15
Optimization and control for CO 2 compression and purication unit in oxy-combustion power plants Bo Jin, Haibo Zhao * , Chuguang Zheng State Key Laboratory of Coal Combustion, Huazhong University of Science and Technology, Luoyu Road 1037, Wuhan, Hubei, 430074, People's Republic of China article info Article history: Received 7 September 2014 Received in revised form 4 January 2015 Accepted 13 February 2015 Available online 11 March 2015 Keywords: CO 2 compression and purication unit Dynamic simulation Control system design Oxy-fuel combustion abstract High CO 2 purity products can be obtained from oxy-combustion power plants through a CO 2 CPU (compression and purication unit) based on phase separation method. To ensure that CPU (with double ash separators) can be operated under optimal conditions, this paper focuses on single variable analysis, multi-variable optimization, dynamic simulation and control system design for CPU in oxy-combustion power plants. It is found that optimal operating conditions are 30 bar, 30.42 C, 24.64 C, and 55 C for multi-stage CO 2 compressor discharge pressure, ue gas temperature after compression, rst ash separator temperature, and second ash separator temperature, respectively. The designed double temperature control structure based on a systematic top-down analysis and bottom-up design method is more suitable than the single temperature control structure under different operating scenarios (load change and ue gas composition ramp change). To bear operating disturbances, operating strategies like elevating temperature, manipulating valves and adjusting setpoints are proposed. Inuence of SOx would be more obvious than that of NOx, whilst the k ij mixing parameters in PengeRobinson property method affects little on process optimization and control system design. Comprehensive dynamic model with specied control system provides possibility to integrate CPU with full-train oxy-combustion power plants. © 2015 Elsevier Ltd. All rights reserved. 1. Introduction To mitigate the greenhouse effect on climate change, CCS (CO 2 capture and storage) technology is proposed as an attractive method for restraining anthropogenic CO 2 emissions from power plants. Generally, CCS can be divided into three categories: post- combustion, pre-combustion (i.e. integrated gasication com- bined cycle, IGCC), and oxy-fuel combustion. Different from other CO 2 capture technologies that used in post-combustion and pre- combustion such as absorption, adsorption, and membrane sepa- ration, oxy-fuel combustion CO 2 capture with CO 2 CPU (compres- sion and purication unit) is a CO 2 enrichment approach [1,2]. The basic idea of CPU is to remove other impurities in ue gas to obtain high purity CO 2 products. Impurities, which depend on power plant conguration and operation [3], may include inert gas (O 2 , Ar, N 2 ), H 2 O, SOx, NOx and other trace elements such as COS, H 2 S, HCN, NH 3 , volatile traces of minerals, Hg and other metals [4]. The presence of these impurities may bring deteriorated effects on operation and equipment. Higher NOx and SOx concentration could contribute to serious corrosion on equipment, external H 2 O may lead to ice formation in pipes and tubes, and higher O 2 content may cause: 1) overheating at the injection point, 2) increased extraction cost, and 3) unknown effects on oil production if for EOR (enhanced oil recovery) application, respectively [4]. To avoid these adverse effects, relevant specication [4e7] for CO 2 products from CPU has been established based on the requirements of environ- mental regulation, cost, technical demands (transport, storage, etc.), and applications. It is widely recognized that the minimum CO 2 product purity exhausting from CPU should be 95 vol.% or above [4,5]. For processing ue gas to meet these requirements, various CPU schemes without or with different purication units were proposed. Three ue gas purication schemes including no purication, partial condensation (cold box), and cold box including distillation were developed by Air Liquide [8]; sour compression process for co-removal of SOx and NOx through re- actions between these components was presented by Air Products [9]; and other three technologies consisting of sulfuric acid process (suitable for ue gas containing high levels of SOx), activated * Corresponding author. Tel.: þ86 027 8754 2417 8208; fax: þ86 027 8754 5526. E-mail address: [email protected] (H. Zhao). Contents lists available at ScienceDirect Energy journal homepage: www.elsevier.com/locate/energy http://dx.doi.org/10.1016/j.energy.2015.02.039 0360-5442/© 2015 Elsevier Ltd. All rights reserved. Energy 83 (2015) 416e430

Optimization and control for CO2 compression and ...cpc.energy.hust.edu.cn/__local/F/29/E2/059D2714F... · Optimization and control for CO2 compression and purification ... system

  • Upload
    lamdang

  • View
    221

  • Download
    1

Embed Size (px)

Citation preview

Page 1: Optimization and control for CO2 compression and ...cpc.energy.hust.edu.cn/__local/F/29/E2/059D2714F... · Optimization and control for CO2 compression and purification ... system

lable at ScienceDirect

Energy 83 (2015) 416e430

Contents lists avai

Energy

journal homepage: www.elsevier .com/locate/energy

Optimization and control for CO2 compression and purificationunit in oxy-combustion power plants

Bo Jin, Haibo Zhao*, Chuguang ZhengState Key Laboratory of Coal Combustion, Huazhong University of Science and Technology, Luoyu Road 1037, Wuhan, Hubei, 430074,People's Republic of China

a r t i c l e i n f o

Article history:Received 7 September 2014Received in revised form4 January 2015Accepted 13 February 2015Available online 11 March 2015

Keywords:CO2 compression and purification unitDynamic simulationControl system designOxy-fuel combustion

* Corresponding author. Tel.: þ86 027 8754 2417 82E-mail address: [email protected] (H. Zhao)

http://dx.doi.org/10.1016/j.energy.2015.02.0390360-5442/© 2015 Elsevier Ltd. All rights reserved.

a b s t r a c t

High CO2 purity products can be obtained from oxy-combustion power plants through a CO2 CPU(compression and purification unit) based on phase separation method. To ensure that CPU (with doubleflash separators) can be operated under optimal conditions, this paper focuses on single variable analysis,multi-variable optimization, dynamic simulation and control system design for CPU in oxy-combustionpower plants. It is found that optimal operating conditions are 30 bar, 30.42 �C, �24.64 �C, and �55 �C formulti-stage CO2 compressor discharge pressure, flue gas temperature after compression, first flashseparator temperature, and second flash separator temperature, respectively. The designed doubletemperature control structure based on a systematic top-down analysis and bottom-up design method ismore suitable than the single temperature control structure under different operating scenarios (loadchange and flue gas composition ramp change). To bear operating disturbances, operating strategies likeelevating temperature, manipulating valves and adjusting setpoints are proposed. Influence of SOxwould be more obvious than that of NOx, whilst the kij mixing parameters in PengeRobinson propertymethod affects little on process optimization and control system design. Comprehensive dynamic modelwith specified control system provides possibility to integrate CPU with full-train oxy-combustion powerplants.

© 2015 Elsevier Ltd. All rights reserved.

