Multiobjective optimization scheme for industrial synthesis gas sweetening plant in GTL process

Availableonlineatwww.sciencedirect.comJoumal ofScience DirectNatural GasChemistryELSEVIERJournal of Natural Gas Chemistry 20(2011)99109Multiobjective optimization scheme for industrial synthesis gassweetening plant in GtL processAlireza Behroozsarand 1* Akbar ZamaniyanIslamic Azad Universit IIkhchi branch. P0 Box. 651335-7996. Tabriz ran2. Department of Natural Gas Conversion, Gas Research Division, Research Institute of Petroleum Industry(RIPI), Tehran 14665-1998, IranManuscript received April 9, 2010: revised October 6, 2010 1AbstractIn industrial amine plants the optimized operating conditions are obtained from the conclusion of occurred events and challenges that arenormal in the working units. For the sake of reducing the costs, time consuming, and preventing unsuitable accidents, the optimization couldbe performed by a computer program. In this papsis of amine plant is performed at first. The optimization ofthis unit is studied using Non-dominated sortingl in order produce sweet gas with CO2 mole percentage less than 2.0%and H2S concentration less than 10 ppm for application in, FischeThe simulation of the plant in HYSYS V.3. I software habeen linked with MATLAB code for real-parameterthe amine process. Three scenarios are selected to corthe effect of (DEA/MDEA)mass composition peratctive junctions. Results show that sour gas temperatuand pressure of 3398C and 14.96 bar, DEA/CO2 molar flow ratitemperature and pressure of 94.92C and 3.0 b40regenerator pressure of 1.5 bar, and ratio of DEA/MDEA=20%for minimizing plant energy consumption, aminecirculation rate, and carbon dioxide recovery.Key wordsamine plant; multiobjective optimization; Non-Don ed Softing Genetic Algorithm; amine circulation rate1. Introductionas the base amine with the addition of one or two more reactive amines such as mea or deaNatural and refinery gases contain typically acid gases inThe major advantage of the amine treatment is that it isconcentrations ranging from a few parts per million to tens of widely commercialized technology in which the hydrocarbdrmolume percentage [1]. The major acid gases are hydrogen loss is almost negligible. However, the operating and capsulfide(H2S) and carbon dioxide(CO2). The typical prod- tal costs shoot up very rapidly as the concentration of carbouct is a CH4-enriched residue stream containing less than 2% dioxide in the feed gas increases [21CO2, which is sold as a pipeline fuel [2, 3]. Carbon dioxide,Technology using alkanolamine solutions, or simplewhich falls into the category of acid gases, is highly corrosive amine solutions, for the removal of hydrogen sulfide and carand rapidly destroys pipelines and equipment; it also reduces bon dioxide from natural gas has been around for decadesthe heating value of a natural gas stream and wastes pipeline Since the 1960s and 1970s, several amines have come intoapacity [2, 4]. Acid gas removal and dehydration are the most general use, but there is little information available on whichcommonly employed processes and various technologies are amine is best suited to a particular service. Many inefficientilable to design engineer for these processes [5]. Acid amine gasoptimized bygas removal technologies include absorption with an aqueous changing their amine solutions [7]. Aqueous alkanolaminealkanolamine solution, cryogenic adsorption and membrane solutions are widely used for the removal of acid gasesprocesses [4, 6]. The chemical solvents and physical solvents such as CO2 and H2S from gas streams. Examples ofor combination of these two have been used extensively in ex- such streams include natural gases, synthesis gases from theisting base load LNG facilities. In the past few years, mixed gasification of coal and heavy oils, and tail gases from sulfuramine solvents for