum-approach-temperature values less than 7 C. The objective function was thus equal to the external cooling load of 5 106 kJ/h (Q-83 in Table 1). The HEN structure shown in Fig. 3 was among the alternative structures. However, the MILP model can only accommodate heaters and coolers at the terminals of the superstructure. Therefore, the place of the cooler (E-105) in the MILP solution was not on the split branch as shown in Fig. 3 but at the end; on Stream 13. However, by using the HEN retrot capability of AspenTech-HX-Net (v6.2, Build 5121), the cooler was moved back to the location shown in Fig. 3 and the total exchanger area (calculated with HX-Nets default parameters) of the HEN was compared. The total HEN area with E-105 at the end (MILP solution) was slightly greater than the total area with E-105 at its original location (Fig. 3). Therefore, it was concluded that the original HEN structure of the TEP (Fig. 3) was already optimal (maximum energy recovery network). The owsheet of the HEN in the form of grid diagram is shown in Fig. 5. It should be noted that what we did here is not a complete synthesis since the source- and target-stream temperatures of the HEN were taken from the base-case solution of the TEP and the locations of side reboilers other than 5th and 8th stages were not considered. In other words, terminal temperatures of the cold streams in Fig. 5 (COLD 14) were all

BP-1 Stream 24 Methane-rich product Stream 1 NGL FEED Cold Utility E-100 TEE-1 E-104 HOT COLD - 1

Stream 22 E-103 TEE-2 Stream 13 to LTS E-102

E-105

from tray 5 COLD - 2 to tray 6 from tray 8 COLD - 3 E-101 COLD - 4 Stream 21 to tray 9 from reboiler to reboiler Ethane-rich product

D e m e t h a n i z e r

Fig. 5. The owsheet of the HEN in the form of grid diagram.

dependent of the DT at the base-case solution. A rigorous synthesis for highly energy integrated complex plants like the TEP, without sacricing from industry-standard (process owsheet simulator based) rigorous thermodynamicand physical-property calculations, is quite difcult. A simultaneous approach requires an equation-oriented process simulator (e.g., ASCEND) where all the mass- and energy-balance equations along with phase-equilibria and thermophysical property equations of the TEP can be made available individually. These process equations may then be treated as equality constraints and may be combined with the equations of the HEN-synthesis formulation, leading to a huge (all internal TEP and HEN variables must be considered as decision variables) infeasible-path MINLP (mixedinteger nonlinear programming) model the (global) solution of which cannot be guaranteed. Such an infeasible-path optimization strategy theoretically can nd a better HEN for the TEP (less cooling, lower total area, lower total cost) perhaps with entirely different HEN-TEP interconnection structure. A sequential approach requires a sequential-modular process simulator (e.g., HYSYS) where the mass- and energy-balance equations along with phase-equilibria and thermophysical property equations of the TEP are not available individually but solved operation by operation following the direction of material ow. Since the process equations cannot be used as equality constraints and cannot be combined with the equations of the HEN-synthesis formulation, a feasible-path optimization strategy must be adopted. Such a feasible-path optimization strategy requires the sequential solution of the HEN-synthesis MI(N)LP problem (outer loop) and the solution of the TEP (inner loop). In each iteration, the side-reboiler locations along the DT (can be represented by integer decision variables)

7064

A.E.S. Konukman, U. Akman / Chemical Engineering Science 60 (2005) 7057 7074


0.09 0.07 0.05 0.03 0.01 25 35 45 55 Natural-Gas Feed Temperature, oC
X-methane Stream-21 X-ethane Stream-24 X-ethane Stream-21 X-methane Stream 24

may be altered and thus the source- and target-stream conditions of the HEN being synthesized are recomputed for the outer loop. If such sequential execution of the HENsynthesis model and the process owsheet simulator can be accomplished, feasible-path optimization strategy as well theoretically can nd a better HEN for the TEP perhaps with entirely different HENTEP interconnection structure. On the other hand, the current status of the commercial process-simulation and HEN-synthesis technologies cannot handle simultaneous process simulation and HEN synthesis automatically for complex plants like the TEP.

1.00 0.95 0.90 0.85 0.80 0.75 0.70 0.65 0.60 0.55 0.50 0.45

Mole Fraction

7. Effects of the feed-stream temperature and pressure The effects of the changing natural-gas-feed (Stream 1) temperature on ethane recovery, methane rejection, and on mole fractions of methane and ethane in the residue (Stream 24) and product (Stream 21) were investigated conveniently by the use of HYSYSs databook capability which automates the repetitive solution of the TEP for varying values of the feed-stream temperature. Fig. 6 shows the ethane recovery and methane rejection as a function of the feed-stream temperature (nominal value was 35 C as given in Table 2). The ethane recovery is inversely related to the feed-stream temperature; recovery increases with decreasing temperature. The methane rejection, however, is directly related to the feed-stream temperature. The scales of the graph indicate that the methane rejection is not very sensitive to the feed-stream temperature. Fig. 7 shows the mole fractions of ethane and methane in the residue and product streams as a function of the natural-gas feed-stream temperature. The mole fraction of methane in the product (Stream 21) decreases signicantly with increasing temperature. The mole fraction of ethane in the product (Stream 21) attains maximum around 35 C. Therefore, 35 C nominal feed temperature seems to be optimal both considering the ethane-recovery and methanerejection levels in the product stream. The methane mole fraction in the residue (Stream 24) is not very sensitive

Fig. 7. Mole fractions of ethane and methane in the residue and product streams as a function of the natural-gas feed-stream temperature.

Stream 13 Temperature, C

-43 -45 -47 -49 -51 -53 -55 25 30 35 40 45 50


o

55

60

Natural-Gas Feed Temperature, C


Fig. 8. Temperature of the HEN outlet (natural-gas stream entering the main TEP) as a function of the natural-gas feed-stream temperature.

