Material stream network modeling, retrofit and optimization for raw natural gas refining systems

Journal of Cleaner Production 142 (2017) 3419e3436

Contents lists available at ScienceDirect

Journal of Cleaner Production journal homepage: www.elsevier.com/locate/jclepro

Material stream network modeling, retrofit and optimization for raw natural gas refining systems Bing J. Zhang a, *, Tao Xu a, Chang He a, Qing L. Chen a, Xiang L. Luo b a

School of Chemical Engineering and Technology, Guangdong Engineering Technology Research Center for Petrochemical Energy Conservation, Sun Yat-Sen University, No. 135, Xingang West Road, Guangzhou, 510275, China b School of Materials and Energy, Guangdong University of Technology, No. 100, Waihuan West Road, Guangzhou Higher Education Mega Center, Guangzhou, 510006, China

a r t i c l e i n f o

a b s t r a c t

Article history: Received 31 July 2016 Received in revised form 21 October 2016 Accepted 22 October 2016 Available online 24 October 2016

The demand for natural gas is increasing in the energy market because of its lower emissions and sustainable development. This increasing demand for natural gas promotes the capacity expansion of raw natural gas refining systems (RNGRSs), resulting in parallel refining processes in a RNGRS. Optimizing the material stream network between these refining processes is very challenging because of the complex thermodynamics, unit operations and utility configurations. An optimization framework is presented for the retrofit of the material stream network between these refining processes to improve the economic performance. The retrofit framework integrates raw natural gas supply, refining processes, utility subsystems and product delivery and is formulated as a mixed-integer nonlinear programming (MINLP) optimization model to obtain an optimal material stream network to increase profit. The model presented here is applied to a Chinese industrial RNGRS and results in an optimal retrofit. A comparison before and after the retrofit demonstrates a significant increase in profit. Crown Copyright © 2016 Published by Elsevier Ltd on behalf of The Royal College of Radiologists. All rights reserved.

Keywords: Raw natural gas Retrofit Optimization Mixed-integer nonlinear programming Process modeling Utility system

1. Introduction Oil, coal and natural gas are the major energy sources that are used to meet the enormous demand for energy before breakthroughs in new energy techniques that will practically replace fossil fuels and reduce greenhouse gas emissions (Demierre et al., 2015). Natural gas is the cleanest of the fossil energy sources and has a considerable reserve. Global natural gas consumption doubled from 1985 to 2014, especially in Asia, as shown in Fig. 1. (International Energy Statistics, 2015). For example, China, facing the challenges of environmental protection and greenhouse gas reduction, will be forced to increase the use of natural gas from 4% to 10% before 2020 (Wang and Lin, 2017). Hence, many RNGRSs have been built and expanded to increase the production of natural gas. A RNGRS simultaneously receives raw natural gas from several exploring fields or wells. The raw natural gas streams have different pressures and chemical compositions for each field or well. The

* Corresponding author. E-mail address: [email protected] (B.J. Zhang).

RNGPR processes the different raw natural gas streams to produce products that cover the various specifications of down-stream customers. The refining processes in a RNGRS include plug catchers, dew point controllers, CO2 treating processes, compressors, blending pipelines, deethanizers, stabilizers and utility subsystems. Continuous capacity expansion leads to parallel refining processes in a RNGRS. Optimization of the material stream network between these refining processes may significantly increase the profitability, reduce the production costs and improve the energy performance. Nevertheless, RNGRSs involve complex material stream networks, thermodynamics and unit operations. The optimization retrofit of such a system is very challenging. In this study, how to optimize and retrofit such a complicated material stream network from raw natural gas supply to refining processes, to utility subsystems, and further to product delivery is addressed with the goal of obtaining better economic performance. 2. Literature review Process systems involve material, energy, water and hydrogen networks, etc. Many publications have investigated the modeling and optimization of these networks (Yong et al., 2016), such as

http://dx.doi.org/10.1016/j.jclepro.2016.10.124 0959-6526/Crown Copyright © 2016 Published by Elsevier Ltd on behalf of The Royal College of Radiologists. All rights reserved.

3420

B.J. Zhang et al. / Journal of Cleaner Production 142 (2017) 3419e3436

Fig. 1. World natural gas consumption by region, 1980e2013.

industrial water network retrofit (Sueviriyapan et al., 2016) and material stream network analysis (Lambrecht and Thißen, 2015). This study proposes the retrofit of material stream network in RNGRSs. The optimal retrofit of integrated chemical process complexes is an important topic that has generated great interest and has been investigated by several studies (Banimostafa et al., 2015). Such work is mainly motivated to process and capacity expansion (Wei et al., 2012), energy efficiency improvements (Kleme
s and Kravanja, 2013), emission reduction for power plants (Pan et al., 2016) and for eco-industrial networks (Kantor et al., 2015). Capacity expansion for the process industry often involves the retrofit and optimization of reactors (Neveu et al., 2013), distillation columns (Long et al., 2015) and their sequence (Caballero and Grossmann, 2014). The retrofit procedures for the integrated chemical processes reported in the literature can be outlined as (1) evaluating the performance of existing processes and equipment to determine bottlenecks; (2) identifying effective debottlenecking measures and initial alternatives to remove the bottlenecks; and (3) establishing a retrofit alternative based on an economic analysis and optimization of the process parameters. The optimization and retrofit of natural gas production systems is reviewed in this study. Diaz et al. (1997) proposed the turboexpansion process model and formulated a MINLP model to optimize the design and debottlenecking of natural gas processing plants. Selot et al. (2008) studied the model of transportation pipelines and compressors and combined them into a short-term operational planning model to optimize a natural gas production system. Further work on the uncertainty of natural gas production networks was performed by Li et al. (2011). Tabkhi et al. (2010) investigated the model of pipeline and compressor for long distance transportation of gas and used these models to minimize fuel consumption in compressor stations. Flores-Salazar et al. (2011) presented a multiperiod mathematical model for gas and oil production systems on the basis of two conceptual process units (well and manifold). The total production cost is simplified as a result of a cost coefficient multiplied by the capacity of the well. Enríquez et al. (2011) presented a retrofit design approach based on

process simulation and Response Surface Methodology (RSM) and applied it to the natural gas liquids recovery process. Zhang et al. (2012) presented a framework for the optimization and retrofit of lu an energy system in the process industry. Üster and Dilaverog (2014) introduced new designs and retrofits for the transmission networks of natural gas to minimize the total investment and operating costs. Kostowski et al. (2015) investigated the possible solutions to improve the thermodynamic performance of a natural gas compressor station equipped with various types of compressors and operated under partial-load conditions. Neseli et al. (2015) investigated the energy recovery from natural gas pressurereducing stations for electricity generation. Arredondo-Ramírez et al. (2016) presented a mathematical programming approach based on disjunctive programming to account for the complex logical relationships in the optimal planning of shale gas exploitation. de Queiroz Fernandes Araújo et al. (2016) proposed and compared chemical absorption and membrane permeation for CO2 separation from high CO2 content natural gas. Recently, Zheng et al. (2010) reviewed the mathematical model and optimization for the operation of natural gas production, gas recovery maximization, pipeline networks, compressor stations, least gas purchase, minimum fuel consumption, the gas market and contract problems, indicating that it is important to develop accurate process models and algorithms for such complex problems. Natural gas production systems include several processes to separate and transport gas streams. The models of these processes have a significant effect on the optimization of the whole system. Ríos-Mercado and BorrazS anchez (2015) also reviewed the optimization of natural gas transportation. The abovementioned works focused on the design and operation of systems for the exploitation, transportation and compression of raw natural gas and less on refining systems. To the best of our knowledge, the retrofit of RNGRSs has seldom been investigated, although RNGRSs have a significant effect on the production cost and utility consumption of natural gas. In this study, a retrofit framework is presented to optimize the material stream network between these refining processes to improve the economic

B.J. Zhang et al. / Journal of Cleaner Production 142 (2017) 3419e3436

performance. The retrofit framework integrates the raw natural gas supply, refining processes, utility subsystems and products distribution and is formulated as a MINLP optimization model to obtain an optimal material stream network that links processes for better profit. The model presented here is applied to a Chinese industrial RNGRS and results in an optimal retrofit network. A comparison before and after the retrofit demonstrates a significant increase in profit. 3. Material stream network of RNGRS 3.1. Description of RNGRS In general, a RNGRS has four parts, as shown in Fig. 2: (1) several raw natural gas transportation terminals; (2) a few parallel naturalgas refining processes; (3) a utility sub-system that supplies the essential energy for the refining processes and consumes fuels from the refining processes; and (4) several customers that receive different products from the system. Raw natural gas is introduced from transport terminals that link exploring fields or wells. A RNGRS generally connects several transport terminals. The costs, chemical compositions and pressures of raw natural gas streams are different for each transport terminal. The refining processes in a RNGRS include plug catchers, dew point controllers, CO2 treating processes, compressors, blending

3421

pipelines, deethanizers, stabilizers and utility sub-systems. Different raw natural gas streams are purified, separated, compressed and blended through these processes. The connections between these refining processes are shown in Fig. 2. Continuous capacity expansion leads to a RNGRS that has a few of the parallel refining processes. These parallel refining processes are often scheduled and operated independently in real industrial operations. This appears to be reasonable, but fails to make use of the advantages of the material stream blending and process network. A utility sub-system is configured to meet the energy requirement of refining processes. Boilers and turbines are installed to convert the chemical energy of fuel into power and heat. Fuel comes from the separators and deethanizers. A portion of the gas from the plug catchers can also be used as boiler fuel if the gas from the separators and deethanizers is not enough for the utility subsystem. There are two pressure levels in a RNGRS. At one level, high-pressure steam generated from boilers is introduced into back-pressure turbines. At the other level, low-pressure steam from the back pressure turbines is used as a heat resource for the CO2 treating processes, deethanizers and stabilizers. Excess lowpressure steam is then transported into condensing turbines to produce power. A RNGRS has three products: sale gas, LPG and naphtha. LPG and naphtha can be considered to be minor products of a RNGRSs. Sale gas products have different specifications according to the requirements of down-stream customers. For example, power plants

Fig. 2. Material stream network of a RNGRS.