the removal of acid gases have received in- plants and petroleum chemical plants. Aqueous solutions ofcreased attention. In most cases, the mixtures contain MDEa alkanolamines reaes and therefore中国煤化工Corresponding author. Tel: +98(412)3459142: Fax: +98(412)3444355: E-mail: behroozsarandar(@CNMHGCopyright(2011, Dalian Institute of Chemical Physics, Chinese Academy of Sciences. All rights reservedoi:10.1016/S1003-9953(1060153-3100Alireza Behroozsarand et al Journal of Natural Gas Chemistry VoL. 20 No I 2011are widely used to remove them. The alkanolamines solution ing HYSYS V.3. I software. The plant is used for carbongroups include monoethanolamine (MEA), diethanolamine dioxide and hydrogen sulfide capture from synthesis gas in(EA), methyldiethanolamine (MDEA), di-propanolamine the Ft process(iron catalyst based). After that, three op-(IPA), and diglycolamine( DGA). Although the acid gas- timization scenarios of plant using Non-Dominated Sortingamine reactions are reversible, irreversible reactions may also Genetic Algorithm-II is performed by considering minimizoccur, resulting in the products from which the amines are not ing the plant energy consumption, amine circulation rate, andeasily recoveredmaximizing carbon dioxide recoveryThe Fischer-Tropsch process (or Fischer-Tropsch Syrthesis) is a catalyzed chemical reaction in which synthesis 2. Amine treating facilitygas(syngas), a mixture of carbon monoxide and hydrogenis converted into liquid hydrocarbons in various forms. TheFigure I represents a simple amine treating facility. Sourmost common catalysts are based on iron and cobalt, although gas is introduced in the absorber where it contacts lean aminenickel and ruthenium have also been used. The principal pur- solution traveling down the column. The acid gas componentpose of this process is to produce synthetic petroleum, typi- CO2 and H2S, is absorbed by the amine solution and the sweetcally from coal, natural gas or biomass. In the FT process, gas leaves the absorber for further processing. The rich amineproduced synthesis gas enters gas sweetening plant to capture solution is sent to a flash tank and absorbed hydrocarbonscarbon dioxide and hydrogen sulfide. According to experexit as the flash-tank vapor. The rich amine flows throughmental data, in iron based FT process, carbon dioxide and hy- the lean/rich exchanger to increase the temperature to aboutdrogen sulfide concentrations in synthesis gas stream must be 90-110C. The hot rich amine is stripped at low pressurelower than 2%(mol%)and 10 ppm, respectively. Because to remove the absorbed acid gases, dissolve hydrocarbons andhigher carbon dioxide and hydrogen sulfide contents cause some water. The energy required to strip the amine is the sumide reactions and deactivate iron catalystsof the sensible heat required to raise the solution temperature,In this paper, an industrial amine plant is simulated us- the energy of absorption, and latent heats [81Amine-makeupOff-gasSour-gasFlash drumFigure 1. Schematic of simple amine sweetening plant [8]The pressure of stripping column should be operated at ashigh as possible to increase the reboiler temperature for optimum stripping [9]. However, the amine degradation temperRCirculation Amine circulation molar flow(kgmole/h)ature should not be exceeded. The stripped or lean amine is(2)sent back through the lean/rich exchanger to decrease its temperature. A pump boosts the pressure such that it is greater CO2 recovery(Reco, )(%)=-( in (CO2)outx 100than the absorber column. Finally, a heat exchanger or aircooler cools the lean solution before the loop is back to theIn this study, the response factors are the plant energy con- 3. Genetic algorithm for multiobjective optimizationsumption(tower energy), amine circulation rate, and carbondioxide recoverTH中国煤化工3.1. Genetic algeCNMHG+(pUmps the new methods, genetic algorithms have