82 80 78 76 74 72 70 68 66 64 62 60 25 30 35 40 45 50 55 60 Natural-Gas Feed Temperature, oC

100.0

99.5

99.0

98.5

Fig. 6. Ethane recovery and methane rejection as a function of the natural-gas feed-stream temperature.

to the feed temperature. However, the ethane mole fraction increases appreciably with increasing feed temperature; indicating the increasing loss of ethane in the residue. Fig. 8 shows the temperature of the HEN outlet (Stream 13), which is the chilled natural-gas stream entering the expander side of the TEP, as a function of the natural-gas feedstream temperature. It is seen that the temperature of Stream 13 depends on the feed-stream temperature almost linearly. This is an important observation since it implies that there is a one-to-one correspondence with the temperature of the outlet stream of the HEN and the ethane-recovery/methanerejection levels. In other words, if the temperature of the HEN outlet can be controlled at a desired level, then it is possible to control the ethane-recovery by inferring its level from the HEN outlet temperature. The temperature of the HEN outlet stream corresponding to the nominal natural-gas feed-stream temperature is 51.5 C. The effect of the changing natural-gas-feed (Stream 1) pressure on the ethane recovery, methane rejection, and on the mole fractions of methane and ethane in the residue (Stream 24) and product (Stream 21) were also investigated conveniently by the use of HYSYSs databook

Ethane Recovery, %

Methane Rejection, %

Mole Fraction

A.E.S. Konukman, U. Akman / Chemical Engineering Science 60 (2005) 7057 7074

7065

86 84 82 80 78 76 74 72 70 68 66 50 55 60 65 70 75 Natural-Gas Feed Pressure, atm 80

100.0 99.5 99.0 98.5 98.0 97.5

8. Effects of the HEN control variables


Methane Rejection, %

Ethane Recovery, %

Fig. 9. Ethane recovery and methane rejection as a function of the natural-gas feed-stream pressure.

0.15 0.13 Mole Fraction 0.11 0.09 0.07 0.05 0.03 0.01 50 55 60 65 70 75 Natural-Gas-Feed Pressure, atm 80
X-methane Stream-21 X-ethane Stream-24 X-ethane Stream-21 X-methane Stream-24

1.00 0.95 0.90 0.85 0.80 0.75 0.70 0.65 0.60 0.55 0.50 0.45 0.40 0.35

Fig. 10. Mole fractions of ethane and methane in the residue and product streams as a function of the natural-gas feed-stream pressure.

capability which automates the repetitive solution of the TEP for varying values of feed-stream pressure. Fig. 9 shows the ethane recovery and methane rejection as a function of the feed-stream pressure (nominal value was 60 atm as given in Table 2). The ethane recovery is directly related to the feed-stream pressure; recovery increases with increasing pressure. The methane rejection, however is inversely related to the feed-stream pressure. The scales of the graph indicate that the methane rejection is not as sensitive as the ethane recovery to the feed-stream pressure. Fig. 10 shows the mole fractions of ethane and methane in the residue and product streams as a function of the naturalgas feed-stream pressure. The mole fraction of methane in the product (Stream 21) increases with pressure. Furthermore, the mole fraction of ethane in the product attains maximum around 60 atm. Therefore, 60 atm nominal feedstream pressure seems to be optimal both considering the ethane-recovery and methane-rejection levels in the product stream. The methane mole fraction in the residue (Stream 24) is not very sensitive to the feed pressure. However, the ethane mole fraction decreases appreciably with increasing feed pressure; indicating the decreasing loss of ethane in the residue with increasing feed pressure.

Since we concentrate on the HEN of the TEP in this work, all control variables (manipulated variables) were selected within the HEN structure. The ve manipulated variables (Fig. 3) are TEE-1 split fraction (Stream 2 to Stream 1), TEE2 split fraction (Stream 8 to Stream 7), BP-1 bypass fraction (Stream 29 to Stream 6), BP-2 bypass fraction (Stream 27 to Stream 22), and the external cooling amount (Q-83). The effects of the changes in these manipulated variables on ethane recovery, methane rejection, and on mole fractions of methane and ethane in the residue (Stream 24) and product (Stream 21) were investigated conveniently by the use of HYSYSs databook capability which automates the repetitive solution of the TEP for varying values of the manipulated variables. TEE-1 and TEE-2 were varied in the range [0.2, 0.8], BP-1 and BP-2 were varied in the range [0, 0.4], and Q-83 was varied in the range [1 106 , 9 106 kJ/h]. To facilitate the presentation of the effects of these variables, their ranges were normalized between [0, 1] in the gures to follow. Figs. 1114 show the effects of the manipulated variables (TEE-1 and TEE-2 split fractions, and BP-1 and BP-2 bypass fractions) on Stream 13 temperature, ethane recovery, methane rejection, and on mole fractions of methane and ethane in the residue and product streams when the ranges of the manipulated variables are normalized. In these gures, the points of intersection of the curves correspond to the nominal values of the manipulated variables (Table 1). Fig. 11 shows that TEE-2 split has almost no effect on the HEN outlet temperature (and hence on the ethane recovery). TEE-1 split has a non-linear and signicant effect only in the normalized range 00.6. BP-1 bypass has an almost linear and signicant effect which is valid throughout the
-42 TEE-1 -44 Stream 13 Temperature,C BP-1 TEE-2 BP-2

Mole Fraction

-46

-48

-50

-52

-54 0.0 0.5 Normalized Range (TEE-1, TEE-2, BP-1, BP-2)


Fig. 11. Temperature of the HEN outlet as a function of the normalized range of the manipulated variables.

1.0

7066

A.E.S. Konukman, U. Akman / Chemical Engineering Science 60 (2005) 7057 7074

80 78

99.8 T EE-1 99.6 BP-1 TEE-2 BP-2

76 Methane Rejection, % TEE-1 BP-1 0.0 0.5 Normalized Range (TEE-1, TEE-2, BP-1, BP-2)
Fig. 12. Ethane recovery as a function of the normalized range of the manipulated variables. Fig. 14. Methane rejection as a function of the normalized range of the manipulated variables.
100 Methane Rejection, %

74 Ethane Recovery, % 72 70 68 66 64 62 1.0 TEE-2 BP-2

99.4

99.2

99.0

98.8

98.6 0.0 0.5 Normalized Range (TEE-1, TEE-2,BP-1, BP-2) 1.0

0.48
Ethane Recovery, %

84 80 76 72 68 64 60 56 1.E+06

99

Ethane Mole Fraction (Stream 21)

98

0.47

97

3.E+06

5.E+06

7.E+06

9.E+06

Q-83 Heat Flow, kJ/h


Fig. 15. Ethane recovery and methane rejection as a function of the cooling load (Q-83 of exchanger E-105).