3422

B.J. Zhang et al. / Journal of Cleaner Production 142 (2017) 3419e3436

require a low CO2 content in the sale gas product. Methanol plants require an appropriate CO2 content in the sale gas product because CO2 is needed for chemical reactions. In civil use, there is a need to control the content of heavy hydrocarbons in the sale gas product. 3.2. Representation of the material stream network for RNGRSs The material stream network in a RNGRS shown in Fig. 2 includes two parts: stream links and process nodes. We use set XCh,s,h0 to denote the existing material stream links of a RNGRS. An element in set XCh,s,h0 shows that a stream s exits process h and enters process h0 . We use another set XGh,s,h0 to denote all of the potential links between two processes in the RNGRS, including existing links. As a result, set XCh,s,h0 is a subset of XGh,s,h0 . We then define set XDh,s,h0 as follows:

XDh;s;h0 ¼ XGh;s;h0  XCh;s;h0

(1)

Hence, set XDh,s,h0 contains links that can be candidate connections for the retrofit of the material stream network. The connections that can be selected for the retrofit must be optimized from the entire system. The definition of all of the symbols used for the sets, parameters and variables can be found in Appendix A. The refining process nodes in the material stream network are able to change the compositions and pressures of the material streams. To generally represent the mass balance and pressure constraints for a real process, a virtual model, as shown on the right of Fig. 3, is presented for a real process. We virtualize the real process into three parts: mixing, separating and splitting. Note that raw natural gas transport terminals, as source nodes of the network, are virtualized as the splitting parts; customers, as end nodes of the network, is virtualized as the mixing parts. The mixing part mixes different material streams and exports only one stream. The mixing part is able to receive streams from different processes according to the network structure. For example, a plug catcher can receive different raw natural gas streams from transportation terminals. The total material and mass balances for individual chemical components in the mixing part are formulated as Eqs. (B1)e(B5), which are listed in Appendix B. There are many constraints for the entire optimization model, so we put them in a special appendix. The separating part can separate the stream from the mixing part into several streams that have different chemical compositions and flow rates. For example, plug catchers separate raw natural gas into a light gas stream and a heavy liquid stream. Meanwhile, a pressure change can be performed in the separating part, such as in

the compressors. The total material balance and mass balances for individual chemical components in the separating part are formulated as Eqs. (B6)e(B12). A real process can have several splitting parts. A splitting part splits a stream from the separating part into several branches that enter different processes. The splitting parts do not change the chemical compositions of the streams, but can lead to branches with different flow rates. For example, the gas stream from dew point controllers can be split into three branches that enter CO2 treating processes, compressors or blending pipelines. The total material balance and mass balances for individual chemical components in the splitting part are formulated as Eqs. (B13)e(B16). The pressure in the material stream network is a significant factor that affects the energy efficiency and product quality in the RNGRS. The pressure of a process must not be greater than that of any stream entering it. In contrast, the pressure of a process is not related to a stream that does not enter it. We model the pressure change from raw natural gas to products by using Eqs. (B17)e(B23). Temperature is another parameter that is just as significant as pressure in the material stream network. It is used at dew point controllers to control the dew point or the contents of heavy hydrocarbons in the gas streams. The temperature at dew point controllers is described in Eqs. (B24)e(B25). The pressure and temperature changes in a process are correlated with the energy requirement in the detailed process models. Hence, Eqs. (B1)e(B25) represent the mass, pressure and temperature constraints for the material stream network of the RNGRS. Detailed process models are proposed in the next section to correlate the mass balance of individual chemical components, as well as the pressure and temperature changes, and utility consumption.

4. Detailed process models The RNGRS shown in Fig. 2 includes two parts: the material stream network and refining processes. The model of the material stream network is presented in the preceding section. Detailed models of the refining processes can be extracted from the fundamentals of thermodynamics and unit operations. Eqs. (B26)e(B54) listed in Tables B1eB8 at Appendix B represent the mass balance of individual chemical components and the utility consumption of refining processes. Plug catchers and separators in the RNGRS are used to separate fluid into gas and liquid streams. The vapor-liquid equilibrium governs the chemical compositions of the gas and liquid streams

Fig. 3. Virtualizing model for refining processes.

B.J. Zhang et al. / Journal of Cleaner Production 142 (2017) 3419e3436

from the plug catchers and separators. For natural gas and the light hydrocarbons system, the correlation method presented by Wilson (1968) can be used, and Eqs. (B26)e(B30) can be used to calculate the chemical compositions in the two processes. Dew point controllers have a behavior similar to that of the plug catchers and separators. Hence, the mass balance of individual chemical components can also be represented using Eqs. (B26)e(B30). Furthermore, dew point controllers consume cold utilities to condense heavy hydrocarbons to meet the specification of sale gas products. The energy requirement is expressed by Eqs. (B31)e(B32). CO2 treating processes have been reviewed (Boot-Handford et al., 2014) and investigated to screen the efficient solvents (Zarogiannis et al., 2016). CO2 treating processes include an absorber and a regenerator and consume hot utilities to remove CO2 from gas streams. The mass balances of the individual chemical components for CO2 treating processes are represented in Eqs. (B33)e(B35). In the RNGRS, the absorber is operated at a high pressure and the regenerator in a low pressure. The pump transports the lean solvent from the regenerator to the absorber and consumes power. The power consumption is calculating using Eq. (B36). The heat required by the regenerator can be expressed in three parts: the heat output from the condenser, heat for the desorption reaction and heat for increasing the temperature of the solvent. These three parts are represented by Eq. (B37). Dehydration is used to remove the very small amount of water in the process streams from the treating processes. Solvents, molecular sieves and silicon are often used to remove trace water. Because of energy cost, silicone is currently widely used. Silicone is re-installed after a certain period. Thus, the dehydration process contributes a fixed cost to the whole system, and there is almost no utility consumption. Compressors are installed in order for the pressures of the sale gas products to satisfy the requirements of customers. Two ideal compression processes, isothermal and isentropic compression, are often used to model practical compression processes (Selot et al., 2008). Isentropic compression is used to represent the compressors and is formulated in Eqs. (B38)e(B39). The isentropic exponent qh of light hydrocarbon fluid was given by Moshfeghian (2013) and is calculated using Eq. (B40). Deethanizers and stabilizers recover the liquid components in raw natural gas. The bottom stream from the deethanizer must control the content of C2 to guarantee the quality of LPG. The purification of both the distillate and bottom streams from the stabilizer must be controlled to satisfy the specifications for LPG and naphtha products. Eqs. (B41) and (B42) express the purification of distillate and bottom streams for the deethanizer and stabilizer. Hence, the heat duties of the reboilers in the two columns are represented by Eqs. (B43) and (B44) (Branan, 2005). The boilers and turbines in the utility subsystems were modeled in our previous publication (Zhang et al., 2015). The constraints of the boilers are described by Eqs. (B45)e(B47), including the energy balances of the fuels and boilers. The correlations between power generation and the steam consumption of turbines are expressed using Eqs. (B48)e(B49). Finally, the balances of power, fuel, highpressure steam and low-pressure steam in the entire RNGRS are represented by Eqs. (B50)e(B53).

3423

profit before (Profitbefore) and after retrofit (Profitafter) and the pipeline cost (Costpipelines) raised by the material stream network retrofit are all accounted for. Depreciation using discount rate is considered for the fixed cost of new pipelines.


OBJ ¼ Profitafter  Profitbefore  Costpipelines

o ð1 þ oÞuseful life  1 (2)

5.2. Maximum profit before retrofit To solve the retrofit problem, the optimal profit before retrofit must be previously obtained for the existing RNGRS. The profit before retrofit is equal to the revenue from products (VP) minus the cost of raw natural gas (VR), the cost of utility consumption (VU) and other fixed costs (p). This is expressed in Eq. (3).

Profitbefore ¼ VP  ðVR þ VU þ pÞ

(3)

The products in a RNGRS include sale gas, LPG and naphtha. The revenue from these products is denoted by Eq. (4). The parameter zh is the price of the product from process h. The variable Fh is the flow rate of the product from process h. The flow rates of products are calculated from the process models.

VP ¼

X

zh Fh h2HNh

(4)

h

The cost of all raw natural gas streams is obtained using Eq. (5). A RNGRS is able to produce power by its own utility subsystem. Power can be uploaded or downloaded from the grid after it balances in the entire system. Hence, the variable VB in Eq. (6) can be negative or positive and is expressed in the utility subsystem model listed in Appendix B.

VR ¼

X

zh Fh h2HRh

(5)

h pow

VU ¼ z

VB

(6)

As a result, we formulate a MINLP model to determine the maximum profit before retrofit. The model to maximize profit before retrofit is denoted as the HP problem and is presented

(HP) max s.t.

Profitbefore Eqs. (3)e(6) (economic constraints); Eqs. (B1)e(B25) (stream network constraints); Eqs. (B26)e(B29) (thermodynamic constraints); Eqs. (B30)e(B45) (process constraints); Eqs. (B46)e(B53) (utility constraints).

below. Note that we must use set XCh,s,h0 instead of XGh,s,h0 in Eqs. (B1)e(B53) when we solve the HP problem to obtain the maximum profit for the existing RNGRS. This is because XCh,s,h0 represents the existing material stream network of a RNGRS.

5. Retrofit MINLP model 5.3. Retrofit cost 5.1. Objective The objective of the model is to obtain the maximum increase in profit by retrofitting the material stream network in the RNGRS. The objective value (OBJ) is calculated using Eq. (2). The maximum

pez The cost of pipeline investment has been investigated by Ye (2008). The correlated function of the cost has been presented by Sanaye and Mahmoudimehr, 2013 for the design and optimization of natural gas systems. Eq. (7) is listed to express the pipeline cost

3424

B.J. Zhang et al. / Journal of Cleaner Production 142 (2017) 3419e3436

for the retrofit of the material stream network in a RNGRS. The specific pipeline investment includes the cost of pipes, installment and construction. In the first term on the right side of Eq. (7), the cost is independent of the capacity. In the second term, however, the cost is shown to be proportional to the capacity. The final term indicates that the cost is proportional to the pipe diameter. The parameters bh,s,h0 , sh,s,h0 , jh,s,h0 can be extracted from the specific RNGRS.