1. Introduction

To mitigate the greenhouse effect on climate change, CCS(CO2 capture and storage) technology is proposed as an attractivemethod for restraining anthropogenic CO2 emissions from powerplants. Generally, CCS can be divided into three categories: post-combustion, pre-combustion (i.e. integrated gasification com-bined cycle, IGCC), and oxy-fuel combustion. Different from otherCO2 capture technologies that used in post-combustion and pre-combustion such as absorption, adsorption, and membrane sepa-ration, oxy-fuel combustion CO2 capture with CO2 CPU (compres-sion and purification unit) is a CO2 enrichment approach [1,2]. Thebasic idea of CPU is to remove other impurities in flue gas to obtainhigh purity CO2 products. Impurities, which depend on powerplant configuration and operation [3], may include inert gas (O2, Ar,N2), H2O, SOx, NOx and other trace elements such as COS, H2S, HCN,NH3, volatile traces of minerals, Hg and other metals [4]. The

08; fax: þ86 027 8754 5526..

presence of these impurities may bring deteriorated effects onoperation and equipment. Higher NOx and SOx concentrationcould contribute to serious corrosion on equipment, external H2Omay lead to ice formation in pipes and tubes, and higher O2 contentmay cause: 1) overheating at the injection point, 2) increasedextraction cost, and 3) unknown effects on oil production if for EOR(enhanced oil recovery) application, respectively [4]. To avoid theseadverse effects, relevant specification [4e7] for CO2 products fromCPU has been established based on the requirements of environ-mental regulation, cost, technical demands (transport, storage,etc.), and applications. It is widely recognized that the minimumCO2 product purity exhausting from CPU should be 95 vol.% orabove [4,5]. For processing flue gas to meet these requirements,various CPU schemes without or with different purification unitswere proposed. Three flue gas purification schemes including nopurification, partial condensation (cold box), and cold boxincluding distillation were developed by Air Liquide [8]; sourcompression process for co-removal of SOx and NOx through re-actions between these components was presented by Air Products[9]; and other three technologies consisting of sulfuric acid process(suitable for flue gas containing high levels of SOx), activated

Page 2: Optimization and control for CO2 compression and ...cpc.energy.hust.edu.cn/__local/F/29/E2/059D2714F... · Optimization and control for CO2 compression and purification ... system

List of abbreviations

AD Aspen Plus DynamicsAP Aspen PlusASU air separation unitC compressorCC_CO2 CO2 product purity controllerCC_CRR CO2 recovery rate controllerCCS CO2 capture and storageCPU compression and purification unitCO2_CO2 CO2 flow rate in CO2 productDTC double temperature controlDT_CO2 dead time for CO2 product purity controllerEOR enhanced oil recoveryF1,2 flash separatorsFC_CO2 CO2 product flow controllerFC_FG flue gas flow controllerFG_CO2 CO2 flow rate in raw flue gasIEAGHG International Energy Agency Greenhouse Gas

IGCC integrated gasification combined cycleLC_F1, F2 level controllers for F1 and F2LCV114, 119 throttle valvesMCC main CO2 compressorMHXs (E1, 2) multi-stream heat exchangersPID proportioneintegrationedifferentiationPV process variableRGA relative gain arraySP setpointSQP sequential quadratic programmingSTC single temperature controlTC_Cooler temperature controller for coolerTC_F1, F2temperature controllers for F1 and F2TC_MCC temperature controller for flue gas after compressionT_S18 S-18 temperatureT_S8 S-8 temperatureV-CO2 CO2 product valveVPSA variable pressure swing absorptionV-VENT vent gas valve

B. Jin et al. / Energy 83 (2015) 416e430 417

carbon process (for lower amounts of SOx), and VPSA (variablepressure swing absorption) process cycle were considered byPaxair [10]. Compared with these options, CPU using partialcondensation method with two flash separators is chosen here foroxy-combustion application since it is an auto-refrigerated processwith characteristics of less power consumptions, lower capitalcosts and easily for operation.

In order to identify its thermodynamic performance, severalprocess simulation studies have been presented. On the steady-state process simulation front, most researches have mainlyfocused on sensitivity analysis, energetic evaluation, and techno-economic assessment. Pipitone and Bolland [4] compared the en-ergy consumption, CO2 product purity and CO2 recovery rate of twoCPU configurations (flash separation and distillation) for naturalgas or pulverized fuel oxy-fuel combustion power plant usingsimulation tool SIMSCI PRO/II. Posch and Haider [7] also simulatedand compared two different CPU configurations in Aspen Plus. Togain impact of main design parameters on performance features,detailed sensitivity analysis was conducted through varying in fluegas composition, plant load and kij mixing parameters (here, itmeans binary parameters in PengeRobinson equation of state toidentify binary interactions between different components). It wasfound that double flash separation unit was with lower power andless cooling requirement, whilst kij mixing parameters were crucialfor calculation of CO2 product purity and CO2 recovery rate. Tofurther improve thermodynamic performance of CPU and reduceits specific energy consumption, energetic evaluation of a CPUplant based on six conceptual configurations was carried out [11].The results indicated that the process with two turbines and twoflash separators was a better conceptual configuration. In addition,taking one-stage flash separation unit as a base case, double-stageflash and three-stage flash separation units were simulated andcompared through process simulation in Aspen Plus and pinchanalysis in PRO_PII [12]. It was found that a better energy perfor-mance can be obtained by adding one flash stage, whereas it mightnot be attractivewhen investment cost is considered. Thus, techno-economic assessment [13] should be conducted to screen moresuitable process configuration for CO2 purification. The resultsshowed that two-stage flash process is the most cost effectivealternative configuration. Although these studies provided goodfoundations for understanding CPUs in oxy-combustion powerplant, how to obtain its optimal operating condition through

optimization method is still limited. And, the proposed approacheswere mainly involved in single variable optimization [7,14].Factually, multi-variable optimization can help to provide a goodinitial condition for dynamic simulation and develop suitablecontrol system to maintain CPU operating around optimalcondition.

Apart from steady-state process simulation mentioned above,very limited studies have been conducted for dynamic processsimulation of CPU. A dynamic CO2 CPU model was established incommercial system UniSim® to improve process design and con-trol, and its robustness was tested by applying disturbances andequipment failures [15]. Another CPU dynamic model with controlsystem developed in Aspen Hysys was ready for integrating CPUmodel into a comprehensive dynamic model of oxy-combustionpower plant [16]. Recently, a mathematical dynamic CPU modelwas implemented in gPROMS to investigate its transient behaviorin response to process disturbances (discharge pressure ofcompressor, flue gas inlet temperature, and flue gas composition)and provide useful information for control system design [17].Before establishing this mathematical model, the control strategyfor CPU was also investigated [18]. The RGA (Relative Gain Array)analysis was applied to identify the most promising control strat-egy, and then uncertain RGA analysis were used to consider processnonlinearities. However, the detailed control structure was notprovided, configured and tested in the dynamic model. Thus,control system design and its related controllability analysis basedon dynamic model is still required to further investigate the dy-namic behavior of CPU or even a full-train oxy-combustion powerplants.

The paper aims to optimize process operation and design acontrol system for CPU, which is organized as follows. In Section 2,the steady-state model for single variable analysis and multi-variable optimization are established in Aspen Plus (AP, version7.1), and corresponding dynamic model for control system design isestablished in Aspen Plus Dynamics (AD, version 7.1). Then,detailed CPU control design procedures, based a systematic top-down analysis and bottom-up design method, is presented inSection 3. Section 4 discusses the transient performance of CPUbased on designed control system through dynamic tests underload change process and planned disturbance. In Section 5, alter-native control strategy, impacts of SOx and NOx levels, and un-certainty analysis for the kij mixing parameters in PengeRobinson

Page 3: Optimization and control for CO2 compression and ...cpc.energy.hust.edu.cn/__local/F/29/E2/059D2714F... · Optimization and control for CO2 compression and purification ... system

Table 1Raw flue gas condition.

B. Jin et al. / Energy 83 (2015) 416e430418

property method are discussed. Finally, conclusions are given inSection 6.

Component (mol.%) Value

CO2 82.40O2 4.19N2 9.50Ar 3.90CO (ppm) 31SO2 (ppm) 14SO3 (ppb) 83NO (ppm) 20NO2 (ppb) 38Mass flow rate/t/hr 717.19Temperature/�C 50Pressure/bar 1.1

Table 2The kij mixing parameters [7].