TEE-1 BP-1 0.46 0.0 0.5 Normalized Range

TEE-2 BP-2 1.0

(TEE-1, TEE-2, BP-1, BP-2)


Fig. 13. Ethane mole fraction as a function of the normalized range of the manipulated variables.

entire range. The effect of BP-2 bypass is less than that of BP-1. Therefore, the control of the HEN outlet temperature should be done with TEE-1 split fraction for higher temperatures of Stream 13, and as the temperature decreases, for example towards 52 C, the control should be done with BP-1 bypass manipulations. Fig. 12 shows that TEE-2 split has almost no effect on the ethane recovery. TEE-1 split has a non-linear and signicant effect only in the normalized range 0.61. BP-1 bypass has a more linear and signicant effect which is valid throughout

the entire range. The effect of BP-2 bypass is less than that of BP-1. Therefore, the control of the ethane recovery should be done with TEE-1 split for lower ethane-recovery levels, and as the recovery increases, for example towards 75%, the control should be done with BP-1 bypass manipulations. Fig. 13 shows that TEE-2 split has almost no effect on the ethane mole fraction in the product stream. TEE-1 split has a very non-linear and signicant effect only in the normalized range 00.6. BP-1 bypass has a signicant effect especially in the normalized ranges 00.3 and 0.71. The effect of the BP-2 bypass is less than that of BP-1. Therefore, the control of the ethane mole fraction should be done with TEE-1 split for lower ethane-mole-fraction levels, and as the mole fraction increases, for example towards 0.5, the control should be done with BP-1 bypass manipulations. Fig. 14 shows that TEE-2 split has almost no effect on the methane rejection. TEE-1 split has a signicant effect only in the normalized range 00.6. BP-1 bypass has the

A.E.S. Konukman, U. Akman / Chemical Engineering Science 60 (2005) 7057 7074

7067

signicant and nearly linear effect in the entire range. The effect of BP-2 bypass is less than that of BP-1. Therefore, the control of the methane rejection should be done with TEE-1 split and/or BP-1 bypass manipulations. The only free service unit in the whole HEN is exchanger E-105. Therefore, the external cooling load (Q-83) may also be considered as the manipulated variable. Fig. 15 shows that Q-83 has more signicant effect on ethane recovery compared to other manipulated variables such as TEE-1 or BP-1.

provided a reasonable initial point is used and the number of primary variables is small. The operability (steady-state control) problem was posed as the minimization of the norm of the ethane-recovery (R) deviation from its desired (set) point (R set ) subject to manipulated-variable bounds under natural-gas feed-stream temperature and pressure disturbances (Eq. (1)).
X1 ,X2 ,X3 ,X4

min

R R set 0.20 X1 0.70, 0.00 X2 , X3 0.40, 1 106 X4 9 106 . (1)

s.t.

9. Operability The operability (steady-state control) problem was dened as follows: When there is a disturbance in the natural-gas feed-stream condition(s), keep the ethane recovery at its desired level by using the manipulated variables in the HEN of the TEP. The four manipulated variables were selected as the TEE1 split fraction (X1 ), BP-1 bypass fraction (X2 ), BP-2 bypass fraction (X3 ), and the external cooling amount Q-83 (X4 ). TEE-2 was not included in the manipulated variables since its effect on ethane recovery was concluded to be negligible compared to those of the others. The nominal values of these manipulated variables had been listed in Table 1. The desired level of ethane recovery, R, (ethane recovery set point) was selected as R set =76.75%, which is the recovery at the basecase operating conditions listed in Table 2. The NLP optimization problem was solved using the HYSYSs optimizer. Among the available optimization methods in HYSYS we used the Mixed Method that attempts to take advantage of the global convergence characteristics of the BOX method and the efciency of the SQP method. It starts the minimization with the BOX method using a very loose convergence tolerance and after convergence, the SQP (Sequential Quadratic Programming) method is used to locate the nal solution using the desired tolerance. The procedure employed in the BOX method is based on the downhill simplex algorithm and is a sequential search technique without derivative evaluations. Although it is not very efcient in terms of the required number of function evaluations, this method can be very robust. The SQP method handles inequality and equality constraints and is considered by many to be the most efcient method

Although Eq. (1) looks simple, the ethane recovery (R) is obtained through the complete and rigorous solution of the TEP (inner loop) at each iteration of the optimizer (outer loop), through a method termed as the feasible-path strategy. We considered ve disturbance scenarios: DS-1: +15 C upset in feed temperature (from 35 to 50 C); DS-2: +10 atm upset in feed pressure (from 60 to 70 atm); DS-3: 5 C upset in feed temperature (from 35 to 30 C) and +10 atm upset in feed pressure (from 60 to 70 atm); DS-4: 20% decrease in cooler load (from 5 106 to 4 106 kJ/h); and DS-5: +20% increase in cooler load (from 5 106 to 6 106 kJ/h). Table 3 lists the ve disturbance scenarios and the corresponding uncontrolled response (ethane-recovery and methane-rejection levels, and HEN outlet (Stream 13) temperature) of the TEP, without optimization. When the cooler load (Q-83) is not considered as a manipulated variable, Table 4 lists the optimal values of the rst three manipulated variables that keep the ethane-recovery level at its set point for each disturbance scenario. Table 5 lists the optimal values of the four manipulated variables, including the cooler load (Q-83), that keep the ethane-recovery level at its set point for each disturbance scenario. Simulation studies showed that TEE-1, BP-1, and BP-2 each alone (one manipulated variable) were not able to keep the ethane recovery at its desired point for some disturbance scenarios (e.g., DS-1). On the other hand the cooler load as the only manipulated variable was able to keep the ethane recovery at its desired point for all disturbance scenarios, however, for some scenarios, the cooling load was greater then its base-case value (Table 1). Thus, it can be concluded that the use of TEE-1, BP-1, and BP-2 accomplishes the

Table 3 Disturbance scenarios and the corresponding uncontrolled response of the TEP Disturbance scenario DS-1 DS-2 DS-3 DS-4 DS-5 % Ethane recovery 68.9 81.6 82.9 73.1 79.6 % Methane rejection 99.6 98.5 98.1 99.4 98.8 HEN outlet temperature, C 46.6 48.3 49.3 49.2 53.5

7068

A.E.S. Konukman, U. Akman / Chemical Engineering Science 60 (2005) 7057 7074

Table 4 Optimal values of the manipulated variables (not including cooler load) Disturbance scenario DS-1 DS-2 DS-3 DS-4 DS-5 TEE-1 split ratio 0.675 0.309 0.251 0.599 0.389 BP-1 bypass fraction 0.041 0.275 0.223 0.105 0.250 BP-2 bypass fraction 0.225 0.273 0.302 0.205 0.229