CSh;s;h0 ¼ bh;s;h0 þ sh;s;h0 Fp;s;p0  0:5 þ jh;s;h0 Fh;s;h0 ðh; s; h0Þ2XDh;s;h0

(7)

In the retrofit, we do not need to install all of the candidate links in set XDh,s,h0 for the material stream network. Here, a link means a stream s exiting process h and then entering process h0 . Hence, we introduce the binary variables Xh,s,h0 and the element (h,s,h0 ) 2 XDh,s,h0 to express whether a link is selected to be installed or not. When it is equal to 0, the corresponding link is not selected to retrofit. When it is equal to 1 and the element (h,s,h0 ) 2 XDh,s,h0 , the corresponding stream link is determined to be installed. Therefore, we use Eqs. (8)e(10) to replace Eq. (7) to denote the pipeline investment of the retrofit.

CTp;s;p0 ¼

X

h Lp;s;p0 bh;s;h0 Xh;s;h0 þ sh;s;h0 Fp;s;p0

p;s;p0

 0:5 i þ jh;s;h0 Fh;s;h0

max Xh;s;h0 Fp;s;p0 ¼ Fp;s;p0

Costpipelines ¼

X

ðh; s; h0Þ2XDh;s;h0

ðh; s; h0Þ2XDh;s;h0

CTp;s;p0

ðh; s; h0Þ2XDh;s;h0

(8)

variables. As a result, the mathematical models only involve bilinear and power items through the above two steps. The global optimization solver ANTIGONE (Misener and Floudas, 2014) in GAMS 24.5.3 is then used to solve the reformulated models. ANTIGONE firstly generates and solves convex relaxations of the nonconvex MINLP that rigorously bound the global solution, then finds feasible solutions via local optimization, and finally divides and conquers the feasible set to generate a sequence of convex relaxations converging to the global optimum. An industrial example is investigated and the solution performance is demonstrated in Section 7. 6. Retrofit strategy The strategy of material stream network retrofit for a RNGRS is summarized in Fig. 4. Five steps are extracted for the material stream network retrofit. The first step, shown in Fig. 4, is to obtain the basic data of RNGRSs, including the raw natural gas supply, demand for products, and price of raw natural gas and products. The second step is to extract the existing and potential material stream networks. At the same time, the models of the refining processes are presented according to the fundamentals of thermodynamics and unit operations. The third step is to obtain the maximum profit before retrofit. This is formulated as problem HP. The reason for calculating the maximum profit before retrofit is to optimize the profit increased by the material stream network

(9) (10)

h;s;h0

5.4. Optimization model for retrofit The profit after retrofit is the same as the expression of the profit before retrofit, as indicated by Eq. (10).

Profitafter ¼ VP  ðVR þ VU þ pÞ

(11)

When the HP problem is solved, we can use profitbefore as a parameter in Eq. (2). Considering the cost of pipelines, the optimization model for the retrofit is then formulated as the MP problem, which is presented below. Note that for the MP problem, the element (h,s,h0 ) in Eqs. (B1)e(B53) belongs to XGh,s,h0 , not to

(MP) max s.t.

obj (profit margin minus pipeline cost) Eqs. (2), (4)e(6) (economic constraints); Eqs. (8)e(11) (pipeline cost constraints); Eqs. (B1)e(B25) (stream network constraints); Eqs. (B26)e(B29) (thermodynamic constraints); Eqs. (B30)e(B45) (process constraints); Eqs. (B46)e(B53) (utility constraints).

XCh,s,h0 . Problems HP and MP are non-convex MINLP mathematical models. The non-convex constraints include Eqs. (7) and (8), B26 and B39, except for bilinear constraints. Two modeling efforts are done to solve the on-convex models. First, we supply the tight bounds for each variables according to the process operation rules and constraints. Second, factorization via introducing new variables is applied to the constraints that involve multiples of multi-

Fig. 4. Strategy of the material stream network retrofit.

B.J. Zhang et al. / Journal of Cleaner Production 142 (2017) 3419e3436

retrofit. The fourth step is to formulate the optimization model MP for the material stream network retrofit. The increase in profit from the retrofit can be maximized by solving the optimization model MP. The final step is to compare the existing material stream network to the optimal network solved in the fourth step, by which the retrofit scheme can be checked. The retrofit scheme can be obtained from the solution result of the binary variable Xh,s,h0 . 7. Examples 7.1. Example description A raw natural gas refining plant from South China is taken as an example and is investigated in this study. The plant receives raw natural gas from two transportation terminals and produces sale gas products for three down-stream customers. The three downstream customers have different specifications for the pressures and chemical compositions of the sale gas products. The raw natural gas refining plant has some parallel processes that have been built for capacity expansion. The data for the raw natural gas and products and the parameters are listed in Tables C1 and C2 in Appendix C. The processes, streams and existing stream network are listed in Tables C3eC5. To protect the competitive edge of this enterprise, some market and process data from the plant have been slightly adjusted. We use n-alkanes to represent the chemical components in the raw natural gas. A complete list of chemical components can also be used for a plant if detailed monitoring data can be applied. The utility subsystem includes three boilers, two back-pressure turbines and two condensing turbines. 7.2. Solution results and discussion The MP and HP models are coded in the GAMS 24.5.3 environment (Brooke et al., 2015) for this example. The mathematical models are executed on an Intel(R) Core(TM) i5-4670 3.40 GHz PC. The global optimization solver ANTIGONE is employed to solve the models. There are 2729 single variables, 136 single binary variables and 3254 single constraints in model HP. There are 3423 single variables, 199 single binary variables and 4137 single constraints in model MP. The solution times of the two models are all less than 12 s. The solution results for the economic data are listed in Table 1. All of the solution results can be found in Tables C6eC12. By comparing Tables C9 and C10, we can observe that two new pipelines need to be installed. One is a pipeline for the gas stream from the third dew point controller to the first CO2 treating process, and the other is a pipeline for the gas stream from the third dew point controller to the second CO2 treating process. Table 1 Solution performance for example. Items

Retrofit after

Retrofit before

Profit [$/h] Pipeline cost [$/h] Cost of raw natural gas [$105/h] Raw natural gas #1 [kmol/h] Raw natural gas #2 [kmol/h] Income from products [$105/h] Sale gas #1 [kmol/h] Sale gas #2 [kmol/h] Sale gas #3 [kmol/h] LPG [kmol/h] Naphtha [kmol/h] CO2 removed [kmol/h] Gas fired [kmol/h] Utility cost [$/h] Power purchased [MW]

133,822 1167 4.248 72,000 45,000 6.3633 30000 54220 0 8861 7373 4534 12012 21709 135.68

121,220 NA 4.248 72,000 45,000 6.2086 24986 29743 32000 7010 7245 9643 6373 18836 117.73

3425

The objective value on the left side of Eq. (2) equals $ 94,622,863. The optimal profit of the raw natural gas refining plant is $121,220 per hour before retrofit, while it is up to $133,822 per hour after retrofit. The cost for new pipeline installation is $1167 per hour. The profit has a significant increase of 9.4% after retrofit. The throughputs of raw natural gas before and after retrofit are all on the upper bounds. This indicates that the raw natural gas refining plant favors installing the maximum capacity for raw natural gas to increase profit. Although the throughput before and after retrofit is the same, the product distribution and utility requirements are significantly different. The total flow rate of all gas products after retrofit is 84,220 kmol/h, while that before retrofit is 86,729 kmol/h. The raw natural gas refining plant does not supply the sale gas #3 product after retrofit. Sale gas #3 has a restrictive limitation on its CO2 content, middle pressure specification and middle price among the three gas products, which can be found in Table C1. Hence, producing sale gas #3 requires the removal of more CO2, resulting in an additional energy requirement. The CO2 removed before retrofit is 9643 kmol/h, and that after retrofit is 4534 kmol/h. The additional energy requirement for CO2 removal does not produce a marginal profit when CO2 is an effective component in the sale gas products, which can be observed in Table 1. More of product sale gas #2 is produced after retrofit. The flow rate of sale gas #2 before retrofit is 29,743 kmol/h, and that after retrofit is 54,220 kmol/h. Sale gas #2 has restrictive specifications on pressure and on the content of heavy hydrocarbons. The price of Sale gas #2 is the highest among the three gas products. Increasing production of sale gas #2 means separating additional heavy hydrocarbons from the raw natural gas. This leads to an energy requirement in the dew point controllers. However, the additional recovery of heavy hydrocarbons can produce the highest valueadded LPG product. Hence, the flow rate of the LPG product after retrofit is 8861 kmol/h, and that before retrofit is 7010 kmol/h. The flow rate of the LPG product is increased by 26.4%. By separating heavy hydrocarbons, the energy requirements in the dew point controllers, deethanizers and stabilizers are increased. As a result, the raw natural gas plant has to download more power from the grid and consume more fuel, as indicated in Table 1. 8. Conclusions RNGRSs include complex material stream networks and refining processes that closely interact in materials and energy. Two MINLP models are presented for the operational optimization and retrofit of RNGRSs. The models integrate material stream networks, thermodynamics, refining processes and utility subsystems. An effective strategy was extracted to guide the material stream network retrofit for RNGRSs. Three conclusions can be drawn. First, the presented MINLP retrofit model and strategy are effective for the optimization and material stream network retrofit of RNGRSs. Second, an optimal material stream network can significantly increase the profitability of RNGRSs. The example given demonstrates a profit increase of 9.4%. Last, the optimal material stream network favors producing the highest pressure stream with the highest price among the sale gas products and avoiding the removal of CO2 from raw natural gas. Acknowledgments This research is supported by the National Natural Science Foundation of China (No. 21376277 and 21276288), and the project of Guangdong Provincial Natural Science Foundation of China (No. 2015A030313112).