Species 1 Species 2 kij value

CO2 Ar 0.123CO2 O2 0.116CO2 N2 �0.0115CO2 SO2 0.0559

2. Model implementation and optimization

2.1. Steady-state model

The unit configuration is based on the prototype proposed bythe IEAGHG (International Energy Agency Greenhouse Gas) R&Dprogramme study [19], as shown in Fig. 1. Raw flue gas is com-pressed by MCC (multi-stage CO2 compressor) with intercoolers to30 bar, and then sent into cold box for gaining 96 mol.% or higherCO2 products. For cold box, two MHXs (multi-stream heat ex-changers) and two flash separators are configured. The pretreatedflue gas is cooled to �24.51 �C and then sent to the first flashseparator for separating CO2 from the impurities. The bottomstream containing about 96 mol.% or more purity of CO2 is the firstflow of CO2 products, whilst the top stream is cooled continuallyto �54.69 �C before it enters into the second flash separator inwhich the rest of CO2 is obtained in the bottom as the second flowof CO2 products. The vent gas from the top of the separator togetherwith the two CO2 products is reheated in MHXs to be about 40 �C.At last, two CO2 flows are mixed and compressed for storage orutilization.

To establish steady-state model, flue gas condition, model andproperty method chosen are identified. Raw flue gas condition asinput, which is from a conceptual 600 MWe oxy-combustion pul-verized-coal-fired boiler [20e23], is presented in Table 1, whereimpurities like Ar, O2, N2, SOx (SO2 and SO3), NOx (NO and NO2),and CO are considered. For model chosen, MCC is represented bymulti-stage compressor model to moderate system pressure anddrive flue gas, MHXs (E1 and E2) byMheatXmodels to identify heattransfer in cold box, and two flash separators (F1 and F2) by Flashmodels to reflect CO2 separation process, respectively. The Pen-geRobinson equation of state is adopted here since it is consideredas the most appropriate equation for CPU [7], and its correspondingmixing parameters kij, which has great effects on CO2 product

Fig. 1. Process flow diagram of CO2 co

purity [7], are listed in Table 2. In addition, simulation results for allstreams are presented in Appendix A.

2.2. Single variable analysis

In order to investigate the effects of operating parameters onsystem performance and provide basic information for optimiza-tion, single variable analysis is first carried out. The consideredvariables include MCC discharge pressure, F1 temperature, and F2temperature, whilst CO2 product purity, CO2 recovery rate, specificpower consumption and specific cooling duty are used as evalu-ating indicators. Here, specific power consumption is defined as theratio of total power consumption including power consumption in

mpression and purification unit.

Page 4: Optimization and control for CO2 compression and ...cpc.energy.hust.edu.cn/__local/F/29/E2/059D2714F... · Optimization and control for CO2 compression and purification ... system

B. Jin et al. / Energy 83 (2015) 416e430 419

MCC and C to total CO2 recovered, specific cooling duty as the ratioof total cooling duty consumed in MCC and Cooler to total CO2recovered, and CO2 recovery rate as the ratio of total CO2 recoveredto that in raw flue gas, respectively. They are formulated as

Specific power consumption ¼ Total power consumptionTotal CO2 recovered

(1)

Specific cooling duty ¼ Total cooling dutyTotal CO2 recovered

(2)

CO2 recovery rate ¼ Total CO2 recoveredTotal CO2 in raw flue gas

(3)

Fig. 2 illustrates the influences of each single operating param-eter on these four factors. When MCC discharge pressure increases

Fig. 2. System performance under variations of three single operating parameters. (a) MC

from 28 bar to 33 bar whilst other parameters are unchanged, CO2product purity is dropped by 0.71% while CO2 recovery rate isimproved by 1.11% (Fig. 2 (a), left). The increase of CO2 recovery rateis derived from the increase in liquid CO2 recovered from F1 islarger than the decrease in liquid CO2 recovered from F2. Mean-while, specific power consumption and specific cooling duty gain aminimum value within this pressure range, which are determinedby total CO2 recovered and their total energy consumption (totalpower consumption and total cooling duty). From Fig. 2(b),although CO2 product purity gets a maximum value whilst CO2recovery rate has a minimum value, they vary within a small range.Specific power consumption and specific cooling duty increasesimultaneously with the increase of F1 temperature from �40 �Cto �15 �C. With respect to the variation of F2 temperature, thetendencies for variations of CO2 product purity and CO2 recoveryrate are operated in an opposite way with those in MCC discharge

C discharge pressure change. (b) F1 temperature change. (c) F2 temperature change.

Page 5: Optimization and control for CO2 compression and ...cpc.energy.hust.edu.cn/__local/F/29/E2/059D2714F... · Optimization and control for CO2 compression and purification ... system

Table 4Comparison of energy performance between optimization case and base case.

Item Base case Optimization case

Specific cooling duty kWh/tCO2 125.88 126.66Specific power consumption kWh/tCO2 116.16 115.86CO2 recovery rate 94.11% 94.20%

Table 5Comparison of exergy analysis results for two cases.

Item Base case Optimization case

ED, kW yD, % ε, % ED, kW yD, % ε, %

Valves 673.79 0.47 2.18 667.79 0.47 2.17Compressors 27023.62 18.91 87.56 27094.45 18.98 88.02Flash separators 0 0 0 0 0 0Heat exchangers 2874.79 2.01 9.31 2825.53 1.98 9.18Others 291.84 0.20 0.95 193.95 0.14 0.63First flash separation 1806.84 1.26 5.85 1777.01 1.24 5.77Second flash separation 1576.13 1.10 5.11 1551.53 1.09 5.04Total exergy destruction 30864.03 30781.28Total exergy loss 15108.60 12442.24Total fuel exergy 142893.16 142785.44Exergy efficiency, % 67.83 69.73

B. Jin et al. / Energy 83 (2015) 416e430420

pressure change, whilst specific power consumption and specificcooling duty gain similar behavior with those in F1 temperaturechange.

In oneword, these operating parameters have specific effects onsystem performance, and similar results were obtained in literature[7,13]. For CO2 product purity and CO2 recovery rate, their behavioris presented in an opposite way under three single factor variations.CO2 recovery rate is more sensitive to F2 temperature change thanchanges in MCC discharge pressure and F1 temperature, because18.05% decrease is found in F2 temperature change, compared with0.5% increase in MCC discharge pressure change and 0.22e0.35%variation in F1 temperature change. With respect to specific powerconsumption and specific cooling duty, they are affected moresignificantly by temperature in flash separators thanMCC dischargepressure. Furthermore, the increase of temperature in flash sepa-rators will make adverse effects on performance of energyconsumption.

2.3. Multi-variable optimization

To obtain optimal operating conditions for CPU, multi-variableoptimization process [24] is conducted. The core methodology isto obtain a minimum or maximum value calculated from objectivefunction through varying multiple objective variables within theircorresponding ranges and setting multiple constraints simulta-neously. Here, the objective function can be formulated to mini-mize the specific power consumption for CPU. As presented in Eq.(1), it means that maximizing the amount of CO2 recovered fromraw flue gas whilst minimizing total power consumption requiredin MCC and C. In Table 3, four objective variables and three oper-ating constraints are specified, which are the most concernedoperating parameters studied in literature [4,7,13]. CO2 productpurity and CO2 recovery rate should be constrained to be greaterthan their minimum values (here, they are set as 96 mol. % and90%). Meanwhile, S-8 temperature should be greater than CO2 tri-ple point temperature (i.e. �56.57 �C) to avoid CO2 solidification.The multi-variable optimization process is accomplished using abuilt-in SQP (gradient-based) local solver, with optimization andconstraint functions in AP. Finally, the optimal operating parame-ters including MCC discharge pressure, MCC outlet temperature, F1temperature, and F2 temperature are attained after three iterationsfor convergence, as shown in Table 3.

Performingmulti-variable optimization runs results in favorablesolutions that cannot be achieved by a single variable analysis.Comprehensive energy and exergy comparisons between optimi-zation case and base case are presented in Tables 4 and 5 to identifyand validate the effectiveness of multi-variable optimization pro-cess. As listed in Table 4, optimization case gets approximate 2.58%decrease of specific power consumption and 0.09 percent increaseof CO2 recovery rate even though its specific cooling duty increasesabout 0.78 kWh/tCO2.