been sucJournal of Natural Gas Chemistry VoL. 20 No. 1 201Icessfully applied to nonlinear optimization problems in many provide a true picture of trade-offs for the decision-makerdimensions where traditional methods are often found to fail()The best-known Pareto front should capture the whole[10]. Moreover, more traditional ones such as deterministic spectrum of the Pareto front. This requires investigating solu-d gradient-based optimization methods do not search the tions at the extreme ends of the objective function space [111parameter space and can tend to converge towards local extreme of the fitness function. It is clearly unsatisfactory for 3.3. Multiobjective GAproblems where the fitness varies non-monotonously with parameters. On the other hand, genetic algorithms are able toBeing a population-based approach, Ga is well suited todepart from local optima due to the variability of the paramsolve multiobjective optimization problems. A generic singleeters within thegene pool"and the element of randomness objective ga can be modified to find a set of multiple non-inherent within the methods. Furthermore, genetic algorithms dominated solutions in a single run. The ability that Ga sido not require knowledge of the gradient of the fitness funcmultaneously searches different regions of a solution spacetions, which makes them particularly suited to optimization makes it possible to find a diverse set of solutions for difficultproblems for which an analytical expression is not known forproblems with non-convex, discontinuous, and multi-modalthe fitness functions3. 2. Multiobjective optimization3.4. Model equations for process parameter optimizationThere are two general approaches to multiple-objectiveThe optimization problem is considered for an industrialoptimization. One is to combine the individual objective funcretrofit amine plant. In this optimization, sour gas flow rate istions into a single composite function or move all but one constant. By study of sensitivity analysis of effective param-objective to the constraint set. In the former case, determi- eters on objective functions(Section 4.1), the trend of thination of a single objective is possible with methods such asmain objective functions, amine circulation rate, plant energyutility theory, weighted sum method. But the problem lies consumption, and carbon dioxide recovery, were investigatedin the proper selection of the weights or utility functions to in wide ranges of operation conditions. It can logically searchharacterize the decision-maker's preferences [ll]. The sec-for operating scenarios that will minimize plant energy conond general approach is to determine an entire Pareto optimalsumption, amine circulation rate and maximize carbon dietion set or a representative subset. A Pareto optimal set ide recovery simultaneously. Performing a constrainedis a set of solutions that are non-dominated with respect to mization with both of them as objectives can identify sucheach other. While moving from one Pareto solution to an- scenarios. The optimization problem can be expressed mathother, there is always a certain amount of sacrifice in one ob- ematically as followingjective(s)to achieve a certain amount of gain in the other(s)MinimizePareto optimal solution sets are often preferred to single solutions because they can be practical when considering real-lifeObject(1)=ENetproblems since the final solution of the decision-maker isObject(2)=Circulation(5)ways a trade-off. Pareto optimal sets can be of varied sizes,but the size of the Pareto set usually increases with theMaximizecrease in the number of objectivesObject(3)=RecoThe ultimate goal of a multiobjective optimization algoSubject to six decision variablesrithm is to identify solutions in the Pareto optimal set. However, identifying the entire Pareto optimal set, for many mul125≤FDEA/Fco2≤14.5tiobjective