Вы находитесь на странице: 1из 29

Energy Conversion and Management 45 (2004) 24952523

www.elsevier.com/locate/enconman

Integration of an iterative methodology for exergoeconomic improvement of thermal systems with a process simulator
Leonardo S. Vieira a, Jo~o L. Donatelli b, Manuel E. Cruz a
a

c,*

CEPEL, Av. Hum s/n, Cidade Universitria, CP 68007, Rio de Janeiro, RJ 21944-970, Brazil a b UFES, Department of Mechanical Engineering, CP 019011, Vitria, ES 29060-970, Brazil o c COPPE/UFRJ, Department of Mechanical Engineering, CP 68503, CT, Cidade Universitaria, Rio de Janeiro, RJ 21945-970, Brazil Received 7 August 2003; accepted 12 November 2003 Available online 24 January 2004

Abstract In this paper, we present the development and automated implementation of an iterative methodology for exergoeconomic improvement of thermal systems integrated with a process simulator, so as to be applicable to real, complex plants. The methodology combines recent available exergoeconomic techniques with new qualitative and quantitative criteria for the following tasks: (i) identication of decision variables that aect system total cost and exergetic eciency; (ii) hierarchical classication of components; (iii) identication of predominant terms in the component total cost; and (iv) choice of main decision variables in the iterative process. To show the strengths and potential advantages of the proposed methodology, it is here applied to the benchmark CGAM cogeneration system. The results obtained are presented and discussed in detail and are compared to those reached using a mathematical optimization procedure. 2003 Elsevier Ltd. All rights reserved.
Keywords: Exergoeconomic improvement; Thermal systems; Process simulator; Optimization; Exergy

1. Introduction The growing concern about conservation of energy and environmental preservation [1] has led to the development and intense application of techniques based on the second law of thermodynamics, such as exergoeconomic analysisan exergy analysis combined with an economic evaluation. Today, exergy is seen as a conuence of energy, environment and sustainable development
*

Corresponding author. Tel.: + 55-21-2562-8403; fax: + 55-21-2562-8383. E-mail address: manuel@serv.com.ufrj.br (M.E. Cruz).

0196-8904/$ - see front matter 2003 Elsevier Ltd. All rights reserved. doi:10.1016/j.enconman.2003.11.007

2496

L.S. Vieira et al. / Energy Conversion and Management 45 (2004) 24952523

Nomenclature Bk Bk1 Bk2 c _ C CRF _ E f i LHV _ m mk n nk NF NK NP p P PEC q r RPc s t T TCI x _ Z Greek a b d D e c g r1 r2 wdrk;x
i

constant in cost Eq. (24) for kth component constant in cost Eq. (24) for kth component constant in cost Eq. (24) for kth component specic exergetic cost exergetic cost ow rate capital recovery factor exergy ow rate exergoeconomic factor interest rate lower heating value of fuel mass ow rate exponent in cost Eq. (24) for kth component useful system life exponent in cost Eq. (24) for kth component number of system fuels number of system components number of system products parameter for sensitivity analysis, expressions (4) and (5) pressure purchased equipment cost parameter for choosing main decision variables relative cost dierence compressor compression ratio standard deviation parameter for hierarchical classication of components temperature total capital investment decision variable investment cost ow rate letters user prescribed tolerance for the iterative process, Eq. (1) multiplying factor for purchased equipment cost relative change in a variable absolute change in a variable exergetic eciency maintenance factor isentropic eciency modied structural bond coecient, Eq. (11) modied structural bond coecient, Eq. (12) sensitivity coecient, Eq. (16)

L.S. Vieira et al. / Energy Conversion and Management 45 (2004) 24952523

2497

wdek;x i s

sensitivity coecient, Eq. (17) yearly plant operating hours

Subscripts a air AC air compressor D exergy destruction F fuel g gas GT gas turbine i ith decision variable, ith system exergy ow rate iter iteration k kth plant component l lower limiting value L loss NK number of components P product s steam total total u upper limiting value Superscript OPT optimum

[2], and exergoeconomic analysis can address environmental issues, reveal the cost formation process of system products and aid system optimization. A critical review of relevant publications between the 1970s and 1990s regarding exergy and exergoeconomic analysis is found in an article by Tsatsaronis [3]. Recent important contributions to this eld are due to Frangopoulos [4], von Spakovsky [5], Lozano and Valero [6], Tsatsaronis [7], Hua et al. [8], Kim et al. [9], Zhang et al. [10], Kwon et al. [11], Sciubba [12] and Rosen and Dincer [13]. According to El-Sayed and Gaggioli [14,15], exergoeconomic methods can be grouped in two classes: algebraic methods and calculus methods. The algebraic methods use algebraic balance equations, always require auxiliary cost equations for each component, focus essentially on the cost formation process and determine average costs. On the other hand, calculus methods use dierential equations, such that the system cost ows are obtained in conjunction with optimization procedures based on the method of Lagrange multipliers, and determine marginal costs. Algebraic methods (e.g., [6,7,9,11,13,16]) are considered subjective with regard to the denition of auxiliary equations. The theory of the exergetic cost by Lozano and Valero [6] is a matrix formulation of the productive structure of the system, adequate for computer implementation. Tsatsaronis [7] introduces an iterative exergoeconomic optimization procedure based on exergoeconomic variables (relative cost dierence, exergoeconomic factor and exergetic eciency) and on minimization of the product cost of each system component treated in thermoeconomic isolation. Lazzaretto and Tsatsaronis [16] present the Specic Exergy Costing/Average Cost