3426

B.J. Zhang et al. / Journal of Cleaner Production 142 (2017) 3419e3436

Appendix A. Nomenclature

XHh,c XLh,c

A.1. Superscripts * critical condition ^ shift variable D amount of change boi boiler bot bottom of column col cold utility hot hot utility ise isentropic enthalpy change low lower bound ohs overheat high-pressure steam out outlet of process pum pump reb reboiler rer refrigeration reu reflux of distillation shs saturated high-pressure steam sls saturated low-pressure steam sol solvent ste steam top distillate of column up upper bound vap vapor wat water

A.2. Sets/Indices B/b set of boilers indexed by b C/c set of chemical components indexed by c H/h set of processes indexed by h S/s set of process streams indexed by s U/u set of steam turbines indexed by u HA subset of set H, representing separators HC subset of set H, representing compressors HD subset of set H, representing dew point controllers HE subset of set H, representing deethanizers HL subset of set H, representing stabilizers HN subset of set H, representing end processes HP subset of set H, representing plug catchers HR subset of set H, representing raw natural gas terminals HS subset of set H, representing customers HT subset of set H, representing CO2 treating processes HU subset of set H, representing virtual utility subsystems SH subset of set S, representing heavy streams SL subset of set S, representing light streams UB subset of set T, representing back pressure steam turbines UC subset of set T, representing condensing steam turbines XCh,s,h0 set of pairs, representing existing connections of stream s from process h to h0 XDh,s,h0 set of pairs, representing candidate retrofit connections of stream s from process h to h0 XGh,s,h0 set of pairs, representing all available connections of stream s from process h to h0 XEh,s set of pairs, representing stream s from process h

XRh,c

set of pairs, representing heavy component c for stabilizers set of pairs, representing light component c for deethanizers and stabilizers set of pairs, representing specifications of customer on components

A.3. Greeks a relative volatility for deethanizers and stabilizers [dimensionless] b, s, j parameter used in process models [dimensionless] g mole volume [m3/kmol] d big data [dimensionless] ε specific heat capacity [MJ/kmol/ C] z price [$/kmol; $/MW/h] h efficiency [dimensionless] q isentropic exponent of compressor [dimensionless] i reflux ratio of CO2 regenerator [dimensionless] k absorption ability of solvent [dimensionless] l blowdown rate of boiler [dimensionless] m maximum temperature change [ C] n bounds [dimensionless] x fixed power requirement [MW] ε latent heat/isentropic enthalpy change/low calorific value/desorption heat [MJ/kmol] o discount rate [dimensionless] p other cost [$/h] r pressure [MPa] t temperature [ C] y specific composition of streams [dimensionless] 4 related to isentropic exponent [dimensionless] c maximum pressure change [MPa] u eccentric factor [dimensionless] A.4. Real variables VB balance of power [MW] OBJ objective [$] VP product income [$] VR raw natural gas cost [$] VU utility cost [$] A.5. Nonnegative variables E power [MW] F flow rate of process stream [kmol/h] G flow rate of steam [kmol/h] K vapor liquid phase equilibrium ratio [dimensionless] M mole fraction [dimensionless] P pressure [MPa] Q heat [MW] T temperature [ C] U minimum reflux ratio for stabilizer and regenerator [dimensionless] A.6. Binary variable X 0,1variable [dimensionless]

B.J. Zhang et al. / Journal of Cleaner Production 142 (2017) 3419e3436

3427

Appendix B. Network, thermodynamics, processes and utility subsystem models

Table B1 Material stream network constraints for RNGRSs. Description

Equation (number)

Mass balance of individual chemical component c for process h Fh;c ¼ Fh Mh;c

cc; ch

(B1)

Handling capacity constraints of process h up nlow h Xh  Fh  nh Xp ch

Summation of component fractions in process h

X

Mh;c ¼ Xh

(B2)

cc; h;HRh

(B3)

c

The constraints of component fractions for raw natural gas Mh;c ¼ yh;c

cc; h2HRh

(B4)

Mh;c  yh;c

h2HSh ; ðh; cÞ2XRh;c

(B5)

The constraints of component fractions for products

Logical relationship of process and stream operations Xh;s;h0  Xh0 Mixing: mass balance constraints for streams and handling capacity

X

ðh; s; h0 Þ2XGh;s;h0

Fh;s;h0 ¼ Fh0

h;HRh ; ðh; s; h0Þ2XGh;s;h0

(B6)

(B7)

h

Mixing: mass balance of individual chemical components

X  Mh;s;h0;c Fp;s;p0 ¼ Fh0;c

cc; h;HRh ; ðh; s; h0Þ2XGh;s;h0

(B8)

h;s

Mixing: mass balance for connecting streams

X

Mh;s;h0;c ¼ Xh;s;h0

cc; h;HRh ; ðh; s; h0Þ2XGh;s;h0

(B9)

c

Mixing: logical constraints for streams Fh;s;h0  d$Xh;s;h0 Separating: mass balance for handling capacity Fh ¼

X

h;HRh ; ðh; s; h0Þ2XGh;s;h0

(B10)

h;HNh ; ðh; sÞ2XEh;s

(B11)

h;HNh ; ðh; sÞ2XEh;s

(B12)

Fh;s

s

Separating: logical constraints for handling capacity Fh;s  d$Xh Separating: mass balance for individual chemical components Fh;c ¼

X  Mh;s;c Fh;s

cc; h;HNh ; ðh; sÞ2XEh;s

(B13)

s

Separating: logical constraints for component fractions

X

Mh;s;c ¼ Xh

h;HNh ;

ðh; sÞ2XEp;s

(B14)

h;HNh ;

ðh; s; h0Þ2XGh;s;h0

(B15)

c

Splitting: mass balance of streams Fh;s ¼

X

Fh;s;h0

h0

Splitting: mass balance for individual chemical components Mh;s;c ¼ Mh;s;h0;c

h;HNh ;

ðh; s; h0Þ2XGh;s;h0

(B16)

3428

B.J. Zhang et al. / Journal of Cleaner Production 142 (2017) 3419e3436 Table B2 Pressure and temperature constraints for RNGRSs. Description

Equation (number)

Pressure constraints for raw natural gas Ph  rh

h2HRh

(B17)

rh  Ph h2HNh

(B18)

Pressure constraints for products

Pressure relationship between processes and streams


 Ph0  Ph;s;h0 þ d$ 1  Xh;s;h0

h0;HRh ;

ðh; s; h0Þ2XGh;s;h0

(B19)

Pressure relationship in processes Phout ¼ Ph þ PhD

h;HNh ; h0;HRh

(B20)

Pressure change in processes PhD  cmax h

h;HNh ; h0;HRh

(B21)

Ph;s  Phout

h;HNh ;

(B22)

Pressure constraints for the outlet of a process ðh; sÞ2XEh;s

Pressure constraints for streams Ph;s;h0  Ph;s

h;HNh ;

ðh; s; h0Þ2XGh;s;h0

(B23)

Temperature relationships in processes D Th ¼ t h  Th

(B24)

ThD  mmax h

(B25)

Temperature changes in processes

Table B3 Models of thermodynamics, plug catchers, separators and dew point controllers. Description

Equation (number)

Vapor-liquid equilibrium ratio (Wilson, 1968)

! Kh;c ¼

5:37ð1þuc Þ Pc*

Ph

e

T*

1Tc

h

cc; h2ðHAh ∪HDh ∪HPh Þ

(B26)

Transformation for vapor-liquid equilibrium ratioa K^h;c  d$Xh Transformation for vapor-liquid equilibrium ratio

cc; h2ðHAh ∪HDh ∪HPh Þ

(B27)

a

K^h;c  Kh;c

cc; h2ðHAh ∪HDh ∪HPh Þ

(B28)

Transformation for vapor-liquid equilibrium ratioa Kh;c K^h;c  d$ð1  Xh Þ  K^h;c

cc; h2ðHAh ∪HDh ∪HPh Þ

(B29)

Phase equilibrium calculation for plug catchers, separators and dew point controllers Mh;s;c ¼ K^h;c Mh;s0;c cc; s2SLs ; s02SHs ; h2ðHAh ∪HDh ∪HPh Þ; ðh; sÞ2XEh;s ; ðh; s0Þ2XEh;s

(B30)

a Vapor-liquid equilibrium ratio Kh,c is never equal to zero according to Eq. (B26). Nevertheless, vapor-liquid equilibrium in plug catchers, dew point controllers and separators only exists if the operating capacity of a process does not equal zero. Hence, Eqs. (B27)e(B29) are used provide an alternative vapor-liquid equilibrium ratio K^p,c for the process operations.

Table B4 Energy requirement of dew point controllers. Description

Equation (number)

Cold utility for dew point controllers Th

# " X X  X  εc Fh;c þ εc Mh;s;c ¼ Qhcold Fh;s c

s

c

(B31)

cc; h2HDh ; s2SHs ; ðh; sÞ2XEh;s Power requirement

. cold Eh ¼ hrer 3600 h Qh

h2HDh

(B32)

B.J. Zhang et al. / Journal of Cleaner Production 142 (2017) 3419e3436

3429

Table B5 Model for CO2 treating processes. Description

Equation (number)

Mass balance for individual chemical components except for CO2 Fh;c ¼

X

Mh;s;c Fh;s



csco2 ; h2HTh ; s2SLs ; ðh; sÞ2XEh;s

(B33)

s

Mass balance for CO2 Mh;s;c ¼ Xh

c ¼ CO2 ; h2HTh ; s2SHs ; ðh; sÞ2XEh;s

(B34)

Solvent flow rate Fhsol ¼ kh Fh;s

s2SHs ; h2HTh ; ðh; sÞ2XEh;s

(B35)

Power consumption Eh ¼

rDh $Fhsol $g

h2HTh

pum

3600hh

Hot duty for regenerators Qhreb ¼ εreu h ih

X s

Fh;s þ εrea h

(B36)

X s

Fh;s þ εsol ThD Fhsol

(B37)

ch2HTh ; s2SLs ; ðh; sÞ2XEh;s

Table B6 Compressors. Description

Equation (number)

Compression ratio Rh ¼

Phout Ph

cc; h2HCh

Power requirement of isentropic compression (Mokhatab et al., 2015) Ehcom

¼

1

hcom h

X

3 2 qh 1 6 7 εc Fh;c 4ðRh Þ qh  15

(B38)

cc; h2HCh

(B39)

c

isentropic exponent of light hydrocarbon gas

qh ¼ 1:3  0:31ð4h  0:55Þ h2HCh

(B40)

Table B7 Deethanizers and Stabilizers. Description

Equation (number)