problems, is practically impossible due to its size27≤ Tsourgas(C)≤35In addition, for many problems, especially for combinatorialoptimization problems, proof of solution optimality is compu10≤ Soure(bar)≤(9)tationally infeasible. Therefore, a practical approach to multiobjective optimization is to investigate a set of solutions(the90< TReg gas(C)<95(10)best-known Pareto set) that represent the Pareto optimal set aswell as possible. With these concerns in mind, a multiobjec3.0< PReg gas(bar)<7.0(11)tive optimization approach should achieve the following three1.5< PRegenerator(bar)< 2.0inflicting goals [121(1)The best-known Pareto front should be as close as posMDEA/DEA)Mass(%)=0%/30%sible to the true Pareto front. Ideally, the best-known ParetoYHa中国煤化工set should be a subset of the Pareto optimal set.CNMHG(14)(2) Solutions in the best-known Pareto set should be uni-formly distributed and diverse over the Pareto front in order toH2S(Mole %)@Sweet gas< 10 ppm102Alireza Behroozsarand et al Journal of Natural Gas Chemistry VoL. 20 No I 2011The range of the feed temperature, feed pressure, preTable 2. Gas sweetening plant operation conditionssure of regeneration gas, regenerator pressure, and DEA/COmolar flow ratio constraints have been selected based on theAmine circulation rate(kgmol/h)39440general industrial amine plants data and sensitivity analysis ofAbsorber column top/bottom pressure(bar)14.4/14.9simulation resultsAbsorber column top/bottom temperaThe ratio of (DEA/MDEA)mass composition percentaStripper column top/bottom pressure(bar)18/2.5Stripper column top/bottom temperature (C)48/124at amine recycling stream is the special decision variable. BeNumber of actual tray (absorber)cause of changing value of this variable among simulationNumber of actual tray(stripper)running is impossible, three scenarios have been predicted forstudying effect of ratio of (DEA/MDEA)mass compositionTable 3 presents results of the model output and induspercentage on optimization problem. Three scenarios aretrial data. It shows that the model is able to predict the plant(1)DEAmine and MDEAmine mass composition per- performance with high accuracycentages in recycling amine stream are equal to 30% and o%respectively.Table 3. Comparison between simulation and plant data(2) DEAmine and MDEAmine mass compositionOperating data Simulation data R.E.(%)centages in recycling amine stream are equal to 20% and1ich amine loading0.43Lean amine loading12.50respectively.CO of sweet gas (mol%)<0.4(3)DEAmine and MDEAmine mass composition per- Absorber column top (c)centages in recycling amine stream are equal to 10% and 20%0Bottom temperature ( C)44.66respectivelyBottom te1.614. Results and discussionRelative error(rE )(%)=Absolute((Operating data-Simulationdata)/Operating data)x 1004/. Simulation and sensitivity analysisIn the follesection effects of process variables onobjective functions are consideredA typical Iranian gas plant(Kangan gas refinery) is se-4.1.1 Sour gas temperatureamine trains for COz and HaS removal. Ekieeditepper at ure( ts effect of sour gas temerature on tail106540t simulator was used to simulate the process. The gas methane flow(FCHa), plant energy consumption(Amine Package thermodynamic model was used. The absorber feed gas composition is shown in Table 1. It shows that2190the feed contains mainly H2, CO, CO2, H2S, and Cha that theCO2 and H2S should be removed. Operating conditions aresummarized in Table 2218057106530 Table 1. Sour gas stream composition dataParameter217.5Mass flow(kg/h)108448.8STD_gas flow(STD m /h)1182222217.0Composition(mol%)Hydrogen29.83216.529,2522242628303234363840106520Feed temperature(℃)0.68106540Methane2.4le-0l1.7342106530·········10652010651O1.62e-04n-Butane1.93e0l106510i-Pentane106500LDEAmine中国煤化工403MDEAmineCNMHGFigure 2. Effect of sour gas temperature on objective functions and10656mmc=1040 kg/m Mw=191i7molar flow22242628303234363840Feed t enper at ure(。