2498

L.S. Vieira et al. / Energy Conversion and Management 45 (2004) 24952523

(SPECO/AVCO) approach, seeking to reduce the subjectivity of fuel and product denitions and cost partitioning. Subsequent improvements of the SPECO/AVCO methodology by Tsatsaronis and co-workers [17,18] involve combination with fuzzy inference systems for a more precise exergoeconomic evaluation of plant components [17] and focus only on the unavoidable exergy destruction and investment costs [18]. However, such improvements are still not free from subjective decisions. In principle, the SPECO/AVCO methodology, as well as others (MOPSA [9], [11], EXCEM [13]), can be applied to the analysis of any system, but they lack some objective criteria for automated computer optimization of complex thermal systems. Calculus methods (e.g. [4,5]) are generally based on the Lagrange multipliers technique and are considered subjective with regard to the mathematical description of the function of each component in the system. A particular diculty in the application of calculus methods to complex systems is the fact that the Lagrange multipliers vary from iteration to iteration when component thermoeconomic isolation [19] is not achieved. This problem has led to the development of the Thermoeconomic Functional Analysis (TFA) [4]. TFA introduces the concept of the functional diagram, and assigns only one function and one product to each component, such that auxiliary cost equations are not needed. With the application of the second law in TFA, the Lagrange multipliers show relatively smaller variations in the optimization process. Despite later simplications of the TFA [5], construction of the functional diagram remains subjective, and convergence diculties are encountered as the system size increases. Additionally, decomposition strategies (e.g. [8,20,21]) based on second law reasoning have been proposed to reduce complexity in the optimization of complete systems. Hua et al. [8] suggest that an energy system be decomposed into a main subsystem and a recovery subsystem, with a reversed exergy costing procedure for the recoverable exergy. El-Sayed [20] proposes the decomposition of the Lagrangean of the system, together with the identication of local and global variables. Benelmir and Feidt [21] present the IEEB method, in which an expression of the optimal unit cost is developed for each component of the system through exergy and cost balance equations. The method is, in fact, a consequence of the application of the Lagrange multipliers technique. Yet, these diverse strategies rest on problem dependent, subjective system decomposition. Most exergoeconomic optimization theories have been applied to relatively simple systems only. Mathematical optimization, exergoeconomic or not, of real thermal systems are large scale problems due to their complicated nonlinear characteristics and because the mass, energy and exergy (or entropy) balance equations must be introduced in the problem as restrictions [22]. To optimize complex systems eciently, it is, thus, desired, if not required, to integrate optimization algorithms with a professional thermodynamic process simulator, such that the optimization task does not have to deal with the mass, energy and exergy balance equations. Furthermore, according to Jaluria [22], it is important to focus on the dominant decision variables, rather than manipulating simultaneously all variables that might aect the solution. In this paper, we present the development and automated implementation of a new approach for design improvement of complex thermal systems based on the integration of an iterative methodology for exergoeconomic improvement with a professional process simulator. As such, the proposed methodology is algebraic and should be easily assimilated and applied by practising engineers of the industrial community. The iterative algorithm eciently combines well known exergoeconomic techniques [6,7] with new qualitative and quantitative criteria for the following tasks: (i) identication of variables that aect system total cost and exergetic eciency;

L.S. Vieira et al. / Energy Conversion and Management 45 (2004) 24952523


feedwater steam

2499

8 7
HRSG

9
preheater

5
natural gas

3 2
compressor

10
combustor

11

11

12 1
air gas turbine

Fig. 1. Schematic of the CGAM cogeneration system (HRSG is the heat recovery steam generator).

(ii) hierarchical classication of components; (iii) identication of predominant terms in the component total cost; and (iv) choice of main decision variables in the iterative process. With regard to task (iv), the proposed method selects in every iteration which of the decision variables are to be modied through an auxiliary mathematical optimization procedure; these variables are called main decision variables. Two alternatives to select the main decision variables are identied and implemented. In Alternative 1, the choice is based on modied structural bond coecients. Structural bond coecients were dened originally by Kotas [23]. In Alternative 2, the choice is based on the sensitivity of two exergoeconomic variablesrelative cost dierence and exergetic eciencywith respect to changes in the decision variables. For comparison and evaluation purposes, the proposed method, designed to treat complex systems with no user interference, is applied in this paper to the benchmark CGAM cogeneration system [24,25], shown schematically in Fig. 1. We remark that, by necessitysee Section 4the physical, thermodynamic and economic models used in this paper are dierent from those of the original CGAM problem. The results obtained are compared to those reached using a mathematical optimization procedure. The main dierence between the proposed method and a mathematical method is that the former does not manipulate all of the components and decision variables in each iteration. This computational advantage is due to the knowledge gained from a preceding exergoeconomic analysis of the system.

2. Proposed iterative methodology for exergoeconomic improvement To start, it is assumed that the system physical, thermodynamic and economic models are known, as well as the design variables and their respective initial values for the improvement process. The proposed method consists of six steps, and requires no user interference. The steps combine and complement the iterative procedure originally developed by Tsatsaronis [3,7,25], the

2500

L.S. Vieira et al. / Energy Conversion and Management 45 (2004) 24952523

denitions of fuel and product and auxiliary cost equations by Lazzaretto and Tsatsaronis [26], and the cost formulation of the theory of the exergetic cost by Lozano and Valero [6]. Iteration stops when changes in the decision variables do not further reduce the objective function, according to the equation _ _ Ctotal;iter1 Ctotal;iter < a; _ Ctotal;iter 1

_ where Ctotal is the system total cost, i.e. the sum of the cost ows of the system products, subscripts iter + 1 and iter represent consecutive iterations and a is a user prescribed tolerance. 2.1. Step 1 _ First, perform an exergoeconomic system analysis to determine all exergy ow rates (Ei ), exergy _ _ destruction ow rates (ED;k ), component exergetic eciencies (ek ), exergetic cost ow rates (Ci ), component product (cP;k ) and fuel (cF;k ) specic exergetic costs, exergy destruction cost ow rates _ _ _ (CD;k ), component investment cost ow rates (Zk ), and system total cost (Ctotal ). Also, for each component k, determine two exergoeconomic variables, the relative cost dierence (rk ) and the exergoeconomic factor (fk ). Detailed information on how to calculate these variables can be found in the article by Tsatsaronis [7] and in the book by Bejan et al. [25]. Denitions of fuel and product and auxiliary cost equations are the same as those proposed by Lazzaretto and Tsatsaronis [26]. Cost balance equations are implemented according to the theory of the exergetic cost by Lozano and Valero [6]. Next, dene the values of the following parameters to be used in the course of the iterative process: lower (xi;l ) and upper (xi;u ) limiting values for each decision variable (xi ), increments (Dx) for numerical evaluation of the sensitivity of the system total cost and system exergetic eciency with respect to the decision variables, parameter for identication of variables that aect total cost (p), parameter for hierarchical classication of components (t), parameter for choosing main decision variables (q) and the tolerance a. Finally, some additional economic parameters must also be dened: the multiplying factor for the purchased equipment cost (b), the useful system life (n), the interest rate (i), the yearly plant operating hours (s) and the maintenance factor (c). The objective function is the specic total cost of the products, expressed by PNF PNK NP X _ _ i1 cFi EFi k1 Zk cPi min ; 2 min PNP _ i1 i1 EPi where NP, NF, and NK are the total number of product ows, fuel ows and system components, respectively. It is observed that when there is just one fuel and the product is constant, as in the CGAM problem, Eq. (2) reduces to [25] P X k CRF cTCIk _ _ total _ k CF _ Z 3 C cF mF LHV; s k where CRF is the capital recovery factor (see Eq. (20), here not applied to the operation and maintenance cost rates), TCI is the total capital investment (see Eq. (18)), and LHV is the lower heating value of the fuel.