Table 3Objective variables, operating constraints, and optimal operating conditions.

Number Variable Range Optimal operatingconditions

Objective variables1 MCC discharge pressure, bar 28 e 33 302 MCC outlet temperature, �C 25 e 50 30.423 F1 temperature, �C �30 to �20 �24.644 F2 temperature, �C �55 to �40 �55Operating constraints1 CO2 product purity, mol.% �96 96.912 CO2 recovery rate, % �90 94.203 S-8 temperature, �C >�56.57 �55.60

For simplicity, the exergy analysis considered here introducesexergy destruction (ED), exergy destruction ratio (yD), exergydestruction distribution (ε), and exergy efficiency (h) as parametersfor evaluating thermodynamic performance. Detailed proceduresfor conducting exergy analysis can be found in recent references[25,26] and Appendix B. Exergy destruction, also called as internalexergy loss, arises from the irreversibility occurring in a system orunits under thermodynamic or techno-economic constraints.While exergy loss (EL), related to the external exergy loss, meansexergy that transferred from system component to ambientbecause of heat loss, cooling, byproducts, ash and wastewater, etc.In addition, yD signifies how much one process or device contrib-utes to the total fuel exergy (EF) of the system, ε reflects the exergydestruction distributed in system, and h provides a true measure ofsystem performance from the thermodynamic viewpoint, respec-tively. These parameters can be formulated as below [25,26].

yD ¼ EDEF;tot

(4)

ε ¼ EDED;tot

(5)

h ¼ 1� ED þ ELEF

(6)

Table 5 presents exergy analysis results (for more details, seeAppendix B). From the base case, it is found that the largest exergydestruction is caused by enormous power consumption occurred incompressors while flash separators gets minimal value becauseonly physical separation is processed. Compared to the base case, itgets a more efficient thermodynamic performance after using theselected optimization method since exergy efficiency for the opti-mization case is 1.9 percent higher than that of the base case. It isattributed to the decrease of ED in equipment except for compres-sors which is related to temperature and power consumption inMCC and C. Despite all these, it gets similar exergy destructiondistribution for equipment and system. Generally, these resultsindicate that the optimization approach is reasonable and effectivefor improving the CPU performance.

Page 6: Optimization and control for CO2 compression and ...cpc.energy.hust.edu.cn/__local/F/29/E2/059D2714F... · Optimization and control for CO2 compression and purification ... system

B. Jin et al. / Energy 83 (2015) 416e430 421

2.4. Dynamic model

The optimization case is selected for dynamic simulation. Beforeexporting to AD for dynamic simulation, some preparations shouldbe conducted. Firstly, type of dynamic simulation should bedetermined according to the purpose of research. In this software,flow driven and pressure driven are two main types for dynamicsimulation. Pressure-driven modeling approach is selected for thisdynamic simulation to identify more rigorous dynamic perfor-mance [20]. Then, in order to represent dynamic characteristics andcontrollability of equipment, some approximate methods [27,28]are used to estimate the geometric sizes of equipment, and someconnections (valves, pumps and compressors) are added. In thecurrent model, an approximate volume for the inlet and outlet on agiven side of heat exchangers can be calculated from Eq. (7) [27]whilst diameter for flash separators can be estimated from “F-Factor” method [28].

V ¼ tR � vSS (7)

where V means volume, tR represents residence time, and vSS il-lustrates steady state volumetric flow rate. It should be mentionedthat these parameters are not real size but estimated from thesteady-state simulation results. If data on the dynamics of plantsare given, these volumes can be adjusted to give a good match tothe dynamic simulation results. The required equipment sizing issummarized in Table 6. For connections, compressor C is used toelevate pressure of stream S-10 to make it equal to that of stream S-19 whilst cooler is use to cool down temperature of stream S-11after compression. From now on, the dynamic model without anycontrols is established in AD and ready for dynamic simulation.

3. Control system design

To avoid possible risks introduced by unexpected disturbances,a closed loop control should be designed and set in the dynamicmodel. A systematical top-down analysis and bottom-up designmethod [29,30] and relevant control concepts [18] are used todesign a basic control system for this CPU model. Here, top-downanalysis includes setting control objectives, defining manipulatedand measured variables, and determining production rate, whilstbottom-up design consists of designing basic control loops likeflow, pressure and level in regular control layer and consideringcomposition, temperature and ratio control loops in supervisorycontrol layer. In order to gain favorable operation for CPU, threecontrol objectives are suggested to be reached. The primary goal isto meet product quality requirement, which means CO2 productpurity must be maintained around its SP (setpoint). As mentionedin Section 2, the second objective is that S-8 temperature must begreater than CO2 freezing point. Additionally, as an optional target,the third one is to maintain CO2 recovery rate at its initial value orabove 90%.

Inputs and outputs, which can be determined via degree offreedom analysis that expressed as the difference between dynamic

Table 6Main equipment sizing.

Model Mode Parameter Value

MCC InstantaneousE1 Dynamic Volume 76 m3/0.6 m3/2.4 m3/11 m3

(S-18/S-14/S-9/S-2)F1 Dynamic Diameter/length 6 m/12 mE2 Dynamic Volume 3.7 m3/0.6 m3/14 m3/86 m3

(S-4/S-13/S-6/S-8)F2 Dynamic Diameter/length 6 m/12 m

or control degrees of freedom and the number of inputs and out-puts with no steady-state effect, are presented in Table 7 to identifymanipulated and measured variables. Here, all parameters will berecalculated after reinitiating for gaining an initial state whensteady-state model is imported for dynamic simulation.

Production rate should be determined because it has significanteffects on control structure [31]. Typically, there exist two criteriafor setting production rate. First, production rate is set by the flowrate of product stream leaving the process. Second, production rateis set by the flow rate of a fresh feed stream entering the process,which is selected in this study because it connects with the up-stream unit operation.

Regulatory control layer is designed to prevent the plant driftingaway from its SP and resist local disturbance which can help su-pervisory control layer handle the effects of disturbances on theprimary outputs. Basic and simple control loops are set in regula-tory control layer. Raw flue gas flow rate control (FC_FG) can berealized by manipulating brake power of MCC, which will makesystem pressure floating. Due to the variations of MCC dischargepressure, MCC outlet temperature should be regulated to form atemperature control loop (TC_MCC) through adjusting cooling dutyin MCC. Liquid level in flash separators ought to be controlledthrough two level controllers (LC_F1 and LC_F2) in which controlactions are applied to their corresponding throttle valves (LCV114and LCV119) since these two parameters vary unsteadily duringoperation. As an influencing factor for CO2 recovery rate, CO2product flow rate is maintained under flow controller (FC_CO2) bymanipulating product valve position (V-CO2), whilst system pres-sure is maintained under system pressure controller (PC_F) actingon vent gas control valve (V-VENT) to help maintain the target ofCO2 product purity.

In the supervisory control layer, two temperature control loops,two composition control loops, and a ratio control loop areinstalled. As suggested in literature [32] and mentioned in Section2, S-8 temperature must be controlled within bounds to avoid CO2solidification. Actually, the temperature is determined by refriger-ation capacity generated by two throttle valves (LCV-114 and LCV-119). Therefore, two temperature control loops (TC_S8 and TC_S18)for S-8 and S-18 streams are considered in the system via twocascade controllers acted on these two valves, respectively. Mostimportantly, CO2 product purity is maintained around SP through acascade controller (CC_CO2) where dead time (DT_CO2) is consid-ered in the primary composition control loop and secondary con-trol action is sent to system pressure controller for manipulatingsystem pressure. To reach the third goal that keeping CO2 recoveryrate above 90%, a cascade and feedforward control loop (CC_CRR) isalso set in the system. Its input signal comes from the flow rates andCO2 concentration of CO2 product (S-21) and raw flue gas (S-1)streams are measured to calculate CO2 recovery rate whilst thesetpoint from raw flue gas flow controller is used as feedforwardsignal. CO2_CO2 and FG_CO2 mean CO2 flow rate in CO2 product andraw flue gas streams, respectively. CO2/FG show the ratio of raw fluegas flow rate to that of CO2 product and manipulated by PIDcontroller (CC_CRR). The detailed control structure is established,as shown in Fig. 3.