QJournal of Natural Gas Chemistry VoL. 20 No. 1 201I103amine circulation rate(RCirculation and carbon dioxide mole DEAmine liquid phase in absorption tower. According to Figpercentage of sweet gas( CO2(%)). The carbon dioxide mole ure 3 the above pressure (10-15 bar)is acceptable value forpercentage of sweet gas represents the carbon dioxide recov- sour gas pressure. According to the literatures [13, 14-17ery behavior. It is obvious that both tail gas methane flow and FTreactions occur at 15-30 bar. But high manufacturing andplant energy consumption are increased and amine circulation facility cost of amine plant in the high pressure urges us toIte is decreased with the increase of the sour gas temperature choose the lowest possible pressure of 15 baras an adjustable input variable. From 22C to 32C of sourgas temperature, the CO2 mole percentage is decreased, but at 4.1.3. DEA/CO2 molar flow ratiohigher temperature, this parameter is increased. It is clear thatthe increase of plant energy consumption and tail gas flow isundesirable because of economic and environmental aspectsFigure 4 presents effect of ratio of DEAmine to feed CO2molar flow on FCHA, ENet, RCirculation and carbon dioxide moleAccording to trend of parameters in Figure 2 and opti- percentage of sweet gas. It is clear that increasing DEA/CO2mization purpose, temperature between 25C to 35C is the molar flow causes all parameters, except carbon dioxide molebest choice for sour gas temperature constraint. In this simpercentage of sweet gas increasedulation, amine circulation rate has been adjusted by ratio ofDEAmine to co molar flow28041.2. Sour gas pressure3.8Figure 3 shows effect of sour gas pressure on FCH, ENetRCirculation and carbon dioxide mole percentage of sweet gasBy increasing sour gas pressure, all parameters, except FCH240are increased. Therefore increasing sour gas pressure has urfavorable effect on optimization objects. It must be noted thatregularly inputting sour gas to amine plant has pressure about10 to 15 bar. High pressure is needed for solving gas phase in222 ress11.512,012.513.013.514.014.515,0155Ratio of FDEA/Fco14000010∈E12000010∈1000001012141618202224262830Feed pressure(bar)80000L1⊥L11.512.012.513.013.514.014.515.01510653010∈Figure 4. Effect of DEA/COz molar flow ratio on objective functions andtail gas molar flow106520品aIn industrial cases, such as hydrogen plant, the carbondioxide mole percentage in feed gas is limited to prevent the106510catalyst deactivation in the reactors So in some cases wemust use high amine circulation rate for this purpose. Using10∈=106500the wide range of DEA/CO2 molar flow ratio in optimization EKproblem for observing plant treatment is proposed10121416182022242628304. 1.4. Regenerati中国煤化工Feed pressure(bar)HCNMHFigure 3. Effect of sour gas pressure on objective functions and tail gas molarFigure 5gas temperature106490RCirculation and carbon dioxide mole percentFeedessur( bar104Alireza Behroozsarand et al Journal of Natural Gas Chemistry VoL. 20 No I 2011age of sweet gas. Increasing regeneration gas temperature de- pressure and high temperature has high yield. Therefore highcreases the four parameters. According to Figure 5 carbon pressure in stripper is suitable for stripping of CO2 and otherdioxide mole percentage of sweet gas is almost constant after light soluble gases from DEAmine liquid phasetemperature above 93C. It is noted that increasing regeneration gas temperature causes that duty of heat exchanger before regeneration tower is increased and mechanical design ofstripper should be performed in higher temperature. So the240三best temperature range could be gained by considering all parameters(90-95°C3.888二230三Regenaration gas pressure(bar)120000l1500000Regeneration gas temperature(C)总105000115000100000⊥⊥A⊥⊥⊥1.40E110000a gas.t eise atFigure 6. Effect