Table 5 Optimal values of the manipulated variables (including cooler load) Disturbance scenario DS-1 DS-2 DS-3 TEE-1 split ratio 0.533 0.514 0.525 BP-1 bypass fraction 0.154 0.244 0.253 BP-2 bypass fraction 0.168 0.229 0.235 Cooler load 106 6.405 3.407 2.748

ethane-recovery control task without extra cooling. The selection of these three manipulated variables in the HEN structure proved to be effective in keeping the ethane recovery at its set point under the disturbances studied. Under the base-case values of the natural-gas feed-stream conditions, the maximum ethane recovery of 82.4% (98.1% methane rejection) was obtained with TEE-1 at its upper bound (0.8) and BP-1 and BP-2 at their lower bounds (zero), without changing the cooler load from its base-case value. Keeping TEE-1 at its upper bound and BP-1 and BP-2 at their lower bounds, the maximum ethane recovery of 90.1% (88.6% methane rejection) was obtained with a very high cooler load of 15 106 kJ/h. 10. Flexibility Operation of a plant can be described by the set of following equality and inequality constraints: h(d, x, u, p, ) = 0, g(d, x, u, p, ) 0, (2) (3)

simulator indicates the satisfaction of Eq. (2). On the other hand, considering the simulation of the TEP with the process simulator, the inequality constraints (Eq. (3)) may be present only if the optimization of the simulator is invoked (e.g., the inequality constraints in Eq. (1) that were imposed on the manipulated variables in the operability analysis of the TEP). Flexibility (Swaney and Grossmann, 1985) of a plant represents the ability to accommodate variations of a set of uncertain parameters. The degree of exibility is determined by the range of parameter variations that the plant can tolerate. A scalar index of exibility can be constructed to measure the size of the feasible operating region in the space of uncertain parameters. Combination of the uncertain parameters lying inside the region permit adjustment of the process to achieve feasible operation. Evaluation of the exibility index (Swaney and Grossmann, 1985) corresponds to inscribing within region a scaled hyperrectangle which may be expressed in terms of the nonnegative scalar variable as ( ) = { |(

)}.

(4)

where d is the vector of design variables (e.g., heatexchanger areas, number of DT stages), x is the vector of state variables (e.g., internal stream temperatures, pressures, compositions in the TEP owsheet), u is the vector of control/free variables (e.g., manipulated variables in the HEN of the TEP), p is the vector of xed parameters (e.g., heattransfer coefcients, locations of side reboilers of the DT), and is the vector of uncertain parameters (e.g., natural-gas feed-stream temperature and pressure of the TEP). Considering the simulation of the TEP with a sequential-modular process owsheet simulator (e.g., HYSYS), the equality constraints (Eq. (2)) correspond to solution of mass- and energy-balance equations for every process module in the owsheet. A successful run (convergence) of the process

Eq. (4) describes the range of the uncertain parameters as deviations from nominal values, , where and + are the scaled deviations from , and the non-negative scalar, , is the exibility. The region ( ) describes the scaled hyperrectangle which is required to be within the feasible region dened by Eqs. (2)(3). The exibility index, , is the maximum of scalar that can be achieved under the conditions dened by Eqs. (2)(3). However, the maximization of subject to Eqs. (2)(4) is a semidenite-programming problem the solution of which is very difcult in general (Swaney and Grossmann, 1985) and impossible using optimization capabilities of commercial process simulators. The maximum hyperrectangle that can be expanded around the nominal values of the uncertain parameters touches the feasible region boundary at a vertex of the hyperrectangle. This

A.E.S. Konukman, U. Akman / Chemical Engineering Science 60 (2005) 7057 7074

7069

vertex can be regarded as a critical point for the feasible operation of the plant since it is the one that limits the maximum exibility. In general, the critical point need not correspond to a vertex of the hyperrectangle for nonconvex feasible regions. An MINLP algorithm (active constraint strategy) for nding nonvertex critical points is available (Grossmann and Floudas, 1987), however its implementation in sequentialmodular process simulators is impossible. When the critical point is a vertex of the hyperrectangle, the simplest approach that can be implemented with the optimization capabilities of commercial process simulators is the vertex enumeration method (Swaney and Grossmann, 1985). In this method the exibility is maximized in each vertex direction, v, belonging the set of all vertex directions, V, (v V) by the following optimization formulation:
v

= max v
u, v

(5a) (5b) (5c)

s.t. Eqs. (2) and (3) = + v.

Then, the exibility index, solutions of Eq. (5):

, is the smallest one among the (6)

= min
vV

variables in the HEN of the TEP. The manipulated variables (set of control variables u in Eq. (5a)) were the TEE-1 split fraction (X1 ), BP-1 bypass fraction (X2 ), BP-2 bypass fraction (X3 ), and the external cooling load Q-83 (X4 ). The desired level of ethane recovery, R, (ethane recovery set point) was R set = 76.75%. The exibility problem, solved using the HYSYSs optimizer (mixed method), was posed as the maximization of the disturbance magnitude, , subject to ethane-recovery-setpoint equality constraint and subject to manipulated-variable bounds. Two disturbance variables (uncertain plant parameters ) were considered; the natural-gas feed stream temperature ( 1 ) and pressure ( 2 ). The nominal values of the disturbance variables are those used for the base-case solution of the TEP, i.e., = 35 C and = 60 atm. The explicit 1 2 formulation is given by Eq. (7). Eq. (7b) symbolizes the feasible-path solution (owsheet convergence) of the TEP with HYSYS for each iteration (including derivative evaluations) of the HYSYS optimizer. Eq. (7c) denotes that the ethane recovery is computed after owsheet convergence as a function of the values of the manipulated variables, X14 , vertex values of uncertain parameters, 1 and 2 , and the scaled distance v along the vertex direction v at each iteration of the HYSYS optimizer.
v

The total number of vertex directions for N uncertain parameters is 2N (v = 1, . . . , 2N combinations of the directed deviations from the nominal values of the uncertain parameters, ). For example, for N = 2 uncertain parameters, the four vertex directions are v V{++, +, +, ). The values in Eqs. (2)(3) are governed by Eq. (5c) that constrains the vector to take only 2N vertex values. The maximization problem in Eq. (5) is solved for each vertex; 2N times, and the minimum of the solutions of Eq. (5) is taken as the exibility index, . Geometrically, solution of Eq. (6) corresponds to nding the direction v for which the distance to the boundary of the feasible operating region is shortest. This method thus requires 2N optimization runs of the process simulator, which may be prohibitive for large number of uncertain parameters, N. The approach named Algorithm I by Halemane and Grossmann (1983) enables replacement of 2N solutions of Eq. (5) with only one solution by repeating the set of constraints (Eq. (5b)) for each vertex (Konukman et al., 2002). However, this approach cannot be implemented in commercial process owsheet simulators since it implies simultaneous solution of different plants (simulation problems) corresponding to individual vertex values of the uncertain parameters. Therefore, the current status of process simulators necessitates the use of vertex enumeration method that require 2N process simulations with optimization. For the TEP plant, we dene the exibility (exibility index) problem as follows: The maximum (limiting) magnitude of disturbance in the natural-gas feed-stream condition(s) (temperature and pressure) such that the ethane recovery can be kept at its desired level via the manipulated

= maximize

v
v

X1 ,X2 ,X3 ,X4 ,

(7a) (7b) (7c) (7d) (7e) (7f) (7g) (7h)

s.t.