Purification constraints for distillate streams

X c

Purification constraints for bottom streams

P c

top

Mh;s;c  yh Xh

c2XHh;c ; h2HLh ; s2SLs ; ðh; sÞ2XEh;s

Mh;s;c  ybot h Xh

c2XLh;c ; h2ðHLh ∪HEh Þ; s2SHs ; ðh; sÞ2XEh;s Minimum reflux ratio Uh þ Xh ¼

X ah;c Mh;s;c c ah;c  bh

(B41)

(B42)

(B43)

h2ðHLh ∪HEh Þ; s2SLs ; ðh; sÞ2XEh;s Reboiler duty Qhreb ¼ sh εreu h Uh

X s

Fh;s þ jh Xh

h2ðHLh ∪HEh Þ; s2SLs ; ðh; sÞ2XEh;s

(B44)

3430

B.J. Zhang et al. / Journal of Cleaner Production 142 (2017) 3419e3436

Table B8 Utility models. Description

Equation (number)

Heat balances for boilers

h

ih









sb nup Xb þ ð1 þ bb ÞGb ð1 þ lb Þεwat tshs  twat þ εshs þ εste tohs  tshs b

Energy balances of fuels Q boi b ¼ Fb

X  εc Ch;c

cb; c; h2HUph

i

¼ Q boi b

cb

(B45)

(B46)

h;c

Hardware capacity constraints for boilers up nlow b Xb  Gb  nb Xb cb

The correlations of power export and steam consumption of turbines 3600$Eu ¼

6 1 5 bu

 εise u 

(B47)

su nup u



1 Xu Gu  nup 6 u

cu

(B48)

Hardware capacity constraints for turbines up nlow u Xu  Eu  nu Xu cu

Fuel balance in the entire RGNGRS

X

X

Fb ¼

b

High-pressure steam balance in the entire RGNGRS

cb; h2ðHAh ∪HEh ∪HPh Þ

(B50)

h

X

Gu 

X

u

Low-pressure steam balance in the entire RGNGRS

Fh

(B49)

cb; u2UBu

Gb

(B51)

b

X Qhreb h

εsls

þ

X

Gu0 þ Gexp 

u0

X

Gu

u

(B52)

p2ðHTh ∪HEh ∪HLh Þ; u2UBu ; u02UCu Power balance in the entire RGNGRS

xþ

X h

Eh  VB þ

X

Eu

cb; cu; h2ðHDh ∪HTh ∪HCh Þ

(B53)

u

Appendix C. Data and solution results for example

Table C1 Raw natural gas and products for example. Items

RNG 1#

RNG 2#

Customer Customer Customer LPG 1# 2# 3#

Naphtha

Upper Mole Rate kmol/h Lower Mole Rate kmol/h Pressure MPa Temperature C Price ($/kmol) mole composition N2 CO2 CH4 C2 C3 C4 C5 C6

72000 45000

/

/

54000 37000

/

/

3.52

3.58

2.8

4.8

4.5

1.4

/

30 3.4

30 4.0

/ 5.0

/ 6.8

/ 6.0

/ 8.2

/ 6.1

0.1144 0.2028 0.5131 0.0323 0.0424 0.0325 0.0514 0.0111

0.0725 0.1271 0.5507 0.0526 0.0571 0.0632 0.0555 0.0213

/ 0.2 / / / / / /

/ 0.06 / / 0.01 0.004 0.003 0.003

/ 0.03 / / / / / /

0.03 0.03

/ / 0.02 / /

Table C2 Data for example. Parameters

Value

High-pressure steam pressure [MPa] High-pressure steam temperature [ C] Low-pressure steam pressure [MPa] Low-pressure steam temperature [ C] Power price [$/MW$h] Cost coefficient of gas pipeline bp,s,p0 Cost coefficient of gas pipeline sp,s,p0 Cost coefficient of gas pipeline jp,s,p0 Cost coefficient of liquid pipeline bp,s,p0 Cost constant of liquid pipeline sp,s,p0 Cost constant of liquid pipeline jp,s,p0 other cost [$/h] Useful life [Year] o Discount rate

4.52 425 0.6 197 162 12.237 0.036 0.313 7.335 0.019 0.172 56,000 1.5 2%

B.J. Zhang et al. / Journal of Cleaner Production 142 (2017) 3419e3436

3431

Table C3 Processes and streams. Processes

Process number

Code of processes

Stream exiting process

Subsea terminal Plug catcher Dew point controller Treating process Dehydration Compression Pipe Blending Separator Deethanizer Stabilizer CO2 product Gas customer LPG product Naphtha product Utility subsystem

2 3 3 3 3 3 3 3 3 3 1 3 1 1 1

SNOD1 SPCA1-SPCA3 SPDC1- SPDC3 SMDE1-SMDE3 SDEH1-SDEH3 SCOM1-SCOM3 SPLE1-SPLE3 SSEP1-SSEP3 SDEE1-SDEE3 SSTA1-SSTA3 ECAR ENOD1-ENOD ELPG ENAP SUTIL

G0 G1 (light stream), L1 G2 (light stream), L2 GA (gas), GE (carbon GB GC GD G3 (light stream), L3 G4 (light stream), L4 G5 (light stream), L5 / / / / /

(heavy stream) (heavy stream) dioxide)

(heavy stream) (heavy stream) (heavy stream)

Table C4 Exiting material stream network. Resource

Stream

Destination

Resource

Stream

Destination

Resource

Stream

Destination

SNOD1 SNOD1 SNOD2 SPCA1 SPCA1 SPCA1 SPCA1 SPCA1 SPCA1 SPCA1 SPCA1 SDPC1 SDPC1 SDPC1 SDPC1 SDPC1 SDPC1 SMDE1 SMDE1 SDEH1 SDEH1 SDEH1 SDEH1 SCOM1 SCOM1 SCOM1 SPLE1 SSEP1 SSEP1 SDEE1 SDEE1 SNOD1 SNOD1 SNOD2 SPCA1 SPCA1 SPCA1 SPCA1 SPCA1 SPCA1

G0 G0 G0 G1 G1 G1 G1 G1 L1 G1 G1 G2 G2 G2 L2 G2 G2 GA GE GB GB GB GB GC GC GC GD L3 G3 L4 G4 G0 G0 G0 G1 G1 G1 G1 G1 L1

SPCA1 SPCA2 SPCA3 SDPC1 SMDE1 SCOM1 SPLE1 SUTIL SSEP1 SPLE2 SPLE3 SMDE1 SCOM1 SPLE1 SSEP1 SPLE2 SPLE3 SDEH1 ECAR SCOM1 SPLE1 SPLE2 SPLE3 SPLE1 SPLE2 SPLE3 ENOD1 SDEE1 SUTIL SSTA1 SUTIL SPCA1 SPCA2 SPCA3 SDPC1 SMDE1 SCOM1 SPLE1 SUTIL SSEP1

SSTA1 SSTA1 SPCA2 SPCA2 SPCA2 SPCA2 SPCA2 SPCA2 SPCA2 SPCA2 SDPC2 SDPC2 SDPC2 SDPC2 SDPC2 SDPC2 SMDE2 SMDE2 SDEH2 SDEH2 SDEH2 SDEH2 SCOM2 SCOM2 SCOM2 SPLE2 SSEP2 SSEP2 SDEE2 SDEE2 SSTA2 SSTA1 SSTA1 SPCA2 SPCA2 SPCA2 SPCA2 SPCA2 SPCA2 SPCA2

G5 L5 G1 G1 G1 G1 G1 L1 G1 G1 G2 G2 G2 L2 G2 G2 GA GE GB GB GB GB GC GC GC GD L3 G3 L4 G4 G5 G5 L5 G1 G1 G1 G1 G1 L1 G1

ELPG ENAP SDPC2 SMDE2 SCOM2 SPLE2 SUTIL SSEP2 SPLE1 SPLE3 SMDE2 SCOM2 SPLE2 SSEP2 SPLE1 SPLE3 SDEH2 ECAR SCOM2 SPLE2 SPLE1 SPLE3 SPLE1 SPLE2 SPLE3 ENOD2 SDEE2 SUTIL SSTA2 SUTIL ELPG ELPG ENAP SDPC2 SMDE2 SCOM2 SPLE2 SUTIL SSEP2 SPLE1

SSTA2 SPCA3 SPCA3 SPCA3 SPCA3 SPCA3 SPCA3 SPCA3 SPCA3 SDPC3 SDPC3 SDPC3 SDPC3 SDPC3 SDPC3 SMDE3 SMDE3 SDEH3 SDEH3 SDEH3 SDEH3 SCOM3 SCOM3 SCOM3 SPLE3 SSEP3 SSEP3 SDEE3 SDEE3 SSTA3 SSTA3 SSTA2 SPCA3 SPCA3 SPCA3 SPCA3 SPCA3 SPCA3 SPCA3 SPCA3

L5 G1 G1 G1 G1 G1 L1 G1 G1 G2 G2 G2 L2 G2 G2 GA GE GB GB GB GB GC GC GC GD L3 G3 L4 G4 G5 L5 L5 G1 G1 G1 G1 G1 L1 G1 G1

ENAP SDPC3 SMDE3 SCOM3 SPLE3 SUTIL SSEP3 SPLE1 SPLE2 SMDE3 SCOM3 SPLE3 SSEP3 SPLE1 SPLE2 SDEH3 ECAR SCOM3 SPLE3 SPLE1 SPLE2 SPLE1 SPLE2 SPLE3 ENOD3 SDEE3 SUTIL SSTA3 SUTIL ELPG ENAP ENAP SDPC3 SMDE3 SCOM3 SPLE3 SUTIL SSEP3 SPLE1 SPLE2

3432

B.J. Zhang et al. / Journal of Cleaner Production 142 (2017) 3419e3436

Table C5 Potential material stream network for pipeline retrofit. Resource

Stream

Destination

Resource

Stream

Destination

Resource

Stream

Destination

SNOD1 SNOD2 SNOD2 SPCA1 SPCA1 SPCA1 SPCA1 SPCA1 SPCA1 SPCA1 SPCA1 SPCA2 SPCA2 SPCA2 SPCA2 SPCA2 SPCA2 SPCA2 SPCA2 SPCA3 SPCA3