Table 7List of main control input/output variables.

Inputs Outputs

Variable Nominal Units Variable Nominal Units

Flue gas flow rate 717,186 kg/hr CO2 product purity 96.91 %MCC discharge

pressure30 bar S-8 temperature �55.50 �C

F1 temperature �24.64 �C CO2 recovery rate 93.08 %F2 temperature �55 �C S-18 temperature �31.25 �C

Page 7: Optimization and control for CO2 compression and ...cpc.energy.hust.edu.cn/__local/F/29/E2/059D2714F... · Optimization and control for CO2 compression and purification ... system

B. Jin et al. / Energy 83 (2015) 416e430422

4. Dynamic analysis

The dynamicmodel is then used to analyze transient behavior ofCPU. Several dynamic scenarios, including load change and opera-tional disturbance (i.e. flue gas composition), are applied to theinputs of model and key parameters like CO2 product purity, streamtemperature, CO2 recovery rate, and system pressure as the outputsare monitored.

4.1. Load change case

To integrate with oxy-combustion boiler system, CPU should beallowed to bear load change occurring frequently in power plant.Therefore, load change processes between 100% and 80% with aramp rate of 2%/min [21] are investigated here. After all PVs (pro-cess variables) reach back to their SPs (setpoints) or gain an offset[33], which are shown in Fig. 4(a) e (c), flue gas flow rate rampsdown from 100% at 10th minute to 80% and up from 80% at 30thminute to 100%. From Fig. 4(b), CO2 product purity decreasesgradually whilst CO2 recovery rate gets a stepwise decrease andthen increases to its SP during the ramp down process from 10thminute to 20th minute, and then they spent about 2.5 min and5 min bringing back to their SPs through cascade control action onsystem pressure controller (shown in right picture of Fig. 4(a)),respectively. With respect to temperature evolutions for S-18 andS-8 streams (see Fig. 4(c), left), they gain an inverse performancethat S-18 temperature increases about 1.75 �C while S-8 tempera-ture increases approximately 0.6 �C during ramp down process. Theoperating temperatures of F1 and F2 appear similar behavior (notshown here) since they are the corresponding upstream flows,respectively. These effects are attributed to the amount of refrig-eration capacity for flash separators and control actions on throttlevalves. Fig. 4(c) also presents variations of specific power con-sumption and specific cooling duty during load change process.During ramp down process, they evolve in an opposite way withthat of load change because the decrease of total CO2 recovered islarger than that of total power consumption and total cooling duty.The total CO2 recovered decreases about 19.93% while the total

Fig. 3. Dynamic model of CPU with designed D

power consumption and total cooling duty are reduced by 17.92%and 18.04%, respectively. In terms of ramp up process, an almostopposite dynamic behavior is attained. In addition, it should benoted that time constant (time period that PV reaches back to itsSP) and offset are related to the chosen PID tuning parameters (i.e.proportional gain Kc, integral time tI, and derivative time tD).

4.2. Flue gas compositions change case

Flue gas composition may vary subjected to variations in oxygenpurity from ASU, excess oxygen, air in-leakage, and fuel type inpower plant operation. In this section, ramp changes are applied toCO2 mole fraction in the flue gas whilst other impurities (O2, Ar, N2,CO, SOx and NOx) are recalculated simultaneously in order tomaintain the sum of mole fraction equal to unity. As presented inleft picture of Fig. 5(a), �5% and þ5% ramp cases from 10th minuteto 15th minute are performed. Decreasing CO2 mole fraction atconstant flue gas flow rate leads to a decrease of CO2 product puritybecause more impurities are presented, while inverse response ofCO2 recovery rate is obtained as shown in Fig. 5(b). To bring CO2product purity back to SP, control action on pressure controller thatdecreasing system pressure through opening vent gas valve (V-VENT) gradually is applied (see the right of Fig. 5(a)). Lasting about13min, CO2 product purity and CO2 recovery rate are turned back totheir SPs. However, with S-8 temperature approaches to its SP atabout 28th minute as shown in Fig. 5(c), CO2 recovery rate appearsa significant stepwise decrease. This adverse effect is attributed tothe unchanged operating temperature (S-18 and S-8 temperaturesreach their SPs), the decreased system pressure, and the decreasesof CO2 content in flue gas. Fig. 5(c) also shows that S-18 and S-8temperatures gain a decrease during 10th minute to 15th minutesince flue gas becomes easier to be cooled down when its heatcapacity decreases, and their time constant are 3 min and 13 min,respectively. More importantly, S-8 temperature attains favorableperformance since its value is always above �56.57 �C which is thecritical point for avoiding CO2 freezing in the second heatexchanger. Meanwhile, specific power consumption and specificcooling duty get an increase as the concurrent results of total

TC (double temperature control) system.

Page 8: Optimization and control for CO2 compression and ...cpc.energy.hust.edu.cn/__local/F/29/E2/059D2714F... · Optimization and control for CO2 compression and purification ... system

Fig. 4. Responses of process variables under load change case. (a) Input change and pressure. (b) CO2 product purity and CO2 recovery rate. (c) Temperature and energyconsumption.

B. Jin et al. / Energy 83 (2015) 416e430 423

energy (power and cooing duty) consumption and total CO2recovered. Similarly, these two parameters have a stepwise in-crease at 28th minute which is owing to the decrease of CO2 re-covery rate. Additionally, specific cooling duty gets a decrease at15th minute due to the increase of S-8 stream temperature whichleads to less cooling duty required in the cooler. In the case ofincreasing CO2 mole fraction, opposite trends in the interestedvariables are observed, except CO2 recovery ratewhich gains robustperformance and reaches its SP after the ramp change.

To bear load change and disturbance, some suggestions can beproposed. During ramp down process, decreasing S-18 and S-8temperatures appropriately would help to condensemore CO2 from

flue gas which could assist to reduce specific energy consumptionsimultaneously. With respect to flue gas composition change, abetter way to avoid CO2 solidification is to elevate the operatingtemperature of flash separators even though CO2 recovery ratewould be sacrificed. If CO2 content in flue gas decreases, decreasingthe SP of S-8 temperature controller or closing LCV119 properlymight be the feasible solution to the sudden decrease of CO2 re-covery rate. In addition, balance between CO2 product purity andCO2 recovery rate ought to be considered since they appear in anapproximate opposite way. To maintain CO2 recovery rate above90%, CO2 product purity target can be suggested to set below theoptimal value.

Page 9: Optimization and control for CO2 compression and ...cpc.energy.hust.edu.cn/__local/F/29/E2/059D2714F... · Optimization and control for CO2 compression and purification ... system

Fig. 5. Responses of process variables under flue gas composition change case. (a) Input change and pressure. (b) CO2 product purity and CO2 recovery rate. (c) Temperature andenergy consumption.

B. Jin et al. / Energy 83 (2015) 416e430424

5. Discussion

5.1. Alternative control structure

Different from DTC (double temperature control) discussed inthe former sections, a STC (single temperature control) strategythat only controlling S-8 temperature can also be applied to CPU. Toidentify the difference between these two control strategies, theirdynamic responses for same parameters and scenarios (loadchange and flue gas composition ramp change) are simulated,compared and analyzed in this section.