Eq. (2) (mass/energy balances),


v

R(X, ,
1 2

) R set = 0,
v v

= =

1 2

+ +

v, v,

0.20 X1

0.70, 0.40, 9 106 .

0.00 X2 , X3 1 106 X4

We rst solved the problem by considering individual temperature and pressure disturbances separately. Thus, e.g., when considering the temperature disturbance only, N = 1 and there are 2N = 2 vertex directions; + indicating feed temperature increase and indicating feed temperature decrease. Table 6 shows the result of the one-dimensional exibility analyses; i.e., the maximum and minimum values of the feed-stream temperatures and pressures that can be tolerated while keeping the ethane recovery at its set point via the manipulated variables in the HEN when the external cooling load (Q-83) was not included as a manipulated variable. In the increasing temperature direction, + , the TEP can tolerate feed temperatures as high as 59.1 C, whereas in the decreasing temperature direction, , the TEP can tolerate feed temperatures as low as 10.3 C (when the feed pressure is at its 60 atm nominal value). In the increasing pressure direction, the TEP can tolerate feed pressures as high as 83.9 atm, whereas in the decreasing pressure direction, the

7070

A.E.S. Konukman, U. Akman / Chemical Engineering Science 60 (2005) 7057 7074

Table 6 One-dimensional exibility of the TEP to feed temperature and pressure variations and the corresponding optimal values of control variables (excluding cooler load as manipulated variable)
Directional exibility, v

TEE-1 split ratio

BP-1 bypass fraction

BP-2 bypass fraction

v = + : v = 24.1 C = Maximum f eed temperature = 59.1 C v = : v = 24.7 C Minimum f eed temperature = 10.3 C v = + : v = 23.9 atm Maximum f eed pressure = 83.9 atm v = : v = 11.7 atm = Minimum f eed pressure = 48.3 atm

0.8 0.2

0.0 0.4

0.0 0.4

0.2 0.8

0.4 0.0

0.4 0.0

Table 7 One-dimensional exibility of the TEP to feed temperature and pressure variations and the corresponding optimal values of control variables (including cooler load as manipulated variable)
Directional exibility, v

TEE-1 split ratio

BP-1 bypass fraction

BP-2 bypass fraction

Cooler load 106

v = + : v = 69.4 C Maximum f eed temperature = 104.4 C v = : v = 42.4 C = Minimum f eed temperature = 7.4 C v = + : v = 50.2 atm Maximum f eed pressure = 110.2 atm v = : v = 27.4 atm = Minimum f eed pressure = 32.6 atm

0.8 0.2 0.2 0.8

0.0 0.4 0.4 0.0

0.0 0.4 0.4 0.0

9 1 1 9

Table 8 Two-dimensional exibility of the TEP to simultaneous feed temperature and pressure variations and the corresponding optimal values of control variables (excluding cooler load as manipulated variable) Vertex directions in T and P ++ + + Directional exibility, v
v v

T C 70.0 40.4 26.4 26.6

P atm 120.0 50.7 74.7 44.3

TEE-1 split ratio 0.2 0.8 0.2 0.8

BP-1 bypass fraction 0.19 0.0 0.4 0.0

BP-2 bypass fraction 0.23 0.0 0.4 0.0

= 100 = 15.5 = v = 24.5 v = 24.0

TEP can tolerate feed pressures as low as 48.3 atm (when the feed temperature is at its 35 C nominal value). According to the standard denition (Swaney and Grossmann, 1985), the exibility index for the feed-temperature uncertainty (Eq. (7e) not used), as computed from Eq. (6), is 24.1 C (minimum of + and ). The exibility index for the feed-pressure uncertainty (Eq. (7d) not used), as computed from Eq. (6), is 11.7 atm. Table 7 shows the result of the onedimensional exibility analyses when the external cooling load (Q-83) was included as a manipulated variable. In the increasing temperature direction, the TEP can tolerate feed temperatures as high as 104.4 C,whereas in the decreasing temperature direction, the TEP can tolerate feed temperatures as low as 7.3 C. In the increasing pressure direction, the TEP can tolerate feed pressures as high as 110.2 atm, whereas in the decreasing pressure direction, the TEP can tolerate feed pressures as low as 32.6 atm. According to the

standard denition (Swaney and Grossmann, 1985), the exibility index for the feed-temperature uncertainty (Eq. (7e) not used), as computed from Eq. (6), is 42.4 C (minimum of + and ). The exibility index for the feed-pressure uncertainty (Eq. (7d) not used), as computed from Eq. (6), is 27.4 atm. The use of cooler load as additional manipulated variable increases the exibility of the TEP. We next solved the problem by considering the temperature and pressure disturbances simultaneously. In this case N = 2 and there are 2N = 4 vertex directions; + indicating increase and indicating decrease in pairs of temperature T and pressure P, respectively {+T + P , +T P , T + P , T P ). In order to scale the temperature and pressure variables, we modied Eq. (7d) and (7e) as % deviations from the nominal values as i = (1 + v v/100) (for i i = 1, 2). Table 8 shows the result of the two-dimensional exibility analysis; i.e., the simultaneous directed varia-

A.E.S. Konukman, U. Akman / Chemical Engineering Science 60 (2005) 7057 7074

7071

% Change in Heat Capacity

tions in feed-stream temperature and pressure that can be tolerated while keeping the ethane recovery at its set point via the manipulated variables in the HEN when the external cooling load (Q-83) was not included as a manipulated variable. For example, in the direction, the TEP can tolerate feed temperatures and pressures jointly as low as 26.6 C and 44.3 atm, and in the + direction, the TEP can tolerate feed temperatures as low as 26.4 C and pressures as high as and 74.7 atm. According to the standard denition, the exibility index for the simultaneous feed-temperature and pressure uncertainties as computed from Eq. (6), is 15.5 (minimum of v values). This corresponds to maximum temperature of 40.4 C and minimum pressure of 50.7 atm.