G0 G0 G0 G1 G1 G1 L1 G1 G1 G1 L1 G1 G1 G1 L1 G1 G1 G1 L1 G1 G1

SPCA3 SPCA1 SPCA2 SDPC2 SMDE2 SCOM2 SSEP2 SDPC3 SMDE3 SCOM3 SSEP3 SDPC1 SMDE1 SCOM1 SSEP1 SDPC3 SMDE3 SCOM3 SSEP3 SDPC1 SMDE1

SPCA3 SPCA3 SPCA3 SPCA3 SPCA3 SPCA3 SDPC1 SDPC1 SDPC1 SDPC1 SDPC1 SDPC1 SDPC2 SDPC2 SDPC2 SDPC2 SDPC2 SDPC2 SDPC3 SDPC3 SDPC3

G1 L1 G1 G1 G1 L1 G2 G2 L2 G2 G2 L2 G2 G2 L2 G2 G2 L2 G2 G2 L2

SCOM1 SSEP1 SDPC2 SMDE2 SCOM2 SSEP2 SMDE2 SCOM2 SSEP2 SMDE3 SCOM3 SSEP3 SMDE1 SCOM1 SSEP1 SMDE3 SCOM3 SSEP3 SMDE1 SCOM1 SSEP1

SDPC3 SDPC3 SDPC3 SDEH1 SDEH1 SDEH2 SDEH2 SDEH3 SDEH3 SSEP1 SSEP1 SSEP2 SSEP2 SSEP3 SSEP3 SDEE1 SDEE1 SDEE2 SDEE2 SDEE3 SDEE3

G2 G2 L2 GB GB GB GB GB GB L3 L3 L3 L3 L3 L3 L4 L4 L4 L4 L4 L4

SMDE2 SCOM2 SSEP2 SCOM2 SCOM3 SCOM1 SCOM3 SCOM1 SCOM2 SDEE2 SDEE3 SDEE1 SDEE3 SDEE1 SDEE2 SSTA2 SSTA3 SSTA1 SSTA3 SSTA1 SSTA2

Table C6 Optimal process capacity and operating parameters before and after retrofit. Items retrofit

Capacity kmol/h

Temperature K

Before

After

Before

SNOD1 SNOD2 SPCA1 SPCA2 SPCA3 SDPC1 SDPC2 SDPC3 SMDE1 SMDE2 SMDE3 SDEH1 SDEH2 SDEH3 SCOM1 SCOM2 SCOM3 SSEP1 SSEP2 SSEP3 SDEE1 SDEE2 SDEE3 SSTA1 SSTA2 SSTA3 SPLE1 SPLE2 SPLE3 SUTIL ENOD1 ENOD2 ENOD3 ELPG ENAP ECAR

72000 45000 27938 44061 45000 25436 39499 37370 2104 39499 31435 1653 32000 29742 1653 32000 29742 2502 4561 13564 2316 4561 11562 2139 3799 8316 24985 29742 32000 6373 24985 29742 32000 7010 7244 9643

72000 45000 27000 45000 45000 20380 32706 35065 30000 3636 2304 26200 3222 1984 26200 28019 1984 2795 12568 12882 2609 7903 11243 2307 5446 8480 30000 54219 0 12011 30000 54219 0 8861 7373 4534

/ / 303.15 303.15 303.15 303.15 303.15 303.15 / / / / / / / / / 303.15 303.15 303.15 / / / / / / / / / / / / / / / /

Temperature change K

Pressure MPa

Pressure change MPa

After

Before

After

Before

After

Before

After

/ / 303.15 303.15 303.15 303.15 303.15 303.15 / / / / / / / / / 303.15 303.15 303.15 / / / / / / / / / / / / / / / /

/ / / / / 0 0 61.091 / / / / / / / / / / / / / / / / / / / / / / / / / / / /

/ / / / / 0 70.699 59.439 / / / / / / / / / / / / / / / / / / / / / / / / / / / /

3.520 3.580 2.939 3.520 3.580 2.939 3.520 3.580 2.939 3.520 3.580 2.939 3.520 3.580 2.939 3.520 3.580 1.461 3.520 3.580

3.520 3.580 3.520 3.520 3.580 3.520 3.520 3.580 3.520 3.520 3.580 3.520 3.520 3.580 3.520 3.520 3.580 2.069 3.520 3.580

2.800 4.800 4.500

2.800 4.800 4.500

2.800 4.800 4.500

2.800 4.800 4.500

/ / / / / / / / / / / / / / 0 0.98 1.22 / / / / / / / / / / / / / / / / / / /

/ / / / / / / / / / / / / / 1.28 1.28 0.92 / / / / / / / / / / / / / / / / / / /

B.J. Zhang et al. / Journal of Cleaner Production 142 (2017) 3419e3436

3433

Table C7 Optimal flowrates and compositions of process streams before retrofit. Streams

SNOD1.G0 SNOD2.G0 SPCA1.G1 SPCA1.L1 SPCA2.G1 SPCA2.L1 SPCA3.G1 SPCA3.L1 SDPC1.G2 SDPC1.L2 SDPC2.G2 SDPC2.L2 SDPC3.G2 SDPC3.L2 SMDE1.GA SMDE1.GE SMDE2.GA SMDE2.GE SMDE3.GA SMDE3.GE SDEH1.GB SDEH2.GB SDEH3.GB SCOM1.GC SCOM2.GC SCOM3.GC SSEP1.G3 SSEP1.L3 SSEP2.G3 SSEP2.L3 SSEP3.G3 SSEP3.L3 SDEE1.G4 SDEE1.L4 SDEE2.G4 SDEE2.L4 SDEE3.G4 SDEE3.L4 SSTA1.G5 SSTA1.L5 SSTA2.G5 SSTA2.L5 SSTA3.G5 SSTA3.L5 SPLE1.GD SPLE2.GD SPLE3.GD

Flowrate kmol/h

72000 45000 25436 2502 39499 4561 37370 7629 25436 0 39499 0 31435 5934 1653 450 32000 7499 29742 1692 1653 32000 29742 1653 32000 29742 186 2316 0 4561 2001 11562 176 2139 762 3799 3245 8316 720 1418 1421 2378 4868 3448 24985 29742 32000

Mole composition N2

CO2

CH4

C2H6

C3H8

C4H10

C5H12

C6H14

0.114 0.073 0.125 0.004 0.127 0.005 0.087 0.004 0.125 0.004 0.127 0.005 0.102 0.007 0.159 0.000 0.157 0.000 0.107 0.000 0.159 0.157 0.107 0.159 0.157 0.107 0.047 0.001 0.127 0.005 0.029 0.001 0.010 0.000 0.030 0.000 0.004 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.128 0.107 0.157

0.203 0.127 0.214 0.087 0.214 0.104 0.139 0.069 0.214 0.087 0.214 0.104 0.111 0.289 0.000 1.000 0.030 1.000 0.060 1.000 0.000 0.030 0.060 0.000 0.030 0.060 0.333 0.067 0.214 0.104 0.290 0.144 0.757 0.010 0.624 0.000 0.512 0.000 0.031 0.000 0.000 0.000 0.000 0.000 0.200 0.060 0.030

0.513 0.551 0.559 0.048 0.566 0.058 0.649 0.067 0.559 0.048 0.566 0.058 0.738 0.181 0.711 0.000 0.698 0.000 0.780 0.000 0.711 0.698 0.780 0.711 0.698 0.780 0.419 0.018 0.566 0.058 0.496 0.051 0.233 0.000 0.345 0.000 0.183 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.569 0.780 0.698

0.032 0.053 0.033 0.021 0.033 0.025 0.055 0.042 0.033 0.021 0.033 0.025 0.039 0.136 0.043 0.000 0.041 0.000 0.042 0.000 0.043 0.041 0.042 0.043 0.041 0.042 0.058 0.018 0.033 0.025 0.104 0.079 0.000 0.020 0.000 0.030 0.206 0.030 0.058 0.000 0.080 0.000 0.051 0.000 0.034 0.042 0.041

0.042 0.057 0.037 0.099 0.034 0.111 0.041 0.135 0.037 0.099 0.034 0.111 0.009 0.209 0.047 0.000 0.043 0.000 0.010 0.000 0.047 0.043 0.010 0.047 0.043 0.010 0.075 0.101 0.034 0.111 0.057 0.187 0.000 0.110 0.000 0.134 0.000 0.260 0.325 0.000 0.357 0.000 0.443 0.000 0.037 0.010 0.043

0.033 0.063 0.018 0.183 0.015 0.185 0.021 0.268 0.018 0.183 0.015 0.185 0.001 0.129 0.023 0.000 0.018 0.000 0.001 0.000 0.023 0.018 0.001 0.023 0.018 0.001 0.038 0.194 0.015 0.185 0.019 0.240 0.000 0.211 0.000 0.222 0.094 0.297 0.566 0.030 0.542 0.030 0.485 0.030 0.018 0.001 0.018

0.051 0.056 0.013 0.445 0.010 0.412 0.007 0.294 0.013 0.445 0.010 0.412 0.000 0.043 0.016 0.000 0.012 0.000 0.000 0.000 0.016 0.012 0.000 0.016 0.012 0.000 0.027 0.478 0.010 0.412 0.005 0.215 0.000 0.518 0.000 0.494 0.000 0.299 0.020 0.771 0.020 0.777 0.020 0.692 0.013 0.000 0.012

0.011 0.021 0.001 0.113 0.001 0.100 0.001 0.121 0.001 0.113 0.001 0.100 0.000 0.006 0.001 0.000 0.001 0.000 0.000 0.000 0.001 0.001 0.000 0.001 0.001 0.000 0.002 0.122 0.001 0.100 0.001 0.083 0.000 0.132 0.000 0.121 0.000 0.115 0.000 0.199 0.000 0.193 0.000 0.278 0.001 0.000 0.001

3434

B.J. Zhang et al. / Journal of Cleaner Production 142 (2017) 3419e3436

Table C8 Optimal flowrates and compositions of process streams after retrofit. Streams