regeneration gas pressure on objective functions and tail91050001.4010000013525095594点3.6Regeneration gas temperature(C)11:66bffect of regenerated gas temperature on objective functions and1.51.61.71.81.92.02.1222.3242.54.1.5. Regeneration gas pressureRegeneration tower pressure(bar)125000Figure 6 presents effect of regeneration gas pressure on120000FCHA, ENet, RCirculation and carbon dioxide mole percentage ofsweet gas. According to Figure 6 minimum ENet, Roculation1150001 sory dioxide mole percentage of sweet gas are obtainedat low pressure of regeneration gas. But at low pressure,11000055methane molar flow of tail gas is increased4.1.6. Regeneration tower pressure40中国煤化工Figure 7 presents effect of regeneration tower pressure onCNMHGRco2, ENet, CIrculation, and carbon dioxide mole percentage Figure 7. Effect of regeneration tower pressure on objective functions and10o060as. As an industrial experience, stripping at lot9O929496981O0Regener at i on gas tenper at ureo QJournal of Natural Gas Chemistry VoL. 20 No. 1 201IOh. 0.1DE2/MDEA mass flow ratioTable 4. Effect of an increase in the decision variable on theobjective functionsDEA/MDEA mass composition ratio a eto 6pt i nratantisnet二吧A3可ME@场effect of this parameter is investigated on final pareto optimasolution sets of optimization problems. AlBestof threeAmine circulationscenarios are shown in figure 8fi tti ngte DEA+30%←MEA=09%10.0l120,0110opt i nal set—D20%—MEA=10%The objective functions were optimized to fulfil the con-Best fitting-DEA=20%-MDEA=10%straintse nSgaPareto optimal set-DEA=10%-MDEA=20%Best fitting. DEA=10%MDEA=20%tting DEA-20%solutionsA MATLAB code for real-parameter NSGA-II described in0.0106g0.0104opt irmalus sets e s era- o/um erM.20%mutation probability of 0. 1. All governing parameters of00102NSGA-ll are presented in Table 5. The different opera-tions were performed for 100 generations to obtain the non-0.0100dominated Pareto optimal solutions. The Pareto optimal solu-tion sets after 100 generations for three scenarios are shown in0.0098Tables 6-82.22.42.6283.03.23.43.63.84.04.2ulation rate x10 (kgmole/h)Table 5. Governing parameters of NSGA-llFigure 8. Effect of DEA/MDEA ratio on Pareto optimal solution setsMethod valNumber of decision variablesNumber of objectives4.2. Optimization resultsMaximum generationsThe optimization problem involves three objective funcReplace proportionOio01 04, ENet, and Circulation. Table 4 shows the effecnt pool sizeof variation in the decision variables on the objective functwo pointCrossover probabilitytions. It has been generated based on a parametric sensitivityMutation methodelectivealysis of the simulated mathematical model of the systemTable 6. Non-dominated Pareto optimal solutions after 50 generations for scenario(1)Ratio of i14,4125.6332021.9096.680.0120211.533.001.9529799.194.0533.24129694.7624826.7715,4434.6114926.7220040015,874.52999633.5514.651.5525608,9495.5189.757439.6115.4434.5714.9240516.639993155014.656.5936267,712,9825250.3512.8834.6514.651.8526712.160.0o94.143.7634.57149839034.893.86321836334.41149288.846.5339692.1514.652.003.938394.991634.6114.821,8133358,3799284.8714.7199130.009834.5714.926.72中国煤化工100.0213.7434.65CNMHG98.2522.212142372293.033231436+3234394.2Circul at i on rate xI 04( kgno106Alireza Behroozsarand et al Journal of Natural Gas Chemistry VoL. 20 No I 2011Table 6( Continue)Reg. FeedCireulation3.032.0025987.163.7212.9032.6113.2793,5319527067,233.55396.3934.5738226100.0033.2414.2294.7397531672.378θ36933.163.5533037,0934.6l93862.0035906.523.7415.5033.3114.6551.9834431.864.0012.8233476.531,9728967073.5929328.063.7534.9614.751.9737295.0732997234.6514.94.693.5828089.1397.638.7633.2413.7594.455.01.9827444.433695.191.8738570.8198.736797.024.3214.0332.5314.92938632320.1598.7891.677014.4893.8631292.394213.9834.764.5830860.943.7298.156233.2414.391.9629122,8132.1014.8233545.5560532.69149227790.744.4296.8134.57149235499.8812.5433.5514.615726320.643.9295.3234367.41Table 