40 35 30 25 20 15 10 5 0 -5 -10 -15 S1 S13 S23 S7 S2z

11. Additional comments on HEN synthesis In the literature, the synthesis, design, and exibility analysis of HENs are done as independent from the main process to which the HEN is attached. The analyses are mostly based on the xed (mean) values of stream properties (density, heat capacity, etc.) and heat-transfer coefcients. However, these properties are generally strong functions of temperature and pressure; especially for highly energy-integrated complex plants like the TEP where extreme low temperatures and high pressures involved result in strong non-ideal and nonlinear phase-equilibria and physical-property behavior. In Figs. 1618, we demonstrate the variations in heat capacities of Streams 1, 7, 13, 22, and 23 in the HEN of the TEP as a function of the natural-gas feed temperature and pressure (external disturbance sources), and TEE-1 split

-20 -25 50 55 60 65 70 75 80 Natural-Gas Feed Pressure, atm


Fig. 17. Deviation in heat capacities of some HEN streams with feed pressure.

2 0 -2 -4 -6 -8 -10 S7 -12 S22 -14 0.2 0.4 0.6 TEE-1 Split Fraction 0.8 S13 S23

8 6 4 % Change in Heat Capacity 2 0 -2 -4 -6 -8 S23 -10 25 30 35 40 45 50 Natural-Gas Feed Temperature, C


Fig. 16. Deviation in heat capacities of some HEN streams with feed temperature.

Fig. 18. Deviation in heat capacities of some HEN streams with TEE-1 split ratio.

S1 S13

S7 S22

ratio (internal disturbance source due to manipulated variable changes). Signicant variations in heat capacities (as high as 40%) are observed around the nominal values (intersection of the curves) of the x-axis variables. Clearly, any HEN synthesis, design, and exibility analysis based on the nominal values of parameters would yield erroneous results for complex plants like the TEP. These gures demonstrate the need to carry out the HEN synthesis and HEN integration tasks simultaneously with rigorous plant simulations using

% Change in Heat Capacity

7072

A.E.S. Konukman, U. Akman / Chemical Engineering Science 60 (2005) 7057 7074 Aguilera, N., Nasini, G., 1995. Flexibility test for heat exchanger networks with uncertain owrates. Computers and Chemical Engineering 19 (9), 10071017. Akman, U., Uygun, K., 1999. Nonlinear model predictive control of retrot heat-exchanger networks. Computers and Chemical Engineering 23S, S231S234. Akman, U., Uygun, K., Uzturk, D., Konukman, A.E.S., 2002. HEN optimizations without using logarithmic-mean-temperature difference. A.I.Ch.E. Journal 48 (3), 596606. Bandoni, J.A., Eliceche, A.M., Mabe, G., Brignole, E., 1989. Synthesis and optimization of ethane recovery process. Computers and Chemical Engineering 13 (45), 587594. Bansal, V., Perkins, J.D., Pistikopoulos, E.N., 2000. Flexibility analysis and design of linear systems by parametric programming. A.I.Ch.E. Journal 46, 335354. Bjork, K.M., Westerlund, T., 2002. Global optimization of heat exchanger network synthesis problems with and without the isothermal mixing assumption. Computers and Chemical Engineering 26 (11), 15811593. Briones, V., Kokossis, A.C., 1999. Hypertargets: a conceptual programming approach for the optimisation of industrial heat exchanger networksPart III. Industrial applications. Chemical Engineering Science 54 (5), 685706. Brooke, A., Kendrick, D., Mearaus, A., 1992. GAMS: A Users Guide. Scientic Press, Palo Alto, CA, USA. www.gams.com. Castier, M., Queiroz, E.M., 2002. Energy targeting in heat exchanger network synthesis using rigorous physical property calculations. Industrial and Engineering Chemistry Research 41 (6), 15111515. Chebbi, R., Al-Qaydi, A.S., Al-Amery, A.O., Al-Zaabi, N.S., AlMansouri, H.A., 2004. Simulation study compares ethane recovery in turboexpander processes. Oil and Gas Journal 102 (4), 6467. Chen, C.L., Hung, P.S., 2004. Simultaneous synthesis of exible heatexchange networks with uncertain source-stream temperatures and ow rates. Industrial and Engineering Chemistry Research 43 (18), 59165928. Chen, C.L., Hung, P.S., 2005. Multicriteria synthesis of exible heat-exchanger networks with uncertain source-stream temperatures. Chemical Engineering and Processing 44 (1), 89100. Cheung, K.Y., Hui, C.W., 2001. Heat exchanger network optimization with discontinuous exchanger cost function. Applied Thermal Engineering 21 (1314), 13971405. Chiu, C.H., 1997. LPG-recovery processes for baseload LNG plant examined. Oil and Gas Journal 95 (47), 5963. Diaz, M.S., Serrani, A., de Beistegui, R., Brignole, E.A., 1995. A MINLP strategy for the debottlenecking problem in an ethane extraction plant. Computers and Chemical Engineering 19 (S1), 175178. Diaz, M.S., Serrani, A., Brignole, E.A., 1997. Automatic design and optimization of natural gas plants. Industrial and Engineering Chemistry Research 36 (7), 27152724. Dimitriadis, V., Pistikopoulos, E.N., 1995. Flexibility analysis of dynamic systems. Industrial and Engineering Chemistry Research 34 (12), 44514462. Dolan, W.B., Cummings, P.T., Le Van, M.D., 1990. Algorithmic efciency of simulated annealing for heat-exchanger network design. Computers and Chemical Engineering 14 (10), 10391050. Eliceche, A.M., Sanchez, M., Fernandez, L., 1998. Feasible operating region of natural gas plants under feed perturbations. Computers and Chemical Engineering 22 (S1), 879882. Fernandez, L., Bandoni, J.A., Eliceche, A.M., Brignole, E.A., 1991. Optimization of ethane extraction plants from natural gas containing carbon dioxide. Gas Separation and Purication 5 (4), 229234. Finn, A.J., Tomlinson, T.R., Johnson, G.L., 2000. Design, equipment changes make possible high C3 recovery. Oil and Gas Journal 98 (1), 3744. Floudas, C.A., 1995. Nonlinear and Mixed-Integer Optimization: Fundamentals and Applications. Oxford University Press, New York, USA.

process owsheet simulators. However, the current technology seems to be far from providing rigorous simulation environments to tackle such tasks.