SNOD1.G0 SNOD2.G0 SPCA1.G1 SPCA1.L1 SPCA2.G1 SPCA2.L1 SPCA3.G1 SPCA3.L1 SDPC1.G2 SDPC1.L2 SDPC2.G2 SDPC2.L2 SDPC3.G2 SDPC3.L2 SMDE1.GA SMDE1.GE SMDE2.GA SMDE2.GE SMDE3.GA SMDE3.GE SDEH1.GB SDEH2.GB SDEH3.GB SCOM1.GC SCOM2.GC SCOM3.GC SSEP1.G3 SSEP1.L3 SSEP2.G3 SSEP2.L3 SSEP3.G3 SSEP3.L3 SDEE1.G4 SDEE1.L4 SDEE2.G4 SDEE2.L4 SDEE3.G4 SDEE3.L4 SSTA1.G5 SSTA1.L5 SSTA2.G5 SSTA2.L5 SSTA3.G5 SSTA3.L5 SPLE1.GD SPLE2.GD SPLE3.GD

Flowrate kmol/h

72000 45000 24204 2795 40341 4658 37370 7629 20380 0 24796 7909 29812 5253 26200 3799 3222 414 1984 320 26200 3222 1984 26200 28019 1984 186 2609 4664 7903 1639 11243 301 2307 2457 5446 2762 8480 898 1409 2805 2641 5157 3322 30000 54219 0

Mole composition N2

CO2

CH4

C2H6

C3H8

C4H10

C5H12

C6H14

0.114 0.072 0.127 0.005 0.127 0.005 0.087 0.004 0.127 0.005 0.163 0.013 0.101 0.007 0.119 0.000 0.113 0.000 0.101 0.000 0.119 0.113 0.101 0.119 0.158 0.101 0.058 0.001 0.026 0.001 0.031 0.001 0.012 0.000 0.003 0.000 0.005 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.125 0.139 0.115

0.203 0.127 0.214 0.104 0.214 0.104 0.139 0.069 0.214 0.104 0.131 0.474 0.114 0.282 0.000 1.000 0.000 1.000 0.000 1.000 0.000 0.000 0.000 0.000 0.116 0.000 0.313 0.090 0.498 0.243 0.278 0.138 0.744 0.004 0.780 0.000 0.561 0.000 0.010 0.000 0.000 0.000 0.000 0.000 0.200 0.060 0.030

0.513 0.551 0.566 0.058 0.566 0.058 0.649 0.067 0.566 0.058 0.683 0.197 0.733 0.175 0.815 0.000 0.827 0.000 0.754 0.000 0.815 0.827 0.754 0.815 0.700 0.754 0.471 0.028 0.334 0.034 0.510 0.053 0.244 0.000 0.109 0.000 0.216 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.578 0.755 0.740

0.032 0.053 0.033 0.025 0.033 0.025 0.055 0.042 0.033 0.025 0.018 0.081 0.041 0.134 0.046 0.000 0.046 0.000 0.064 0.000 0.046 0.046 0.064 0.046 0.021 0.064 0.052 0.023 0.072 0.054 0.100 0.076 0.000 0.026 0.107 0.030 0.219 0.030 0.067 0.000 0.058 0.000 0.049 0.000 0.035 0.033 0.053

0.042 0.057 0.034 0.111 0.034 0.111 0.041 0.135 0.034 0.111 0.004 0.130 0.010 0.216 0.015 0.000 0.012 0.000 0.048 0.000 0.015 0.012 0.048 0.015 0.005 0.048 0.061 0.115 0.051 0.165 0.056 0.184 0.000 0.130 0.000 0.240 0.000 0.244 0.334 0.000 0.466 0.000 0.402 0.000 0.035 0.010 0.037

0.032 0.063 0.015 0.185 0.015 0.185 0.021 0.268 0.015 0.185 0.000 0.061 0.001 0.136 0.003 0.000 0.001 0.000 0.025 0.000 0.003 0.001 0.025 0.003 0.000 0.025 0.027 0.196 0.013 0.162 0.019 0.243 0.000 0.221 0.000 0.235 0.000 0.322 0.569 0.000 0.456 0.000 0.529 0.000 0.016 0.002 0.018

0.051 0.055 0.010 0.412 0.010 0.412 0.007 0.294 0.010 0.412 0.000 0.040 0.000 0.045 0.002 0.000 0.000 0.000 0.008 0.000 0.002 0.000 0.008 0.002 0.000 0.008 0.018 0.440 0.007 0.279 0.005 0.220 0.000 0.497 0.000 0.405 0.000 0.291 0.020 0.801 0.020 0.814 0.020 0.713 0.010 0.001 0.006

0.011 0.021 0.001 0.100 0.001 0.100 0.001 0.121 0.001 0.100 0.000 0.003 0.000 0.006 0.000 0.000 0.000 0.000 0.001 0.000 0.000 0.000 0.001 0.000 0.000 0.001 0.001 0.107 0.000 0.062 0.001 0.085 0.000 0.121 0.000 0.090 0.000 0.113 0.000 0.199 0.000 0.186 0.000 0.287 0.001 0.000 0.001

Table C9 Optimal material stream network for operation before retrofit (kmol/h). Stream direction

Flowrate

Stream direction

Flowrate

Stream direction

Flowrate

SNOD1.G0.SPCA1 SNOD1.G0.SPCA2 SNOD2.G0.SPCA3 SPCA1.G1.SDPC1 SPCA1.L1.SSEP1 SPCA2.G1.SDPC2 SPCA2.L1.SSEP2 SPCA3.G1.SDPC3 SPCA3.L1.SSEP3 SDPC1.G2.SMDE1 SDPC1.G2.SPLE1 SDPC2.G2.SMDE2 SDPC3.G2.SMDE3 SDPC3.L2.SSEP3 SMDE1.GA.SDEH1 SMDE1.GE.ECAR

27938.68 44061.32 45000 25436.38 2502.29 39499.54 4561.78 37370.53 7629.46 2104.7 23331.63 39499.49 31435.71 5934.81 1653.92 450.77

SMDE2.GA.SDEH2 SMDE2.GE.ECAR SMDE3.GA.SDEH3 SMDE3.GE.ECAR SDEH1.GB.SCOM1 SDEH2.GB.SCOM2 SDEH3.GB.SCOM3 SCOM1.GC.SPLE1 SCOM2.GC.SPLE3 SCOM3.GC.SPLE2 SSEP1.G3.SUTIL SSEP1.L3.SDEE1 SSEP2.L3.SDEE2 SSEP3.G3.SUTIL SSEP3.L3.SDEE3 SDEE1.G4.SUTIL

32000 7499.49 29742.75 1692.96 1653.92 32000 29742.75 1653.92 32000 29742.75 186 2316.34 4561.78 2001.94 11562.34 176.75

SDEE1.L4.SSTA1 SDEE2.G4.SUTIL SDEE2.L4.SSTA2 SDEE3.G4.SUTIL SDEE3.L4.SSTA3 SSTA1.G5.ELPG SSTA1.L5.ENAP SSTA2.G5.ELPG SSTA2.L5.ENAP SSTA3.G5.ELPG SSTA3.L5.ENAP SPLE1.GD.ENOD1 SPLE2.GD.ENOD2 SPLE3.GD.ENOD3

2139.58 762.66 3799.12 3245.95 8316.38 720.99 1418.59 1421.09 2378.02 4868.02 3448.35 24985.55 29742.75 32000

B.J. Zhang et al. / Journal of Cleaner Production 142 (2017) 3419e3436

3435

Table C10 Optimal material stream network after retrofit (kmol/h). Stream direction

Flowrate

Stream direction

Flowrate

Stream direction

Flowrate

SNOD1.G0.SPCA1 SNOD1.G0.SPCA2 SNOD2.G0.SPCA3 SPCA1.G1.SDPC1 SPCA1.G1.SMDE1 SPCA1.L1.SSEP1 SPCA2.G1.SDPC2 SPCA2.G1.SPLE1 SPCA2.L1.SSEP2 SPCA3.G1.SDPC3 SPCA3.G1.SMDE3 SPCA3.L1.SSEP3 SDPC1.G2.SPLE1 SDPC2.G2.SCOM2 SDPC2.L2.SSEP2 SDPC3.G2.SMDE1 SDPC3.G2.SMDE2

27000 45000 45000 20380.9 3823.71 2795.37 32706.38 7634.65 4658.96 35065.66 2304.87 7629.46 20380.85 24796.89 7909.48 26176.28 3636.34

SDPC3.L2.SSEP3 SMDE1.GA.SDEH1 SMDE1.GE.ECAR SMDE2.GA.SDEH2 SMDE2.GE.ECAR SMDE3.GA.SDEH3 SMDE3.GE.ECAR SDEH1.GB.SCOM1 SDEH2.GB.SCOM2 SDEH3.GB.SCOM3 SCOM1.GC.SPLE2 SCOM2.GC.SPLE2 SCOM3.GC.SPLE1 SSEP1.G3.SUTIL SSEP1.L3.SDEE1 SSEP2.G3.SUTIL SSEP2.L3.SDEE2

5253.03 26200.34 3799.65 3222.26 414.07 1984.53 320.34 26200.34 3222.26 1984.53 26200.34 28019.15 1984.49 186 2609.42 4664.65 7903.79

SSEP3.G3.SUTIL SSEP3.L3.SDEE3 SDEE1.G4.SUTIL SDEE1.L4.SSTA1 SDEE2.G4.SUTIL SDEE2.L4.SSTA2 SDEE3.G4.SUTIL SDEE3.L4.SSTA3 SSTA1.G5.ELPG SSTA1.L5.ENAP SSTA2.G5.ELPG SSTA2.L5.ENAP SSTA3.G5.ELPG SSTA3.L5.ENAP SPLE1.GD.ENOD1 SPLE2.GD.ENOD2

1639.44 11243.05 301.84 2307.58 2457.28 5446.51 2762.69 8480.36 898.21 1409.37 2805.41 2641.1 5157.66 3322.7 30000 54219.49

Table C11 Optimal results of utility subsystems before retrofit. Process

Heat duties MJ/h

Boiler A Boiler B Boiler C Turbine A Turbine B Turbine C Turbine D SDPC1 SDPC2 SDPC3 SMDE1 SMDE2 SMDE3 SCOM1 SCOM2 SCOM3 SDEE1 SDEE2 SDEE3 SSAT1 SSAT2 SSAT3

1.8837E þ 5 6.2125E þ 5 4.1417E þ 5 / / / / / / / 7356.720 1.2239E þ 5 24175.543 / / / 1555.875 5551.712 34965.010 33701.814 65816.922 2.6863E þ 5