L.S. Vieira et al. / Energy Conversion and Management 45 (2004) 24952523

2501

2.2. Step 2 Perform a numerical sensitivity analysis of the eect of each decision variable (xi ) on the system _ total cost (Ctotal ) and system exergetic eciency (e). Decision variables are thus grouped into four _ dierent classes: (1) those that aect Ctotal only, (2) those that aect e only, (3) those that aect _ _ total and e and (4) those that aect neither Ctotal nor e. Class-1 variables are modied at the both C end of the iteration. Class-4 variables are not modied in the iteration. The sensitivity analysis is performed for each xi according to the following expressions: if xi De > p ) xi affects exergetic efficiency; Dxi e _ xi DCtotal > p ) xi affects system total cost: _ Dxi Ctotal 4

if

The left-hand sides of expressions (4) and (5) are numerically evaluated and compared to the value of the parameter p dened in Step 1. 2.3. Step 3 Identify the most important components, i.e. those that exert the larger inuence on the system total cost. Three tests are performed to categorize the components hierarchically into main, secondary and remainder components:
NK 1 X _ _ _ _ Zi CD;i tsZi CD;i ) component k is main; if Zk CD;k > _ _ NK i1 NK 1 X _ _ _ _ if Zk CD;k > Zi CD;i tsZi CD;i _ _ NK i1

and 7

NK 1 X _ _ _ _ Zk CD;k < Zi CD;i tsZi CD;i ) component k is secondary; _ _ NK i1 NK 1 X _ _ _ _ if Zk CD;k < Zi CD;i tsZi CD;i ) component k is remainder; _ _ NK i1

where the parameter t is dened in Step 1. Components are ranked in decreasing order in terms of _ _ the sum of the investment and exergy destruction cost rates (Zk CD;k ). The main components exert the largest inuences on the system total cost and are considered rst in the improvement process. The remainder components exert the smallest eects on the system total cost and are not considered in the iteration. It is observed that the rank of a component may vary in the course of the improvement process. For example, a main component in one iteration can become secondary in the next iteration or vice-versa.

2502

L.S. Vieira et al. / Energy Conversion and Management 45 (2004) 24952523

2.4. Step 4 Identify, for each main and secondary component, the cost element (investment or exergy destruction) that exerts the larger inuence on the component total cost, according to the following tests: if _ CD;k _ > q ) CD;k is predominant; _ _ Zk CD;k _ Zk _ > q ) Zk is predominant; _ _ Zk CD;k 9

if

10

where the parameter q, dened in Step 1, is chosen in the range 1 > q P 0:5. 2.5. Step 5 In this step, two distinct alternatives are identied for the choice of the main decision variables for each component. Alternative 1 is based on modied structural bond coecients, dened as r1k;xi _ DCD;total =Dxi ; _ DCD;k =Dxi _ _ DZ CD total =Dxi : _ _ DZ CD k =Dxi 11

r2k;xi

12

The coecients r1k;xi and r2k;xi are modied relative to the original structural bond coecients dened by Kotas [23]. From Eqs. (11) and (12), it is seen that the modied coecients r1k;xi and r2k;xi are based, respectively, on the cost of exergy destruction and on the total cost (investment cost plus exergy destruction cost) of component k. Positive values of r1k;xi (respectively, r2k;xi ) indicate that a change in the decision variable xi that causes a reduction in the exergy destruction (respectively, total) cost of component k also leads to a reduction in the exergy destruction (respectively, total) cost of the entire system. The coecients are used for selecting decision variables: when the component exergy destruction cost is far superior to its investment cost (Eq. (9) is satised), decision variables are selected for positive values of r1k;xi ; otherwise, decision variables are selected for positive values of r2k;xi . Alternative 1 has the useful feature of being able to handle component cost equations in any analytical form. Alternative 2 requires that component cost equations be in the analytical form  nk ek _ mk EP;k ; 13 PECk Bk 1 ek where PECk is the purchased equipment cost. Eq. (13) imposes that the PECk should depend on _ the component exergetic eciency, ek , and the product exergetic ow rate, EP;k . In Eq. (13), Bk , nk and mk are component specic constants, which may be derived by curve tting to appropriate cost data. Similarly to the methodology by Tsatsaronis [7], Alternative 2 is based on the relative

L.S. Vieira et al. / Energy Conversion and Management 45 (2004) 24952523

2503

deviations between the actual and the optimal values of exergetic eciency and relative cost dierence for each main and secondary component: dek ek eOPT k 100; eOPT k
OPT rk rk 100: OPT rk

14

drk

15

OPT The optimal values eOPT and rk in Eqs. (14) and (15) are determined considering component k k isolated from the system (see Section 4.2 for details) and expressing PECk by an equation of the form of Eq. (13). It is proposed here that these equations be applied to the main and secondary components only, as dened in Step 3. In Alternative 2, the choice of main decision variables is based on negative values for the sensitivity coecients wdek;x and wdrk;x given by
i i

wdek;x

Ddek xi ; dek Dxi Ddrk xi : drk Dxi

16 17

wdrk;x
i

The value of wdek;x is used when Eq. (9) is satised (i.e. when the exergy destruction cost is i predominant), otherwise the value of wdrk;x is used. The proposed Eqs. (16) and (17) permit i automation of the methodology. 2.6. Step 6 For the main and secondary components, perform a mathematical optimization for the selected decision variables that aect the system total cost and/or exergetic eciency (Step 2). Any conventional mathematical optimization method can be chosen at this point [27]. The objective function is the specic total cost of the system. Consider rst the decision variables of the main components and then those of the secondary components. Finally, consider the decision variables that aect the system total cost only. In the optimization process, decision variables are bounded, and they are modied within the limits of validity of the component cost equations.

3. Integration with a process simulator In order that the proposed iterative methodology for exergoecomomic improvement be applicable to complex systems, it must exploit the computational power of a process simulator. Here, the IPSEpro simulator [28] has been chosen. The IPSEpro contains the basic thermal simulator module PSE, which can be coupled to the optional modules Model Development Kit (MDK) and PSE-Excel. The integration of the methodology with the IPSEpro is depicted in Fig. 2.

2504

L.S. Vieira et al. / Energy Conversion and Management 45 (2004) 24952523


MS-EXCEL IPSEpro

Worksheets
decision variables all variables sending

PSE
simulation environment

receiving

running

Macros
Visual Basic routines

PSE-Excel
(interface)

MDK

Fig. 2. Integration of the improvement routine with the IPSEpro simulator program.