Input changes to represent load change process are the same asthose of double temperature control strategy. Comparisons ofprocess responses between these two control strategies are pre-sented in Fig. 6. Apart from almost the same transient behavioroccurred in CO2 recovery rate, other parameters obtain theirsmaller or larger variation ranges. Since pressure and S-8 temper-ature fluctuate within a narrow range, CO2 product purity variesaround more closely to its SP. However, it would lead to inverseeffects on energy performance, in which more power consumptionand cooling duty are required after ramp down process as pre-sented in Fig. 6(c).

Page 10: Optimization and control for CO2 compression and ...cpc.energy.hust.edu.cn/__local/F/29/E2/059D2714F... · Optimization and control for CO2 compression and purification ... system

B. Jin et al. / Energy 83 (2015) 416e430 425

Similar to process responses under DTC, the system performanceusing STC for flue gas composition ramp change is illustrated in Fig. 7.When CO2 mole fraction in flue gas decreases gradually, the CO2product purity, pressure and temperature will decrease whilst CO2recovery rate, specific power consumption, and specific cooling dutywill increase. Then, these parameters begin to reach to a new steadystate or their SPs after control actions on corresponding actuators.However, CO2 product purity gets a rapid decrease at 25th minutebecause F1 operating temperature (not shown here) decreasesapproximately 13 �C which is the result of no temperature controllerto force it back to its initial value. To bring CO2 product purity back to

Fig. 6. Comparison of two control strategies under load change case. (a) CO2 product puritconsumption and specific cooling duty.

its target, vent gas control valve (V-VENT) is opened to further reducethe system pressure. As a result, this action further aggravates thedecrease of CO2 recovery rate to below 90% as shown in Fig. 7(a).From the viewpoint of satisfying the third control objective, STCmaynot be a suitable choice.

5.2. Influences of SOx and NOx levels

Step changes for SOx and NOx (i.e. SO2 and NO) levels are con-ducted to identify their impacts on system performance. Here, theeffects are shown in Fig. 8 when step increasing SO2 and NO

y and CO2 recovery rate. (b) System pressure and S-8 temperature. (c) Specific power

Page 11: Optimization and control for CO2 compression and ...cpc.energy.hust.edu.cn/__local/F/29/E2/059D2714F... · Optimization and control for CO2 compression and purification ... system

Fig. 7. Comparison of two control strategies under flue gas composition change case. (a) CO2 product purity and CO2 recovery rate. (b) System pressure and S-8 temperature. (c)Specific power consumption and specific cooling duty.

B. Jin et al. / Energy 83 (2015) 416e430426

concentration to100 timesof their initial values.As shown inFig. 8(b),CO2 product puritygains a stepwisedecreasewhilst CO2 recovery rategets a stepwise increase at20thminute, and then their values reach totheir setpoints after about 10min since control signal from CC_CO2 issent to pressure controller to decrease system pressure which ispresented in Fig. 8(a). Comparedwith these two cases, it is found thatsystemperformance forSO2case is changedmoreobvious thanthat inNO case, which might be attributed to thermodynamic property andinteractions among components. However, it should be mentionedthat higher SO2 and NOwould contribute to serious corrosion issuesdue to the presence of some water contents.

5.3. Uncertainty analysis

When compared these two cases without and with kij mixingparameters, consistent conclusion as discussed in Ref. [7] can befound that these parameters does not influence the specific powerconsumption and specific cooling duty while are crucial for thecalculation of CO2 product purity and CO2 recovery rate. Whenchanging one operating parameter whilst maintaining other pa-rameters unchanged, variations of studied evaluating indicatorsunder simulation without kij are similar to those of PengeRobinsonstate of equation property method with kij. For multi-variable

Page 12: Optimization and control for CO2 compression and ...cpc.energy.hust.edu.cn/__local/F/29/E2/059D2714F... · Optimization and control for CO2 compression and purification ... system

Fig. 8. Responses of process variables under SO2 and NO step change case. (a) Input change and pressure. (b) CO2 product purity and CO2 recovery rate.

B. Jin et al. / Energy 83 (2015) 416e430 427

optimization, its optimization procedures are unchanged while theoptimal operating conditions would be changed to 29.96 bar,29.91 �C, �24.63 �C, and �55 �C for multi-stage CO2 compressordischarge pressure, flue gas temperature after compression, firstflash separator temperature, and second flash separator tempera-ture, respectively. Furthermore, simulation results for CO2 productpurity, CO2 recovery rate and specific power consumption areturned to be 96.02%, 94.50%, 114.68 kWh/tCO2, respectively.

Followed with variation of steady-state simulation results andoptimal operating conditions, some modifications would beapplied to dynamic model, control system design and transientresponses. As equipment sizing is based on steady-state simulationresults, the estimation of geometric sizes would be recalculatedwhile other dynamic preparations do not need any modification.Fortunately, it would not alter the proposed control structure eventhough initial conditions for dynamic simulation seem differentfrom those in the original case, because control system is highlyrelated to system inherent characteristics. Certainly, it wouldchange dynamic results rather than the trends of variations whendynamic scenarios are applied to dynamic model with designedcontrol system. Thus, reasonable property method is necessary forgaining accurate results but it should not make obstructions toprocess optimization and control system design.

6. Conclusion

The paper focuses on multi-variable optimization, control sys-tem design, and dynamic analysis of a CO2 CPU (compression andpurification unit) for oxy-combustion power plants. Simultaneousmulti-variable optimization and control system design for the CPU

are firstly realized in this paper using commercial simulationplatform Aspen Plus and Aspen Plus Dynamics. Effects of operatingparameters on system performance and optimal operating condi-tions are identified through single variable analysis and multi-variable optimization, respectively. A double temperature controlsystem is designed using a systematical top-down analysis andbottom-up design method and applied to the established dynamicCPU model with the optimal operating parameters. The designeddouble temperature control system is capable of investigating thedynamic behavior of CPU under different operating cases. Theseresults provide guidelines for operators to manipulate corre-sponding devices to approach the desired operating conditions.

From multi-variable optimization results, the optimal values ofmulti-stage CO2 compressor discharge pressure,flue gas temperatureafter compression, firstflash separator temperature, and secondflashseparator temperature are 30 bar, 30.42 �C, �24.64 �C, and �55 �C,respectively. The effectiveness and correctness of optimizationmethod are validated through exergy analysis, in which the exergyefficiency increases about 1.9 percent, specific power consumptiondecreases about 2.58%, and CO2 recovery rate increases about 0.09percent point, respectively. For comprehensive dynamicbehavior, it isshowed that the parameters interested are more sensitive to flue gascomposition change than that of load change. CO2 product purity andCO2 recovery rate attain similar performance between load changeand flue gas composition ramp change even though the decrease ofCO2 recovery rate occurs in the CO2 mole fraction ramp down case.However, they appear an inverse response during each case, whichmay indicate that the primary control objective should be carefullydetermined. In the present study, CO2 product purity is considered asthemost important goalwhile CO2 recovery rate ismerely required to

Page 13: Optimization and control for CO2 compression and ...cpc.energy.hust.edu.cn/__local/F/29/E2/059D2714F... · Optimization and control for CO2 compression and purification ... system

B. Jin et al. / Energy 83 (2015) 416e430428

be maintained above 90%. On the temperature responses front,different trends are gained from different operating cases since themain reasons for these temperature variations are related to refrig-eration capacity, control actions, and flue gas property. Moreover,particular attention should be paid to S-8 temperature which de-termines the risk of CO2 solidification (when below its triple pointtemperature). Thus, one feasible suggestion for avoiding this risk is toelevate its operating temperature. In addition, energy performanceinvolved in two different cases is also distinctive. Load downing andCO2 content decrease in flue gas are adverse to effective utilization ofenergy. In order to bear these operating disturbances, SPs in tem-perature controllers, valves or even CO2 product purity are suggestedto bemanipulated to help reduce the penalty of CO2 recovery rate andspecific energy consumption.