High pressure steam kmol/h

Low pressure steam kmol/h

Power MW

Production

Consumption

Production

Consumption

Production

Consumption

2664 9000 6000 / / / / / / / / / / / / / / / / / / /

/ / / 10089 7575 / / / / / / / / / / / / / / / / /

/ / / 10089 7575 / / / / / / / / / / / / / / / / /

/ / / / / 1624 0 / / / 196 3254 643 0 0 0 41 148 930 896 1750 7143

/ / / 24 18 5 0 / / / / / / / / / / / / / / /

/ / / / / / / 0 0 12.539 0.064 1.127 0.244 0 72.873 70.958 / / / / / /

Table C12 Optimal results of utility subsystems after retrofit. Process

Boiler A Boiler B Boiler C Turbine A Turbine B Turbine C Turbine D SDPC1 SDPC2 SDPC3 SMDE1 SMDE2 SMDE3 SCOM1 SCOM2 SCOM3 SDEE1 SDEE2 SDEE3 SSAT1 SSAT2 SSAT3

Heat duties MJ/h

1.8837E 6.2125E 4.1417E / / / / / / / 62010 6758 4574 / / / 2582 28297 38189 42251 1.1266E 3.0593E

þ5 þ5 þ5

þ5 þ5

High pressure steam kmol/h

Low pressure steam kmol/h

Power MW

Production

Consumption

Production

Consumption

Production

Consumption

2664 9000 6000 / / / / / / / / / / / / / / / / / / /

/ / / 10089 7575 / / / / / / / / / / / / / / / / /

/ / / 10089 7575 / / / / / / / / / / / / / / / / /

/ / / / / 1624 0 / / / 1649 180 122 0 0 0 69 752 1015 1123 2995 8134

/ / / 24 18 8 0 / / / / / / / / / / / / / / /

/ / / / / / / 0 13.904 11.317 0.541 0.062 0.046 69.640 73.009 4.080 / / / / / /

3436

B.J. Zhang et al. / Journal of Cleaner Production 142 (2017) 3419e3436

References Arredondo-Ramírez, K., Ponce-Ortega, J.M., El-Halwagi, M.M., 2016. Optimal planning and infrastructure development for shale gas production. Energy Conv. Manag. 119, 91e100. Banimostafa, A., Papadokonstantakis, S., Hungerbühler, K., 2015. Retrofit design of a pharmaceutical batch process considering “green chemistry and engineering” principles. AIChE J. 61 (10), 3423e3440. Boot-Handford, M.E., Abanades, J.C., Anthony, E.J., Blunt, M.J., Brandani, S., Dowell, N.M., et al., 2014. Carbon capture and storage update. Energy Environ. Sci. 7, 130e189. Branan, C.R., 2005. Rules of Thumb for Chemical Engineers, third ed. Elsevier, Amsterdam. Brooke, A., Kendrick, D., Meeruas, A., Raman, R., 2015. GAMS: a User's Guide. GAMS Development Corporation, Washington (DC). Caballero, J.A., Grossmann, I.E., 2014. Optimal synthesis of thermally coupled distillation sequences using a novel MILP approach. Comput. Chem. Eng. 61, 118e135. de Queiroz Fernandes Araújo, O., de Carvalho Reis, Alessandra, de Medeiros, J.L., do Nascimento, J.F., Grava, W.M., Musse, A.P.S., 2016. Comparative analysis of separation technologies for processing carbon dioxide rich natural gas in ultradeepwater oil fields. J. Clean. Prod. http://dx.doi.org/10.1016/ j.jclepro.2016.06.073. Demierre, J., Bazilian, M., Carbajal, J., Sherpa, S., Modi, V., 2015. Potential for regional use of East Africa's natural gas. Appl. Energy 143, 414e436. Diaz, M.S., Serrani, A., Bandoni, J.A., Brignole, E.A., 1997. Automatic design and optimization of natural gas plants. Ind. Eng. Chem. Res. 36 (7), 2715e2724. Enríquez, A.H., Tanco, M., Kim, J.K., 2011. Simulation-based process design and integration for the sustainable retrofit of chemical processes. Ind. Eng. Chem. Res. 50 (21), 12067e12079. zquez-Roma n, R., Grossmann, I.E., Iglesias-Silva, G., 2011. Flores-Salazar, M.A., Va A multiperiod planning model for gas production system. J. Pet. Sci. Eng. 77, 226e235. International Energy Statistics. 2015. http://www.eia.gov/cfapps/ipdbproject/ iedindex3.cfm?tid¼3&pid¼26&aid¼2. Kantor, I., Betancourt, A., Elkamel, A., Fowler, M., Almansoori, A., 2015. Generalized mixed-integer nonlinear programming modeling of eco-industrial networks to reduce cost and emissions. J. Clean. Prod. 99, 160e176. Kleme
s, J.J., Kravanja, Z., 2013. Forty years of Heat Integration: Pinch Analysis (PA) and Mathematical Programming (MP). Curr. Opin. Chem. Eng. 2, 461e474.  ski, P., 2015. Energy and exergy reKostowski, W.J., Kalina, J., Bargiel, P., Szuflen covery in a natural gas compressor station e A technical and economic analysis. Energy Conv. Manag. 104, 17e31. Lambrecht, H., Thißen, 2015. Enhancing sustainable production by the combined use of material flow analysis and mathematical programming. J. Clean. Prod. 105, 263e274. Li, X., Armagan, E., Tomasgard, A., Barton, P.I., 2011. Stochastic pooling problem for natural gas production network design and operation under uncertainty. AIChE J. 57, 2120e2135. Long, N.V.D., Minh, L.Q., Nhien, L.C., Lee, M., 2015. A novel self-heat recuperative dividing wall column to maximize energy efficiency and column throughput in retrofitting and debottlenecking of a side stream column. Appl. Energy 159, 28e38. Misener, R., Floudas, C.A., 2014. ANTIGONE: Algorithms for coNTinuous/Integer Global Optimization of Nonlinear Equations. J. Glob. Optim. 59, 503e526.

Mokhatab, S., Poe, W.A., Mak, J.Y., 2015. Handbook of Natural Gas Transmission and Processing: Principles and Practices, third ed. Gulf Professional Publishing, MA. Moshfeghian, M., 2013. Variation of Ideal Gas Heat Capacity Ratio with Temperature and Relative Density (Tip of the Month). John M. Campbell & Co., Norman, OK, USA. Neseli, M.A., Ozgener, O., Ozgener, L., 2015. Energy and exergy analysis of electricity generation from natural gas pressure reducing stations. Energy Conv. Manag. 99, 109e120. Neveu, P., Tescari, S., Aussel, D., Mazet, N., 2013. Combined constructal and exergy optimization of thermochemical reactors for high temperature heat storage. Energy Conv. Manag. 71, 186e198. Pan, M., Aziz, F., Li, B., Perry, S., Zhang, N., Bulatov, I., Smith, R., 2016. Application of optimal design methodologies in retrofitting natural gas combined cycle power plants with CO2 capture. Appl. Energy 161, 695e706. Ríos-Mercado, R.Z., Borraz-S anchez, C., 2015. Optimization problems in natural gas transportation systems: A state-of-the-art review. Appl. Energy 147, 536e555. Sanaye, S., Mahmoudimehr, J., 2013. Optimal design of a natural gas transmission network layout. Chem. Eng. Res. Des. 91 (12), 2465e2476. Selot, A., Kuok, L.K., Robinson, M., Mason, T.L., Barton, B.I., 2008. A Short-Term Operational Planning Model for Natural Gas Production Systems. AIChE J. 54 (2), 495e515. Sueviriyapan, N., Suriyapraphadilok, U., Siemanond, K., Quaglia, A., Gani, R., 2016. Industrial wastewater treatment network based on recycling and rerouting strategies for retrofit design schemes. J. Clean. Prod. 111, 231e252. Tabkhi, F., Pibouleau, L., Hernandez-Rodriguez, G., Azzaro-Pantel, C., Domenech, S., 2010. Improving the performance of natural gas pipeline networks fuel consumption minimization problems. AIChE J. 56, 946e964. lu, S¸., 2014. Optimization for design and operation of natural gas Üster, H., Dilaverog transmission networks. Appl. Energy 133, 56e69. Wang, T., Lin, B., 2017. China's natural gas consumption peak and factors analysis: a regional perspective. J. Clean. Prod. 142 (P2), 548e564. http://dx.doi.org/ 10.1016/j.jclepro.2016.04.095. Wei, Z.Q., Zhang, B.J., Wu, S.Y., Chen, Q.L., Hui, C.W., 2012. A hydraulics-based heuristic strategy for capacity expansion retrofit of distillation systems and an industrial application on a light-ends separation plant. Chem. Eng. Res. Des. 90 (10), 1527e1539. Wilson, G., 1968. A modified Redlich-Kwong equation of state applicable to general physical data calculations. In: 65th AIChE National Meeting. pez, R.A., 2008. A cost function for the natural gas transmission industry. Eng. Ye Econ. 53 (1), 68e83. Yong, J.Y., Kleme
s, J.J., Varbanov, P.S., Huisingh, D., 2016. Cleaner energy for cleaner production: modelling, simulation, optimisation and waste management. J. Clean. Prod. 111, 1e16. Zarogiannis, T., Papadopoulos, A.I., Seferlis, Panos, 2016. Systematic selection of amine mixtures as post-combustion CO2 capture solvent candidates. J. Clean. Prod. 136 (PB), 159e175. http://dx.doi.org/10.1016/j.jclepro.2016.06.073. Zhang, B.J., Wu, S.Y., Chen, Q.L., 2012. An optimization procedure for retrofitting process energy systems in refineries. Comput. Aided Chem. Des. 31, 1005e1009. Zhang, B.J., Liu, K., Luo, X.L., Chen, Q.L., Li, W.K., 2015. A multi-period mathematical model for simultaneous optimization of materials and energy on the refining site scale. Appl. Energy 143, 238e250. Zheng, Q.P., Rebennack, S., Iliadis, N.A., Pardalos, P.M., 2010. Optimization models in the natural gas industry. In: Rebennack, S., Pardalos, P.M., Pereira, M.V., Iliadis, N.A. (Eds.), Handbook of Power Systems I, Chap. 6. Springer, pp. 121e148.