In this work, both modules MDK and PSE-Excel are used in addition to the Advanced Power Library (APP-Lib), which contains models of the components of thermal systems. The MDK module permits programming and modication of component models in the APP-Lib library. With MDK, it is possible to incorporate the exergy property calculation into the simulator. The PSE-Excel module permits integration of the improvement routine with the PSE simulator through the Dynamic Data Exchange (DDE) protocol. Also, through Excel worksheets, it is possible to exchange variables between IPSEpro and the improvement routine, written in the VBA (Visual Basic) language. Control of data exchange can be eected in two dierent ways: rst, through the additional menu commands that become available during installation of the PSE-Excel integration module and second, through Excel macros series of special commands to exchange input and output data and run IPSEproalso available during the installation of PSE-Excel. The latter procedure has been chosen in this work because it is necessary to control the data exchange from within the improvement routine. The simulator determines all mass, energy and exergy ow rates of the system and is called by the improvement (VBA) routine each time a modication of any decision variable is necessary. To prevent failure of the VBA routine due to IPSEpros internal error messages caused by selection of unfeasible thermodynamic data, a penalty function is used, whose value is added to the objective function. Note that the improvement routine does not have to deal with the thermodynamic balance equations as restrictions. As previously mentioned, the proposed method (Alternatives 1 and 2) requires that an auxiliary mathematical optimization procedure be used to modify the main decision variables in the iterative process. In principle, any mathematical optimization method will do. In this work, the exible polyhedron method developed by Nelder and Mead [27] has been selected and modied to comply with the bounded intervals for the decision variables.

L.S. Vieira et al. / Energy Conversion and Management 45 (2004) 24952523

2505

4. Application to the CGAM cogeneration system The application of the proposed exergoeconomic improvement procedure to the benchmark CGAM cogeneration system [24,25] is now described. Because the CGAM cogeneration system is relatively simple, it is possible to compare the results obtained with the proposed method (Alternatives 1 and 2) to those reached using a mathematical optimization procedure. It is important to remark that the physical, thermodynamic and economic models used in this paper are dierent from those of the original CGAM problem: here, the physical and thermodynamic models are solved with the IPSEpro program, not through simplied balance and property equations, and the economic model consists of component cost equations derived from ts to data points obtained from the original CGAM cost equations [24]. To compare Alternative 1 with Alternative 2, investment data are determined using component cost equations in the form of Eq. (13) (required by Alternative 2 only). In the next three subsections, the modication of the original cost equations, the exergoeconomic analysis and the application of the improvement procedure to the CGAM system are reported. 4.1. Cost equations The principal costs of a thermal system are the capital investment, the operation and maintenance, and the fuel costs. The economic analysis performed here is a simplication of the revenue requirement method [25]. A simplied economic model is assumed, based on the capital recovery factor (CRF), and considering that the total capital investment (TCI) in a plant is given by the sum of all the purchased equipment costs (PEC) multiplied by a constant factor b [25]. This multiplying factor supposedly incorporates all direct and indirect costs of the plant. The total capital investment in a plant is, thus, given by X X X TCIk bPECk b PECk : 18 TCI
k k k

_ IN The investment cost rate for each component, Zk , is expressed by _ IN TCIk CRF ; Zk s CRF i1 in ; 1 in 1 19 20

where CRF is the capital recovery factor, i is the interest rate, and s is the yearly plant operating hours. Consistently with the CGAM problem, the operation and maintenance cost rate for each _ OM component, Zk , is now expressed by _ OM cCRF TCIk ; Zk s 21

where c is the maintenance factor, here assumed constant. _ _ IN The system total cost rate, excluding fuel costs, Z, is the summation over all components of Zk , _ OM Eq. (19), and Zk , Eq. (21),

2506

L.S. Vieira et al. / Energy Conversion and Management 45 (2004) 24952523

_ Z

X
k

_ Zk

P X _ IN Z OM k CRF1 cTCIk : _ Zk k s k

22

_ The system total cost, including fuel costs, Ctotal , is expressed by P X k CRF1 cTCIk _ _ _ _ cF mF LHV; Ctotal Zk CF s k

23

_ where cF , LHV and mF are, respectively, the specic cost, lower heating value and mass ow rate _ of the fuel. Because, for the CGAM problem, the system products are constant, the cost Ctotal as given by Eq. (23) is the objective function to be optimized, i.e. minimized. As previously mentioned, to permit comparison between results from Alternatives 1 and 2, the component cost equations in the form of Eq. (13) are derived from ts to data points obtained from the original CGAM economic model [24]. However, for the turbine and combustion chamber, to evaluate correctly the strong eect of the combustion chamber outlet temperature T4 on component costs, Eq. (13) has been slightly modied to read  n k ek _ mk EP;k : PECk Bk expBk1 T4 Bk2 24 1 ek The modied cost equations are determined by simulating the turbine and combustion chamber individually with IPSEpro and by calculating their exergetic eciencies and product ow rates for dierent input operating conditions. Table 1 shows the limits for the input data of all components of the CGAM cogeneration system, which are also applicable to the values of the decision variables in the improvement process. The values of the parameters Bk , Bk1 , Bk2 , nk and mk have been determined through ecient nonlinear estimation procedures available with the Statistica 5.0
Table 1 Limits for the input data of all components of the CGAM cogeneration system, applicable to the cost equations and improvement process Component Compressor Input variable RPc gAC _ ma (kg/s) _ ma (kg/s) T4 (C) DP (%) mg (kg/s) _ gGT T4 (C) RPc _ ms (kg/s) _ mg (kg/s) Value Minimum 7 0.700 25 25 800 5 25 0.700 600 6.4 10 125 Maximum 27 0.900 425 425 1400 5 425 0.900 1400 24.5 50 425

Combustor

Turbine

HRSG

L.S. Vieira et al. / Energy Conversion and Management 45 (2004) 24952523

2507

program [29], and are shown in Table 2 together with the correlation index (equal to 1 for a perfect t). 4.2. Exergoeconomic analysis An exergoeconomic analysis of the thermal system is performed in Step 1 of the methodology. Here, we analyse the CGAM cogeneration system, seen in Fig. 1 to have ve components and 12 ow streams. The denitions of fuel and product, presented in Table 3, and auxiliary cost
Table 2 Values of parameters for the purchased equipment cost equations Component Compressor Combustor Turbine HRSG Preheater Number of cases 65 29 98 37 24 Bk 51.9 299.9 181.3 98645.7 44839.1 Bk1 0.014 0.035 Bk2 )19.898 )53.799 nk 2.499 1.038 1.450 0.783 0.917 mk 1.002 1.002 1.004 0.649 0.371 Correlation index 0.999 0.996 0.999 0.971 0.992

Table 3 Denitions of fuel and product ow rates for the system components Component Compressor Combustor Turbine HRSG Preheater Product _ _ E2 E1 _ E4 _ _ E11 E12 _ _ E9 E8 _ _ E3 E2 Fuel _ E12 _ _ E10 E3 _ 4 E5 _ E _ _ E6 E7 _ _ E5 E6