7. Non-dominated Pareto optimal solutions after 50 generations for scenario(2Ratio of FDEAFco2.8714.9634.8026947.9412.5823390.2894.9223727.084.3814.7134.8014.7132160.4814.7234.9614.9031487.723.6999.1934.5714.7532417.231213.5434.8415.0094.965.422.0028886.43.41984533.5914.9294.9215326123.494.268.841542295595594.921535929.241634.0614.9694.9215326483.14.5299.1334.960229400.414.4712.5834.141925051413.4397.6827701.401.0514.9694.96271878133.59150094.9228545,533.6215429228307.8530434.3714.9093.866.4533590.1099721214.961.7224742.203.6697.47126334.1494.9234.80中国煤化工6977414.9293.71CNMHGI34.0614.9632785393.83Journal of Natural Gas Chemistry VoL. 20 No. 1 201ITable 7( Continue)Ratio of FDEAF12.581.8575.3897.17154228652.0734.1494.143.001.9727410.6514.7134.804544429985.710566.573732.339294.926.221947568.343.5398.02349214.8889.9433210.7434.7628382.124.4134.0614.9626693.944.6093866.3330153,222,8934.181.5328754.864999.174332.1414.9294.926.2219430981.704.0399.332.3344444001.9728067,3231.6311.2494.293.021.7524481823.8215423964.94.0794.9225595.1698.27349214.8894926.4731313,453,57989712.5834.1494.613.35Table 8. Non-dominated Pareto optimal solutions after 50 generations for scenario(3)Pop. No. Ratio of FDEA/Fco, TFeed PFeed14.514.6194.061.9726204.4393.6934.923.2224412.114.2494.8394.69234095034.9214.923.1327654.043.6396.2334.4114.822965423.9692.7823542.73349214.7594,416928801,9125101.5294.27154194.8817326965.304.1528943,7297.2114.8433.6793.7825667.454.7632.373.03238588993.1394.731.6127920.543.7696.9714.7519327191.273.7412.86324514.8294.8819824699.303.5894.0712711433456634.2994.7627484.103.6296.0614.5994.8819827260.2996.0314.826314.8815.5094.9613.3433.5114.894.764802.8314.8224930.763.3594.10129234.9614.7534.926000252680424220.513.5728072.843.5594.101.5223714.714.4334.8814.751.6225552.704.295.6914.8225581.5694.7414.7594.6527323.7296.1434.9614.9894.8828002,2296949625383.7532.3314.9094.9293.7813.6234.8414.76中国煤化93.63CNMHG3934.3714.7694.733.246859514.49966l4.5108Alireza Behroozsarand et al. Journal of Natural Gas Chemistry Vol, 20 No. I 2011Table 8( Continue)Ratio of FDEAFco,4.9614.751873.66232690814.8494.733.2226383.493.5831.8614.823.433.0327596.4034.9214.92560626627,283.5634.5314.821.9824393.243.51934134.0614.7823070.014.7834.9214.7694.7325048.4595.3313.329398Figure 8 shows the Pareto set of optimal solutions obfunctions value. Figure ( a)shows the effect of DEA/MDEAined for the problem formulated above with considering mass composition percentage ratio on amine solution circu3.0 DEA/MDEA ratio as decision variable. The values of ENet, tion rate. By increasing MDEA mole fraction of amine soRCirculation and Rco, are plotted. The points in the Pareto set lution, circulation rate is decreased. Figure 9(b) represents(Figure 8)indicate the minimum possible of ENet, RCirculation the effect of DEA/MDEA mass composition percentage ratioor maximum possible of Rco, with the given operating con- on energy consumption. It is clear that by increasing ratio ofstraints. The benefit of a multiobjective optimization is evi- DEA/MDEA in amine solution, the net energy consumptionsT 2 dent upon observing the wide choice of operating points avail- is decreased. This is because according to the previous figure,le in the Pareto-optimal set. The cpu time taken to generate circulation rate is decreased when ratio of dEA/MDEa is inone set of Pareto-optimal solutions(such as those in Figure 8) creased. o However, Figure 9(c)shows that increasing ratio ofis 100 min on the Pentium(R)4 CPU 3.00GHz, 512 MB of dEA/MdeA has unsuitable effect on CO 2 recovecEA=10%AM computerIt seems that DEa= 20%(Mw%) and mo The final optimal set results of NSGA-ll algorithm(TaMW%)are the desired cebles 6-8)have been sorted smallest to largest of objective in special applications there may be some objective