12. Conclusions A natural-gas turbo-expander plant (TEP) operating under ethane-recovery mode is investigated with emphasis on its heat-exchanger network (HEN) that strongly interacts with the plant. In the analyses, the process owsheet simulator HYSYS was used along with its optimization capability. Bypass streams around heat exchangers were added to increase exibility and disturbance-rejection ability of the HEN and thus the TEP. Together with the bypasses, the split fractions in the HEN and external cooling load were used as the manipulated variables in investigating the operability and exibility of the TEP. The emphasis was given to the assessment of exibility of the process in maintaining desired level of ethane recovery under natural-gas feed-stream temperature and pressure disturbances. The heat-exchanger bypasses and split fraction as the manipulated variables were able to control the ethane recovery at its set point for wide range of natural-gas feed-stream disturbances. It was shown that, for the TEP, any HEN synthesis, design, and exibility analysis based on the nominal values of physical properties would yield erroneous results due to signicant variations in heat capacities around the nominal operating conditions. For the rst time in the literature, the operability and exibility index computations were carried out with rigorous computations of the HEN and plant simultaneously using a commercial process owsheet simulator and its optimization capability. For highly energy-integrated complex plants like the TEP, the difculties or impossibilities of automated HEN synthesis and exibility analysis with process owsheet simulators are discussed.

Acknowledgements Financial supports provided by Gebze Institute of Technology Research Foundation (Project No: 2005A11), Bogazici University Research Foundation (Project No: 01A503 and 03A502), and TUBITAK (The Scientic and Technical Research Council of Turkey) (Project No: MISAG-83) are gratefully acknowledged. The authors thank the anonymous reviewers for their comments, which led to improvements over the rst version of this paper. References
Abbass, H.A., Wiggins, G.A., Lakshmanan, R., Morton, W., 1999. Heat exchanger network retrot via constraint logic programming. Computers and Chemical Engineering 23S, S127S130.

A.E.S. Konukman, U. Akman / Chemical Engineering Science 60 (2005) 7057 7074 Floudas, C.A., Grossmann, I.E., 1986. Synthesis of exible heat exchanger networks for multiperiod operation. Computers and Chemical Engineering 10 (2), 153168. Frausto-Hernandez, S., Rico-Ramirez, V., Jimenez-Gutierrez, A., Hernndez-Castro, S., 2003. MINLP synthesis of heat exchanger networks considering pressure drop effects. Computers and Chemical Engineering 27 (89), 11431152. Furman, K.C., Sahinidis, N.V., 2001. Computational complexity of heat exchanger network synthesis. Computers and Chemical Engineering 25 (910), 13711390. Furman, K.C., Sahinidis, N.V., 2002. A critical review and annotated bibliography for heat exchanger network synthesis in the 20th century. Industrial and Engineering Chemistry Research 41 (10), 23352370. Furman, K.C., Sahinidis, N.V., 2004. Approximation algorithms for the minimum number of matches problem in heat exchanger network synthesis. Industrial and Engineering Chemistry Research 43 (14), 35543565. Glemmestad, B., Skogestad, S., Gundersen, T., 1997. On-line optimization and choice of optimization variables for control of heat exchanger networks. Computers and Chemical Engineering 21 (S), 379384. Grossmann, I.E., Floudas, C.A., 1987. Active constraint strategy for exibility analysis in chemical processes. Computers and Chemical Engineering 11 (6), 675693. Grossmann, I.E., Caballero, J.C., Yeomans, H., 1999. Mathematical programming approaches to the synthesis of chemical process systemsreview. Korean Journal of Chemical Engineering 16 (4), 407426. Gundersen, T., Naess, L., 1988. The synthesis of cost optimal heat exchanger networks: an industrial review of the state of the art. Computers and Chemical Engineering 12 (6), 503530. Gundersen, T., Sagli, B., Kiste, K., 1991. Problems in sequential and simultaneous strategies for heat exchanger network synthesis. In: Puigjaner, L., Espuna, A. (Eds.), Computer-Oriented Process Engineering. Elsevier, Amsterdam, pp. 105116. Halemane, K.P., Grossmann, I.E., 1983. Optimal process design under uncertainty. A.I.Ch.E. Journal 29 (3), 425433. Heggs, P.J., Vizcaino, F., 2002. A rigorous model for evaluation of disturbance propagation through heat exchanger networks. Chemical Engineering Research and Design 80 (A3). HX-Net v6.2, Hyprotech, Aspen Technology, Inc. (www.aspentech.com), 2004. Conceptual Engineering Product v6.2, Heat Integration Guide. HYSYS v3.2, Hyprotech, Aspen Technology, Inc. (www.aspentech.com), 2003. User Guide. Karafyllis, I., Kokossis, A., 2002. On a new measure for the integration of process design and control: the disturbance resiliency index. Chemical Engineering Science 57 (5), 873886. Konukman, A.E.S., Camurdan, M.C., Akman, U., 2002. Simultaneous exibility targeting and synthesis of minimum-utility heat-exchanger networks with superstructure-based MILP formulation. Chemical Engineering and Processing 41 (6), 501518. Lee, R.J., Yao, J., Elliot, D.G., 1999. Flexibility, efciency to characterize gas-processing technologies. Oil and Gas Journal 97 (50), 9094. Li, S.J., Yao, P.J., 2001. Synthesis of heat exchanger network considering multipass exchangers. Chinese Journal of Chemical Engineering 9 (3), 242246. Li, Z.H., 2002a. Modeling and optimization for heat exchanger networks synthesis based on expert system and genetic algorithm. Chinese Journal of Chemical Engineering 10 (3), 290297. Li, Z.H., 2002b. Method for incorporation of controllability in heat exchanger network synthesis by integrating mathematical programming and knowledge engineering. Chinese Journal of Chemical Engineering 10 (6), 711716. Li, Z.H., Luo, X., Hua, B., Roetzel, W., 2004. Synthesis of exible heat exchanger networks with stream splits based on rangers of stream supply temperatures and heat capacity owrates. Chinese Journal of Chemical Engineering 12 (2), 240246.