Table 4 Matrix representation of system costs; the numeric entries form the 12 12 matrix A Stream 1 Compressor Combustor Turbine HRSG Preheater Air Fuel Water Turbine Turbine HRSG Preheater 1 0 0 0 0 1 0 0 0 0 0 0 2 )1 0 0 0 1 0 0 0 0 0 0 0 3 0 1 0 0 )1 0 0 0 0 0 0 0 4 0 )1 1 0 0 0 0 0 _ 1=E4 0 0 0 5 0 0 )1 0 1 0 0 0 _ 1=E5 0 0 _ 1=E5 6 0 0 0 1 )1 0 0 0 0 0 _ 1=E6 _ 1=E6 7 0 0 0 )1 0 0 0 0 0 0 _ 1=E7 0 8 0 0 0 1 0 0 0 1 0 0 0 0 9 0 0 0 )1 0 0 0 0 0 0 0 0 10 0 1 0 0 0 0 1 0 0 0 0 0 Shaft 11 1 0 )1 0 0 0 0 0 0 _ 1=E11 0 0 12 0 0 )1 0 0 0 0 0 0 _ 1=E12 0 0

2508

L.S. Vieira et al. / Energy Conversion and Management 45 (2004) 24952523

equations are the same as those proposed by Lazzaretto and Tsatsaronis [26]. Cost balance equations are implemented according to the theory of the exergetic cost by Lozano and Valero [6], such that the cost ow rates are determined by solving the linear system A C Z. The matrix A represents the cost balance of the system and is shown in Table 4. The last four rows of A cor_ respond to the auxiliary cost equations. The vector C fCi ; i 1; . . . ; 12g contains all the cost _ i ; i 1; . . . ; 12g contains all the external economic assignments ow rates, and the vector Z fZ for the system (component investment, operation and maintenance and fuel cost rates).
Table 5 Parameters (Par.) used in the iterative process Par. Dxi =xi p Value 1% 0.1 Dxi =xi Equation (4) and (5) (4) and (5) Description/function Increments on decision variables (sensitivity analysis) Identication of variables which aect total cost and system exergetic eciency

0.5

(6)(8)

Classication of components

2/3

(9) and (10)

Identication of predominant term in total cost (investment or exergy destruction)

a b n i s c

0.00001 6.32 10 12.7 8000 0.06

(1) (18) (20) (20) (19) (21)

End of iterative process Multiplying factor for purchased equipment cost Useful system life (years) Interest rate (%) Yearly plant operating hours Maintenance factor

L.S. Vieira et al. / Energy Conversion and Management 45 (2004) 24952523

2509

_ Exergy destruction ow rates (ED;k ), component exergetic eciencies (ek ), component product _ (cP;k ) and fuel (cF;k ) specic exergetic costs, exergy destruction cost ow rates (CD;k ), component _ k ) and relative cost dierences (rk ) are, respectively, calculated by the investment cost ow rates (Z following equations: _ _ _ _ ED;k EF;k EP;k EL;k ; 25

Table 6 Denition of Cases 14, corresponding to four dierent sets of initial values for the decision variables Decision variable RPc T3 (C) T4 (C) gAC gGT Initial value of the objective function (US$/h) Case 1 8.000 626.85 1126.85 0.800 0.800 1870.62 Case 2 20.000 525.00 1225.00 0.800 0.800 2132.76 Case 3 18.000 500.00 1200.00 0.900 0.900 2080.97 Case 4 12.000 500.00 1000.00 0.700 0.700 13863.39

Table 7 Final values for the decision variables obtained using a mathematical optimization procedure Decision variable RPc T3 (C) T4 (C) gAC gGT Number of function calls Final value of the objective function (US$/h) Case 1 7.000 634.36 1226.98 0.869 0.842 1509 1650.93 Case 2 8.195 608.02 1226.98 0.873 0.837 1414 1647.01 Case 3 8.201 607.71 1226.98 0.873 0.837 3235 1647.01 Case 4 8.201 609.01 1226.98 0.873 0.838 2414 1647.01

Math. (571)

Alt. 2 (1329)

Alt. 1 (706)

Total cost (US$/h)

1800 1750 1700 1650 1600 0 250 500 750 1000 1250 1500 1750

Number of function calls


Fig. 3. Optimization (Math.) and improvement (Alternatives 1 and 2) results for Case 1.

2510

L.S. Vieira et al. / Energy Conversion and Management 45 (2004) 24952523

ek

_ _ _ EP;k ED;k EL;k 1 ; _ _ EF;k EP;k _ CP;k ; _ EP;k

26

cP;k

27

Math. (659)

Alt. 2 (830)

Alt. 1 (568)

Total cost (US$/h)

1800 1750 1700 1650 1600 0 250 500 750 1000 1250 1500

Number of function calls

Fig. 4. Optimization (Math.) and improvement (Alternatives 1 and 2) results for Case 2.

Math. (630)

Alt. 2 (683)

Alt. 1 (628)

Total cost (US$/h)

1800 1750 1700 1650 1600 0 500 1000 1500 2000 2500 3000

Number of function calls

Fig. 5. Optimization (Math.) and improvement (Alternatives 1 and 2) results for Case 3.

Math. (1704)

Alt. 2 (1772)

Alt. 1 (2561)

Total cost (US$/h)

1800 1750 1700 1650 1600 0 500 1000 1500 2000 2500 3000 3500

Number of function calls

Fig. 6. Optimization (Math.) and improvement (Alternatives 1 and 2) results for Case 4.

L.S. Vieira et al. / Energy Conversion and Management 45 (2004) 24952523

2511

cF;k

_ CF;k ; _ EF;k

28 29 30 31

_ _ CD;k cF;k ED;k ; _ _ IN _ OM CRF1 cTCIk ; Zk Zk Zk s rk _ cP;k cF;k 1 ek Zk : _ cF;k ek cF;k EP;k

_ The exergetic cost ow rate Ci is given by the product of the respective specic exergetic costs and the exergy ow rates. Optimal values of exergetic eciency and relative cost dierence for each main and secondary component are determined considering the component thermoeconomically isolated from the system and are given by the following equations [25]: eOPT k 2 1 ; 1 Fk 32

31=nk 1 CRF1 ck bBk nk 5 Fk 4 ; _ 1m scF;k EP;k k

33

Table 8 Application of Alternative 1 to Case 1: cost rates resulting from an exergoeconomic analysis performed at the beginning of iterations 13 _ _ _ _ _ _ CP (US$/h) CF (US$/h) Z (US$/h) CD (US$/h) Z CD (US$/h) Iteration 1 Compressor Combustor Turbine HRSG Preheater Iteration 2 Compressor Combustor Turbine HRSG Preheater Iteration 3 Compressor Combustor Turbine HRSG Preheater 831.78 3161.93 1924.52 567.96 907.15 1110.07 2898.54 1971.01 695.97 652.13 1069.94 2724.59 1942.92 671.56 516.28 1160.35 3126.52 1841.95 433.78 743.89 931.35 2876.99 1754.73 541.25 562.62 952.60 2703.75 1729.63 522.76 432.04 67.61 35.40 82.57 134.18 163.26 70.41 21.56 216.28 154.71 89.50 79.62 20.84 213.29 148.80 84.25 122.56 508.50 135.23 145.52 81.59 80.66 518.76 56.91 160.31 87.94 81.01 495.94 57.93 160.31 66.42 190.17 543.91 217.80 279.70 244.84 151.08 540.32 273.19 315.03 177.44 160.63 516.78 271.22 309.12 150.67