functionsDEA=0%DEA-20%-MDEA=10%EA=10%DEA-10%-MDEA-20DEAIEA=20%4.0Population numberPopulation numberDEA30%0-MDEA0%DEA=20%-MDEA=10%DEA-30%-MDEA-0%DEA=10%-MDEA=20%A=10%H中国煤化工CNMHGFigure 9. Comparison of three objective functions in three scenarioPopul ation nunberJournal of Natural Gas Chemistry VoL. 20 No. 1 201Ithat are more important than the others or some industrial con- Latin lettersstraints may exist. For example, CO2 recovery must be higher DAbsorber Absorber tower diameter(than 97%. In these cases user may select the practical optimal DRegenerator Regenerator tower diameter(m)solution sets that satisfy the industrial constraintsENNet plant heat load(kJ/h)Figure 9(d)represents the comparison of Pareto optimaMolar flow of streams(kgmole/h or lbmole/h),sets of three scenarios. According to this figure, Scenarios(2)i=CH4, CO2, DEAand (3) have lower circulation rate than the Scenario(1). Onthe other hand, Scenario(2)has the highest CO2 recoveryLAbsorber Absorber tower length(m)As mentioned above, it seems that DEa= 20%(Mw%)LRegenerator Regenerator tower length(m)and MDEa=10%(MW%) are the better choice for operating condition of amine plant. Therefore, three sets of soluLs pressure(bar)tions are selected as optimal solutions from Table 7(RelatedPRegenerator Regeneration tower pressure(bar)to DEa= 20%(Mw%)and MDEa= 10%(Mw%) scenario Q Condunser Duty of condenser(kJ/h)shown in Table 10). In the solution set Number l, the plant QrDuty of pump(kJ/energy consumption is minimized, in the solution set Number REboiler Duty of reboiler(kJ/h)2, the amine circulating rate is minimized and in solutionirculationNumber 3, the carbon dioxide recovery is maximizedCarbon dioxide recovery(%)urgasSour gas temperature(C)5. ConclusionsTReg gas Regeneration gas temperature (C)A阳 sweetening plant was simulated and optimized by Gree体such as amine solution circulation rate, CO2 mole percent ofP Densweet gas, and plant energy consumption affect the processeconomy. The simulation results have good agreement withexperimental data. By analyzing process parameters, trendReferencesd constraint of adjustable variables have been obtained. Thecirculation ratea nd ro And minimizing the energy load, amine [2] Datta A K, Sen PK. J Membr Sci, 2006, 283(1-2): 290,5096process has been optimized for obtaining CO2 mole fraction in [1 Abdi M A, Meisen A Ind Eng Chem Res, 1999, 38(8)the sweet gas below 2ased methane flow results show that[3] Wauquier J P Petroleum Refining: Separation Processes. Parisre and pof 33.98C and 14.96 barEditions Technip, 2000DEA/CO2 molar flow ratio of 12.58, regeneration gas temper[4] Dortmundt D, Doshi K In: Recent Development in CO2 Reature and pressure of 94.92C and 3.0 bar, regenerator presmoval Membrane Technology, UOP LLC Des Plaines: Illinoissure of 1, 53 bar and ratio of dea/mdea= 20%0/10%o are thebest values for minimizing plant energy consumption, amine[5] Geankoplis C J. Transport Processes and Unit Operations. Newcirculation rate and carbonJersey: Prentice Hall, 1993dioxide recovery.[6] Bhide b d, voskericyan A, Stern S A. J Membr Sci, 1998Nomenclature140(1):27[7] Khakdaman H, Zoghi A, Abedinzadegan M A. Petrol Tech 2.Acronyms2005,10(5):113Abs[8] Unsford M, Bullin A. In: Proceedings of the 1996 AIChE SpringDEa DiethanolamineNational Meeting. New YorkDGA D[9 Kohl A L, Riesenfeld F C. Gas Purification. 4th Ed. Texas: GulfPublishing ceIPa Di-propanolamine[10 Harris S D, Elliott L, Ingham D B,Fischer -TComput Meth Appl Mech Eng, 2000, 190(8-10): 1065[11] Konak A, Coit D W, Smith A E1ggtL Gas to liq91(9)E. Deb K. thiele l, EvolaMDEA Methyldiethanolamine8(2):173[13] Davis B H. Catal Today, 2003, 84(1-2)[14] Donnelly T J, Satterfield C N. Appl Catal, 1989, 52(1)NSGA Non-sorting genetic algorithm[15] Eliason S A, Bartholomew C H App/ Catal A, 1999, 186(1-2)PRSV Peng-robinson-stryjek-verRE Relative error[6」 Madon r j,Ta中国煤化工3[I7| Raje A P, daHHCNMHG