7073

Lin, B., Miller, D.C., 2004. Solving heat exchanger network synthesis problems with tabu search. Computers and Chemical Engineering 28 (8), 14511464. Linnhoff, B., Flower, J.R., 1978a. Synthesis of heat exchanger networks: I. Systematic generation of energy optimal networks. A.I.Ch.E. Journal 24 (4), 633642. Linnhoff, B., Flower, J.R., 1978b. Synthesis of heat exchanger networks: II. Evolutionary generation of networks with various criteria of optimality. A.I.Ch.E. Journal 24 (4), 642654. Mehra, Y.R., Gaskin, T.K., 1999. Guidelines offered for choosing cryogenics or absorption for gas processing. Oil and Gas Journal 97 (9), 6267. Mehta, R.K.C., Devalkar, S.K., Narasimhan, S., 2001. An optimization approach for evolutionary synthesis of heat exchanger networks. Chemical Engineering Research and Design 79 (A2), 143150. Mikkelsen, J., Qvale, B., 2001. A combinatorial method for the automatic generation of multiple near-optimal heat exchanger networks. Chemical Engineering Research & Design 79 (A6), 663672. Mizutani, F.T., Pessoa, F.L.P., Queiroz, E.M., Hauan, S., Grossmann, I.E., 2003. Mathematical programming model for heat-exchanger network synthesis including detailed heat-exchanger designs. 2. Network synthesis. Industrial and Engineering Chemistry Research 42 (17), 40194027. Nagy, A.B., Adonyi, R., Halasz, L., Friedler, F., Fan, L.T., 2001. Integrated synthesis of process and heat exchanger networks: algorithmic approach. Applied Thermal Engineering 21 (1314), 14071427. Nishida, N., Liu, Y.A., Lapidus, L., 1977. Studies in chemical process design and synthesis: III. A simple and practical approach to the optimal synthesis of heat exchanger networks. A.I.Ch.E. Journal 23 (1), 7793. Oliveira, S.G., Liporace, F.S., Araujo, O.Q.F., Queiroz, E.M., 2001. The importance of control considerations for heat exchanger network synthesis: a case study. Brazilian Journal of Chemical Engineering 18 (2), 195210. Ozkan, S., Dincer, S., 2001. Application for pinch design of heat exchanger networks by use of a computer code employing an improved problem algorithm table. Energy Conversion and Management 42 (18), 20432051. Papalexandri, K.P., Pistikopoulos, E.N., 1993. An MINLP retrot approach for improving the exibility of heat exchanger networks. Annals of Operations Research 42 (14), 119168. Pettersson, F., 2005. Synthesis of large-scale heat exchanger networks using a sequential match reduction approach. Computers and Chemical Engineering 29 (5), 9931007. Pintaric, Z.N., Kravanja, Z., 2004. A strategy for MINLP synthesis of exible and operable processes. Computers and Chemical Engineering 28 (67), 11051119. Ravagnani, M.A.S.S., Silva, A.P., Arroyo, P.A., Constantino, A.A., 2005. Heat exchanger network synthesis and optimisation using genetic algorithm. Applied Thermal Engineering 25 (7), 10031017. Serna, M., Jimenez, A., 2004. An area targeting algorithm for the synthesis of heat exchanger networks. Chemical Engineering Science 59 (12), 25172520. Serna-Gonzalez, M., Ponce-Ortega, J.M., Jimenez-Gutierrez, A., 2004. Two-level optimization algorithm for heat exchanger networks including pressure drop considerations. Industrial and Engineering Chemistry Research 43 (21), 67666773. Shivakumar, K., Narasimhan, S., 2002. A robust and efcient NLP formulation using graph theoretic principles for synthesis of heat exchanger networks. Computers and Chemical Engineering 26 (11), 15171532. Sorsak, A., Kravanja, Z., 2002. Simultaneous MINLP synthesis of heat exchanger networks comprising different exchanger types. Computers and Chemical Engineering 26 (45), 599615. Swaney, R., Grossmann, I.E., 1985. An index for operational exibility in chemical process design. Part-I: formulation and theory. A.I.Ch.E. Journal 31 (4), 621630.

7074

A.E.S. Konukman, U. Akman / Chemical Engineering Science 60 (2005) 7057 7074 Wilkinson, J.D., Hudson, H.M., 1982. Turboexpander plant designs can provide high ethane recovery without inlet CO2 removal. Oil and Gas Journal 80 (18), 281285. Wilkinson, J.D., Hudson, H.M., 1992. Improved NGL recovery designs maximize operating exibility and product recoveries. Proceedings of the 71st Annual Meeting of the Gas Processors Association, Anaheim, CA, March 18. Wilkinson, J.D., Hudson, H.M., 1993. Improving gas processing prots with retrot designs for better ethane rejection/recovery. Presented at the Permian Basin Regional Meeting of the Gas Processors Association, May 13, 1993, Midland, TX, USA. Yee, T.F., Grossmann, I.E., 1990a. Simultaneous optimization models for heat integrationI. Area and energy targeting and modeling of multistream exchangers. Computers and Chemical Engineering 14 (10), 11511164. Yee, T.F., Grossmann, I.E., 1990b. Simultaneous optimization models for heat integrationII. Heat exchanger network synthesis. Computers and Chemical Engineering 14 (10), 11651184.

Tantimuratha, L., Kokossis, A.C., 2004. Flexible energy management and heat exchanger network design. Annals of Operations Research 132 (1), 277300. Tantimuratha, L., Asteris, G., Antonopoulos, D.K., Kokossis, A.C., 2001. A conceptual programming approach for the design of exible HENs. Computers and Chemical Engineering 25 (46), 887892. Uzturk, D., Konukman, A.E.S., Boyaci, C., Akman, U., 1996. Operability of an autothermal reactor linked to a exible heat-exchanger network. Computers and Chemical Engineering 20 (S), S943S948. Varvarezos, D.K., Biegler, L.T., Grossmann, I.E., 1995. Modeling uncertainty and analyzing bottleneck characteristics in multiperiod design optimization. Computers and Chemical Engineering 19 (5), 497511. Wang, W.B., 1985. Optimization of expander plants. Ph.D. Thesis, University of Tulsa, Tulsa, OK, USA. Wei, G.F., Yao, P.J., Luo, X., Roetzel, W., 2004. Study on multi-stream heat exchanger network synthesis with parallel genetic/simulated annealing algorithm. Chinese Journal of Chemical Engineering 12 (1), 6677.

Das könnte Ihnen auch gefallen