2512
OPT rk

L.S. Vieira et al. / Energy Conversion and Management 45 (2004) 24952523

nk 1 Fk : nk

34

4.3. Improvement procedure The proposed methodology (Alternatives 1 and 2) described in Section 3 requires an auxiliary mathematical optimization procedure to modify the main decision variables in the iterative process. While any conventional mathematical method could be chosen, the easy to implement exible polyhedron method [27] has been used in this work, with slight modications in order to comply with the allowable range of values of the decision variables imposed by the cost equations. Table 5 shows the values of the parameters used for the selection of main components and main decision variables in the proposed iterative process, and also, it indicates the values of the economic parameters used in all our calculations. We point out that, dierently from the CGAM problem [24], the value of the parameter b is 6.32, as recommended by Bejan et al. [25] for a new system. Finally, the CGAM system is simple enough to permit mathematical optimization. To compare the proposed improvement procedure with a mathematical optimization method, we have also chosen the exible polyhedron method to act on the same decision variables to optimize the whole

Table 9 Application of Alternative 1 to Case 1: modied structural bond coecients r1 (Eq. (11)) and r2 (Eq. (12)) at the beginning of iterations 13 Decision variable RPc r1 Iteration 1 Compressor Combustor Turbine HRSG Preheater Iteration 2 Compressor Combustor Turbine HRSG Preheater Iteration 3 Compressor Combustor Turbine HRSG Preheater )18.83 1.06 )1.20 0.90 0.17 9.54 1.46 )2.27 0.58 0.30 9.20 1.35 )1.97 0.64 0.25 r2 )0.11 0.19 )0.30 0.38 0.81 )0.38 0.55 3.95 1.12 0.34 )0.28 0.48 4.60 0.91 0.29 T3 (C) r1 )1.54 1.24 )1.55 0.28 0.36 )2.11 2.11 )1.98 0.24 1.68 )2.11 1.98 )1.98 0.25 2.10 r2 1.04 )0.74 1.13 )1.79 0.12 0.30 )0.21 0.75 )0.30 0.04 0.43 )0.27 0.98 )0.34 0.04 T4 (C) r1 )0.28 3.19 )0.29 0.13 0.14 )0.81 3.86 )0.78 0.21 0.49 )0.72 3.68 )0.69 0.21 0.42 r2 0.34 )17.63 0.60 )0.81 0.26 )0.02 0.14 0.26 0.04 )0.04 0.02 )0.12 )0.34 )0.03 0.02 gGT r1 1.79 1.95 0.43 0.69 0.58 )0.48 )1.26 0.11 0.19 0.72 )0.60 )1.77 0.12 0.23 0.67 r2 0.47 0.67 0.92 )5.67 3.39 20.26 3.72 0.22 4.20 )31.29 )22.24 4.51 0.22 5.20 756.64 gAC r1 0.13 )7.13 )9.02 0.56 )0.55 0.00 0.01 0.01 )0.01 0.00 0.00 0.00 0.00 0.00 0.00 r2 0.16 )1.50 )0.05 0.06 0.95 0.23 0.94 )1.47 1.53 0.91 0.23 0.99 )1.51 1.66 0.90

L.S. Vieira et al. / Energy Conversion and Management 45 (2004) 24952523

2513

CGAM system. We remark that the mathematical strategy does not make use of a preceding exergoeconomic analysis of the system.

5. Results and discussion The results obtained with the application of the proposed exergoeconomic improvement procedure, Alternatives 1 and 2, to the CGAM system are now presented and discussed in detail and are compared with those reached using a mathematical optimization method. As remarked in Section 4.1, to establish the same basis for comparison, investment costs are determined with equations in the form of Eqs. (13) and (24). Because the physical, thermodynamic and economic models used in this paper are dierent from those of the original CGAM problem, the present results for the optimized costs and values of the decision variables should not be compared to those in Ref. [24].
Table 10 Application of Alternative 1 to Case 1: results at the end of iteration 1 RPc T3 (C) T4 (C) gGT gAC Total cost (US$/h) 1870.62 NFa

Initial value Eect of decision variable on system Main components Selection of decision variables Improved results Secondary component Selection of decision variables Improved results Secondary component Selection of decision variables Improved results

8.000

626.85

1126.85

0.800

0.800

No eect Only Cost and Cost and Cost and eciency eciency eciency eciency Combustor _ CD is predominant 8.000 HRSG p 602.56 p 1226.95 p 0.900 0.800 p 8.000 Preheater 602.56 p 8.000 585.65 1226.95 p 1226.97 0.900 p 0.900 0.831 p 0.838 1667.26 290 1670.32 164 1683.30 129

Remainder components Turbine Compressor Last improvement Selection of decision variables Improved results Final value 8.000 585.65 1226.97 0.900 0.838 1667.26 290

selected variables. a NFcumulative number of function calls.

2514

L.S. Vieira et al. / Energy Conversion and Management 45 (2004) 24952523

The chosen decision variables are: compressor compression ratio (RPc), compressor outlet temperature (T3 ), turbine inlet temperature (or combustor outlet temperature, T4 ), compressor isentropic eciency (gAC ) and turbine isentropic eciency (gGT ). The sophisticated built in thermodynamic balance equations solvers of the simulator permit reduction of the number of decision variables of the optimization problem to only ve. Furthermore, in the proposed improvement procedure, these ve decision variables are not simultaneously modied throughout the iterative process. The number of function calls, i.e. objective function evaluations, is equal to the number of calls to the simulator. It is observed that a conventional mathematical approach applied directly to this problem, without the use of a simulator and adopting simplied mass and energy balances as restrictions to the optimization problem, would require the choice of more than 25 decision variables. Integration of the mathematical method to the simulator permits reduction of the number of decision variables to ve; however, all ve of them must be simultaneously modied by the method throughout the entire process.

Table 11 Application of Alternative 1 to Case 1: results at the end of iteration 2 RPc T3 (C) T4 (C) gGT gAC Total cost (US$/h) 1667.26 NFa

Initial value Eect of decision variable on system Main components Selection of decision variables Improved results Secondary component Selection of decision variables Improved results Secondary component Selection of decision variables Improved results

8.000

585.65

1226.97

0.900

0.838

290

Only Only eciency eciency Combustor _ CD is predominant HRSG p 9.287 p 9.287 Preheater p 9.287 566.04 p 566.04 p 566.04

Cost and Cost and Only eciency eciency eciency p 1226.97 p 1226.97 p 1226.98 0.900 p 0.900 p 0.900 0.833 1664.33 614 p 0.834 p 0.834 1664.33 542 1664.33 470

Remainder components Turbine Compressor Last improvement Selection of decision variables Improved results Final value 9.287 566.04 1226.98 0.900 0.833 1664.33 614

selected variables. a NFCumulative number of function calls.