Single temperature control strategy is also studied as alterativecontrol strategy to compare with double temperature controlstrategy. During load change, these two control strategies gainsimilar dynamic behavior, but single temperature control strategyshows narrow variation range and higher specific energy con-sumption. For flue gas composition change, they also appear similarperformance while the final CO2 recovery rate is below 90% forsingle temperature control strategy. It may indicate that singletemperature control strategy is not the suitable option. Impacts of

Table 1Detailed stream conditions for initial process simulation model as sketched in Fig. 1.

Stream no. S-1 S-2 S-3

Composition, mol.%

CO2 82.40 82.40 82.40O2 4.19 4.19 4.19N2 9.50 9.50 9.50Ar 3.90 3.90 3.90CO (ppm) 31 31 31SO2 (ppm) 14 14 14SO3 (ppb) 83 83 83NO (ppm) 20 20 20NO2 (ppb) 38 38 38Flow rate, kg/hr 717,186 717,186 717,186Temperature, �C 50 35 �24.51Pressure, bar 1.10 30 29.72

Stream No. S-8 S-9 S-10

Composition, mol.%

CO2 95.60 95.60 95.60O2 1.26 1.26 1.26N2 2.08 2.08 2.08Ar 1.06 1.06 1.06CO (ppm) 7 7 7SO2 (ppm) 4 4 4SO3 (ppb) 2 2 2NO (ppm) 11 11 11NO2 (ppb) 2 2 2Flow rate, kg/hr 186,062 186,062 186,062Temperature, �C �55.06 �43.15 23.40Pressure, bar 8.74 8.54 8.33

Stream No. S-15 S-16 S-17

Composition, mol.%

CO2 24.29 24.29 97.51O2 17.51 17.51 0.69N2 41.62 41.62 1.21Ar 16.56 16.56 0.60CO (ppm) 136 136 4SO2 (ppm) 0.0772 0.0772 24SO3 (ppb) 0.002 0.002 150NO (ppm) 75 75 4NO2 (ppb) 0.004 0.004 68Flow rate, kg/hr 118,450 118,450 412,675Temperature, �C 23.40 23.36 �24.51Pressure, bar 28.9 28.8 29.72

SOx and NOx levels are identified through dynamic simulationunder their step change processes. It gains an interesting findingthat influence of SOx would be more obvious than that of NOx eventhough higher contents lead to more corrosion for equipment.Through uncertainty analysis, a conclusion can be drawn that theproperty method is crucial for calculating accuracy but it affectslittle on process simulation and control system design.

In general, CO2 recovery rate control may be challenging sincethere exist strong interactions between different process variables.Certainly, some improvements and potential control strategiesshould be investigated to make dynamic model approach to realplant, and more efforts are required to optimize the developedcontrol system.

Acknowledgment

This work was supported by “The National Natural ScienceFoundation of China (51390494)”, and “National Key Basic Researchand Development Program (2011CB707300)”.

Appendix A

S-4 S-5 S-6 S-7

63.89 63.89 95.60 95.608.49 8.49 1.26 12.619.66 19.66 2.08 2.087.95 7.95 1.06 1.0664 64 7 72 2 4 41 1 2 239 39 11 111 1 2 2304,512 304,512 186,062 186,062�24.51 �54.69 �54.69 �43.1529.72 29.45 29.45 29.24

S-11 S-12 S-13 S-14

95.60 95.60 24.29 24.291.26 1.26 17.51 17.512.08 2.08 41.62 41.621.06 1.06 16.56 16.567 7 136 1364 4 0.0772 0.07722 2 0.002 0.00211 11 75 752 2 0.004 0.004186,062 186,062 118,450 118,45097.00 40 �54.69 �43.1518.59 18.49 29.45 29.17

S-18 S-19 S-20 S-21

97.51 97.51 96.91 96.910.69 0.69 0.87 0.871.21 1.21 1.48 1.480.60 0.60 0.74 0.744 4 5 524 24 17 17150 150 104 1044 4 6 668 68 47 47412,675 412,675 598,736 598,736�31.12 23.40 28.46 28.3618.8 18.59 18.49 18.39

Page 14: Optimization and control for CO2 compression and ...cpc.energy.hust.edu.cn/__local/F/29/E2/059D2714F... · Optimization and control for CO2 compression and purification ... system

B. Jin et al. / Energy 83 (2015) 416e430 429

Appendix B

Exergy is the maximum theoretical useful work obtained as thesystem state changes toward the dead state reversibly whenexcluding other heat sources. Exergy is also a comprehensivemeasure that coordinates quality with quantity of energy. The totalexergy (E) is the sum of physical exergy (EPH), chemical exergy(ECH), kinetic exergy (EKN), and potential exergy (EPT). EKN and EPT

are relating to velocity and elevation, respectively. For thermody-namic system like CPU, only EPH and ECH are left for calculation. TheReference-Substance Model proposed by Gaggioli and Petit (1977)is chosen as the reference state to carry out exergy analysis, whichis shown in Table 2.

Table 2Environmental reference state

Temperature T0 298.15 K

Pressure P0 1 atm

Composition

Air constituents Mole fraction Condensed phase

N2 0.7567 H2O LiquidO2 0.2035 CaCO3 SolidH2O 0.0303 CaSO4$2H2O SolidAr 0.0091H2 0.0001CO2 0.0003

EPH is defined as the maximum theoretical useful work obtainedas a system passes from its initial state (T and P) to the environmentstate (T0 and P0). The unit physical exergy (ePH, kJ/kmol) is calcu-lated as

Table 3Exergy streams for CO2 compression and purification unit

Stream no. S-1 S-2 S-3 S-4 S-5 S-6 S-7

Exergy, kW

Physical exergy 266.21 38493.08 44597.22 17637.46 22082.80 12422.41 12036.76Chemical exergy 74652.45 74652.45 74652.45 24871.75 24871.75 22205.80 22205.80Exergy 74918.66 113145.53 119249.67 42509.22 46954.55 34628.22 34242.56

Stream No. S-8 S-9 S-10 S-11 S-12 S-13 S-14

Exergy, kW

Physical exergy 11752.50 6512.85 6042.42 8543.14 8228.64 8029.95 7942.65Chemical exergy 22205.80 22205.80 22205.80 22205.80 22205.80 4297.80 4297.80Exergy 33958.30 28718.65 28248.22 30748.94 30434.45 12327.75 12240.45

Stream No. S-15 S-16 S-17 S-18 S-19 S-20 S-21

Exergy, kW

Physical exergy 7781.96 7774.14 25684.72 25465.96 18400.29 26570.78 26524.71Chemical exergy 4297.80 4297.80 51056.86 51056.86 51056.86 73251.07 73251.07Exergy 12079.77 12071.95 76741.57 76522.82 69457.15 99821.85 99775.78

ePH ¼ ðh� h0Þ � T0ðs� s0Þ (1)

where h and s mean unit enthalpy (kJ/kmol) and unit entropy (kJ/kmol�1 K�1), respectively. The subscript “0” means the referencestate.

ECH arises from the departure of the chemical composition of asystem from the environment. When calculating the chemicalexergy, there are different cases depending upon material streams.For the condensed phase in the reference environment model, itsunit chemical exergy is equal to 0. For a gaseous component in the

reference environment model, its unit chemical exergy (eCH, kJ/kmol) can be expressed as

eCH ¼ �RT0 ln x0 (2)

in which R is the gas constant, 8.314 J K�1 mol�1, and x0 means themole fraction of the gaseous component in the environmentmodel.It should be noted that all of the gaseous components are treated asideal gases.