L.S. Vieira et al. / Energy Conversion and Management 45 (2004) 24952523

2515

Results are presented for four dierent cases, indicated in Table 6. Each case possesses its own set of initial values for the decision variables and a corresponding initial value for the objective function. For each case, the improvement process is executed using Alternative 1 and Alternative 2, and an optimization is eected with the chosen mathematical method. In all cases, to reduce signicantly the possibility of achieving local minima, the iterative process is restarted with an initial polyhedron with a dierent conguration from that used in the preceeding iterations. Note that the high initial value of the objective function for Case 4 is indicative that the initial values of the decision variables are thermodynamically inconsistent. Therefore, the chosen set of values is considered unfeasible for the optimization problem. A penalty function is added to the objective function whenever unfeasibility is observed. Table 7 shows the nal results obtained using the mathematical method of the exible polyhedron integrated with the IPSEpro program. The objective function minimum is 1647.01 US$/h, obtained for Cases 2, 3 and 4. By comparing Cases 1 and 2, one observes that large deviations in RPc may correspond to small variations of the objective function. The graphs in Figs. 36 show
Table 12 Application of Alternative 1 to Case 1: results at the end of iteration 3 RPc T3 (C) T4 (C) gGT gAC Total cost (US$/h) 1664.33 NFa

Initial value Eect of decision variable on system Main components Selection of decision variables Improved results Secondary component Selection of decision variables Improved results Secondary component Selection of decision variables Improved results Remainder components Last improvement Selection of decision variables Improved results Final value Combustor _ CD is predominant

9.287

566.04

1226.98

0.900

0.833

614

No eect Only Only eciency cost p 9.287 HRSG 566.04 1226.98

Cost and Only eciency eciency

0.900 p

0.833 p

1664.33

614

9.287 Turbine _ Z is predominant 9.287 Preheater Compressor

566.04 p 566.04

1226.98

0.900 p

0.833

1664.33

647

1226.98

0.900

0.833

1664.33

671

p 9.287 9.287 566.04 566.04 1226.98 1226.98 0.900 0.900 0.833 0.833 1664.33 1664.33 706 706

selected variables. a NFcumulative number of function calls.

2516

L.S. Vieira et al. / Energy Conversion and Management 45 (2004) 24952523

the evolution of the objective function as the optimizing process progresses. Also, the legends indicate the number of function calls before restart of the iterative process for the mathematical method (and for both alternatives as well). It is seen in the graphs that the objective function is substantially reduced after about 700 function calls for all cases. Furthermore, the restart provides additional signicant objective function reduction (more than 5%) for Case 3. It is veried that the amount of time consumed by the simulator is the major portion of time of the entire optimizing process. Even for the simple CGAM system, the mathematical approach may fail to reach the minimum if the restart is not performed, as evidenced by Case 3. The results obtained with application of the proposed method to the CGAM system are presented below in great detail for Case 1, rst for Alternative 1 and second for Alternative 2. Because similar operations and considerations apply to the other cases, only the nal results are shown for Cases 24. Application of Alternative 1 to Case 1 requires four iterations. Table 8 shows the values of the _ _ _ _ _ _ cost rates CP , CF , Z, CD and Z CD resulting from an exergoeconomic analysis performed at the beginning of the rst three iterations. The fourth iteration corresponds to the restart, and because, in this case, it does not provide further objective function reduction, it is not shown in the tables. Table 9 displays the calculated modied structural bond coecients r1k;xi and r2k;xi for each component and decision variable at the beginning of the three iterations. Tables 1012 show, respectively, for iterations 13, the evolution of the hierarchical classication of components, the selection of main decision variables and the improved results. It is seen that for Case 1, the combustor is always the main component, the HRSG is a secondary component and the compressor is a remainder component. The preheater is secondary in the rst and second iterations,
Table 13 Application of Alternative 2 to Case 1: cost rates resulting from an exergoeconomic analysis performed at the beginning of iterations 1, 2 and 3 _ _ _ _ _ _ CP (US$/h) CF (US$/h) Z (US$/h) CD (US$/h) Z CD Iteration 1 Compressor Combustor Turbine HRSG Preheater Iteration 2 Compressor Combustor Turbine HRSG Preheater Iteration 3 Compressor Combustor Turbine HRSG Preheater 831.78 3161.93 1924.52 567.96 907.15 1063.15 2859.50 1872.18 689.18 582.48 1046.31 2941.93 1896.27 672.84 731.25 1160.35 3126.52 1841.95 433.78 743.89 892.41 2838.42 1705.46 561.79 509.86 925.20 2917.87 1744.88 520.64 627.21 67.61 35.40 82.57 134.18 163.26 83.37 21.08 166.72 127.39 72.62 75.24 24.06 151.39 152.20 104.05 122.56 508.50 135.23 145.52 81.59 73.23 533.00 66.12 196.86 93.50 80.00 505.91 73.15 156.49 90.34 190.17 543.91 217.80 279.70 244.84 156.60 554.08 232.84 324.25 166.12 155.25 529.97 224.54 308.69 194.38

L.S. Vieira et al. / Energy Conversion and Management 45 (2004) 24952523

2517

and remainder in the third. The turbine is remainder in the rst two iterations, and secondary in the third. It is worth noting that the selected decision variables associated with each main and secondary component change between iterations. In each iteration, not all of the decision variables are selected simultaneously. Also, it is observed that towards the end of the iterative process (Table 12), just a few decision variables are selected. The minimum obtained is 1664.32 US$/h, only 0.81% higher than 1650.93 US$/h (see Table 7), the value obtained using a mathematical procedure beginning from the same initial values for the decision variables. Application of Alternative 2 to Case 1 requires ve iterations, only the rst three of which are _ _ _ _ _ _ relevant to be shown here. Table 13 shows the values of the cost rates CP , CF , Z, CD , and Z CD