On the other hand, for a gaseous mixture, its unit chemicalexergy (eCH, kJ/kmol) can be presented as

eCH ¼X

xkeCHk þ RT0

Xxkln xk (3)

in which the subscript “k” means the gaseous component in themixture, and xk means the mole fraction of the k gaseous compo-nent. For a non-gaseous mixture, its eCH is the weighted sum of unitchemical exergies of all components in the mixture.

The chemical exergy of a substance which is not present in theenvironment model or difficult to find its chemical exergy can becalculated through considering an idealized reaction of the sub-stance with other substances for which the chemical exergy areknown, and the basic formula is described as

eCH ¼ �(X

pvpgp �

Xr

vrgr

)þ(X

pvpeCHp �

XR

vreCHr

)(4)

Here, v and g mean the chemical stoichiometric number and unitGibbs free energy, and subscribe “p” and “r” mean the product andreactant, respectively.

Based on these exergy analysis methods mentioned above,exergy results for all streams in CO2 compression and purificationunit are listed in Table 3.

References

[1] Yan J, Anheden M, Lindgren G, Str€omberg L, Vattenfall Utveckling A,Vattenfall A. Conceptual development of flue gas cleaning for CO2 capturefrom coal-fired oxyfuel combustion power plant. In: Proceedings of 8th In-ternational Conference on Greenhouse Gas Control Technologies; 2006.

[2] Yan J, Anheden M, Faber R, Starfelt F, Preusche R, Ecke H, et al. Flue gascleaning for CO2 capture from coal-fired oxyfuel combustion power genera-tion. Energy Proc 2011;4(2011):900e7.

[3] Liu H, Shao Y. Predictions of the impurities in the CO2 stream of an oxy-coalcombustion plant. Appl Energy 2010;87(10):3162e70.

[4] Pipitone G, Bolland O. Power generation with CO2 capture: technology for CO2purification. Int J Greenh Gas Control 2009;3(5):528e34.

Page 15: Optimization and control for CO2 compression and ...cpc.energy.hust.edu.cn/__local/F/29/E2/059D2714F... · Optimization and control for CO2 compression and purification ... system

B. Jin et al. / Energy 83 (2015) 416e430430

[5] De Visser E, Hendriks C, de Koeijer G, Liljemark S, Barrio M, Austegard A, et al.D 3.1. 3 DYNAMIS CO2 quality recommendations. In: DYNAMIS Project (6thFramework Programme), The Netherlands; 2007. p. 16e35.

[6] Metz B, Davidson O, de Coninck H, Loos M, Meyer L. IPCC special report oncarbon dioxide capture and storage. Geneva (Switzerland): IntergovernmentalPanel on Climate Change; 2005. Working Group III.

[7] Posch S, Haider M. Optimization of CO2 compression and purification units(CO2CPU) for CCS power plants. Fuel 2012;101:254e63.

[8] Darde A, Prabhakar R, Tranier J-P, Perrin N. Air separation and flue gascompression and purification units for oxy-coal combustion systems. EnergyProc 2009;1(1):527e34.

[9] White V, Torrente-Murciano L, Sturgeon D, Chadwick D. Purification ofoxyfuel-derived CO2. Energy Proc 2009;1(1):399e406.

[10] Shah M, Degenstein N, Zanfir M, Kumar R, Bugayong J, Burgers K. Near zeroemissions oxy-combustion CO2 purification technology. Energy Proc 2011;4:988e95.

[11] Ritter R, Kutzschbach A, Stoffregen T. Energetic evaluation of a CO2 purifica-tion and compression plant for the oxyfuel process. In: Proc of 1st OxyfuelCombustion Conference, Cottbus, Germany; 2009.

[12] Fu C, Gundersen T. Reducing the power penalty related to CO2 conditioning inoxy-coal combustion plants by pinch analysis. Chem Eng 2011;25:581.

[13] Fu C, Gundersen T. Techno-economic analysis of CO2 conditioning processesin a coal based oxy-combustion power plant. Int J Greenh Gas Control 2012;9:419e27.

[14] Besong MT, Maroto-Valer MM, Finn AJ. Study of design parameters affectingthe performance of CO2 purification units in oxy-fuel combustion. Int J GreenhGas Control 2013;12:441e9.

[15] Pottmann M, Engl G, Stahl B, Ritter R. Dynamic simulation of oxyfuel CO2processing plants. Energy Proc 2011;4:951e7.

[16] Kuczynski K, Fitzgerald F, Adams D, Glover F, White V, Chalmers H, et al.Dynamic modelling of oxyfuel power plant. Energy Proc 2011;4:2541e7.

[17] Chansomwong A, Zanganehb KE, Shafeenb A, Douglas PL, Croiset E, Ricardez-Sandoval LA. Dynamic simulation and control of a CO2 compression and pu-rification process for oxy-coal-fired power plants. In: 3rd Oxyfuel CombustionConference; 2013.

[18] Chansomwong A, Zanganeh KE, Shafeen A, Douglas PL, Croiset E, Ricardez-Sandoval LA. A decentralized control structure for a CO2 compression, capture

and purification process: an uncertain relative gain array approach. In: WorldCongress; 2011. p. 8558e63.

[19] Dillon DJ, White V, Allam RJ, Wall RA, Gibbins J. Oxy-combustion processes forCO2 capture from power plant. IEA Greenhouse Gas R&D Programme; 2005.

[20] Jin B, Zhao H, Zheng C. Dynamic simulation for mode switching strategy in aconceptual 600 MWe oxy-combustion pulverized-coal-fired boiler. Fuel2014;137:135e44.

[21] Jin B, Zhao H, Zheng C. Dynamic modeling and control for pulverized-coal-fired oxy-combustion boiler island. Int J Greenh Gas Control 2014;30:97e117.

[22] Xiong J, Zhao H, Zheng C. Thermoeconomic cost analysis of a 600MWe oxy-combustion pulverized-coal-fired power plant. Int J Greenh Gas Control2012;9:469e83.

[23] Xiong J, Zhao H, Zheng C, Liu Z, Zeng L, Liu H, et al. An economic feasibilitystudy of O2/CO2 recycle combustion technology based on existing coal-firedpower plants in China. Fuel 2009;88(6):1135e42.

[24] Zebian H, Gazzino M, Mitsos A. Multi-variable optimization of pressurizedoxy-coal combustion. Energy 2012;38(1):37e57.

[25] Xiong J, Zhao H, Zheng C. Exergy analysis of a 600 MWe oxy-combustionpulverized-coal-fired power plant. Energy Fuels 2011;25(8):3854e64.

[26] Bejan A, Tsatsaronis G, Moran M. Thermal design & optimization. New York:John Wiley & Sons; 1996.

[27] Aspen Technology Inc. Aspen Dynamics™ 11.1 user guide. 2001.[28] Luyben WL. Plantwide dynamic simulators in chemical processing and con-

trol. CRC Press; 2002.[29] Larsson T, Skogestad S. Plantwide control e a review and a new design pro-

cedure. Model Identif Control 2000;21(4):209e40.[30] Skogestad S. Control structure design for complete chemical plants. Comput

Chem Eng 2004;28(1):219e34.[31] Luyben ML, Tyreus BD, Luyben WL. Plantwide control design procedure.

AICHE J 1997;43(12):3161e74.[32] Chansomwong A, Zanganeh K, Shafeen A, Douglas P, Croiset E, Ricardez-

Sandoval L. Dynamic modelling of a CO2 capture and purification unit for anoxy-coal-fired power plant. Int J Greenh Gas Control 2014;22:111e22.

[33] Luyben WL. Process modeling, simulation and control for chemical engineers.McGraw-Hill Higher Education; 1989.