Table 14 Application of Alternative 2 to Case 1: sensitivity coecients wdrk;x (Eq. (16)) and wdek;x (Eq. (17)) at the beginning of i i iterations 1, 2 and 3 Iteration 1 wdrk;xi (%) Variable RPc Compressor Combustor Turbine HRSG Preheater Variable T3 Compressor Combustor Turbine HRSG Preheater Variable T4 Compressor Combustor Turbine HRSG Preheater Variable gGT Compressor Combustor Turbine HRSG Preheater Variable gAC Compressor Combustor Turbine HRSG Preheater 0.91 )0.39 4.35 )3.05 9.93 0.31 )1.85 8.40 )32.63 28.74 )4.25 1.54 )24.86 43.02 )27.58 )4.48 1.23 )98.08 )36.34 17.74 )5.46 1.37 27.09 )18.33 )5.13 wdek;xi (%) 4.21 )0.17 2.31 )1.48 4.55 )1.48 )1.57 4.04 )18.43 10.61 )1.63 0.89 )13.78 19.90 )13.04 )4.30 0.26 )86.49 )21.03 7.50 39.39 0.29 12.43 )10.37 )2.80 Iteration 2 wdrk;xi (%) )3.41 )0.26 )0.62 )3.18 )12.42 )0.56 )1.12 )0.58 )12.91 )20.41 16.30 0.64 3.01 17.74 28.42 17.52 1.46 34.23 )17.96 )12.03 25.53 1.40 )3.04 )5.74 14.67 wdek;xi (%) 1.12 )0.12 )0.16 )1.61 )6.47 )0.32 )0.99 )0.34 )7.07 )12.05 )0.50 0.30 0.92 8.75 14.47 )2.87 0.25 10.11 )10.45 )6.81 15.19 0.24 )1.83 )3.69 6.65 Iteration 3 wdrk;xi (%) )6.41 )0.27 )0.71 )10.06 116.97 )1.20 )1.30 )0.93 )47.52 290.08 30.01 0.87 3.59 69.30 )96.86 32.15 1.50 36.63 )61.15 151.60 43.34 1.41 )3.41 )26.77 )66.20 wdek;xi (%) 1.39 )0.13 )0.21 )5.22 47.02 )0.55 )1.18 )0.54 )28.20 93.91 )0.53 0.49 1.21 30.89 )117.99 )3.16 0.28 11.55 )38.75 57.24 17.89 0.26 )2.01 )15.34 )42.06

2518

L.S. Vieira et al. / Energy Conversion and Management 45 (2004) 24952523

resulting from an exergoeconomic analysis performed at the beginning of the rst three iterations. Table 14 displays the calculated sensitivity coecients wdek;x and wdrk;x for each component and i i decision variable at the beginning of the iterations. Tables 1517 show, respectively, for iterations 13, the evolution of the hierarchical classication of components, the selection of decision variables and the improved results. It is seen that, for Case 1, similarly to Alternative 1, the combustor is always the main component, the HRSG is a secondary component, and the compressor is a remainder component. The preheater is secondary in the rst iteration, and the turbine is secondary in the second and third iterations. The selected decision variables associated with each main and secondary component also change between iterations. In each iteration, not all of the decision variables are selected simultaneously. It is observed that for all iterations (Tables 15 17), just a few decision variables are selected. The minimum obtained is 1647.16 US$/h, 0.23% lower than 1650.93 US$/h (see Table 7).

Table 15 Application of Alternative 2 to Case 1: results at the end of iteration 1 RPc T3 (C) T4 (C) gGT gAC Total cost (US$/h) 1870.62 NFa

Initial value Eect of decision variable on system Main components Selection of decision variables Improved results Secondary component Selection of decision variables Improved results Secondary component Selection of decision variables Improved results Combustor _ CD is predominant

8.000

626.85

1126.85

0.800

0.800

No eect Only eciency p 8.000 HRSG 636.85 p 8.000 Preheater 557.73

Cost and Cost and Cost and eciency eciency eciency

1126.85

0.800 p

0.800 p 0.840 p

1868.86

36

1126.85 p

0.881

1705.58

264

8.000

557.73

1226.98

0.881

0.847

1663.98

350

Remainder components Turbine Compressor Last improvement Selection of decision variables Improved results Final value 8.000 557.73 1226.98 0.881 0.847 1663.98 350

selected variables. a NFCumulative number of function calls.

L.S. Vieira et al. / Energy Conversion and Management 45 (2004) 24952523 Table 16 Application of Alternative 2 to Case 1: results at the end of iteration 2 RPc T3 (C) T4 (C) gGT gAC Total cost (US$/h) 1663.98

2519

NFa

Initial value Eect of decision variable on system Main components Selection of decision variables Improved results Secondary component Selection of decision variables Improved results Secondary component Selection of decision variables Improved results Combustor _ CD is predominant

8.000

557.73

1226.98

0.881

0.847

350

No eect Cost and Only eciency cost p 8.000 HRSG 600.23 p 8.000 Turbine _ Z is predominant 8.000 611.60 p 611.60 1226.98 1226.98 1226.98

Only Only eciency eciency

0.881 p 0.872

0.847 p 0.838 p

1650.19

379

1647.24

511

0.872

0.838

1647.24

581

Remainder components Preheater Compressor Last improvement Selection of decision variables Improved results Final value p 8.000 8.000 611.60 611.60 1226.98 1226.98 0.872 0.872 0.838 0.838 1647.24 1647.24 586 586

selected variables. a NFcumulative number of function calls.

Figs. 36 show, respectively, for Cases 14, a comparison between the results obtained using the mathematical optimization procedure and the proposed method, Alternatives 1 and 2. In general terms, with one exception or another, we can make the following remarks. First, signicant improvement of the objective function is reached with no more than 700 function calls. Second, Alternative 1 exhibits faster improvement than Alternative 2 in the early stages of the iteration process, and third, the mathematical optimization method reduces the objective function at a faster rate than Alternatives 1 and 2. Nevertheless, the initial faster reduction does not mean a smaller number of calls to the simulator to reach the minimum system total cost, and also, it does not mean smaller nal values for the objective function. Table 18 summarizes the nal values for the decision variables and the objective function, the total number of function calls and the relative dierence between each nal value of the objective function and the reference value. The reference value is the smallest minimum achieved, obtained with the mathematical approach applied to Case 4 (1647.01 US$/h). Based on the total number of

2520

L.S. Vieira et al. / Energy Conversion and Management 45 (2004) 24952523

Table 17 Application of Alternative 2 to Case 1: Results at the end of iteration 3 RPc T3 (C) T4 (C) gGT gAC Total cost (US$/h) 1647.24 NFa

Initial value Eect of decision variable on system Main components Selection of decision variables Improved results Secondary component Selection of decision variables Improved results Secondary component Combustor _ CD is predominant HRSG

8.000

611.60

1226.98

0.872

0.838

586

Only Only eciency eciency p 8.141 p 8.141 Turbine _ Z is predominant p 610.01 p 610.01

Cost and Cost and Only eciency eciency eciency

1226.98

0.872 p

0.838 p 0.838

1647.16

662

1226.98

0.872

1647.16

842

Selection of decision variables Improved results Remainder components Preheater Compressor Last improvement Selection of decision variables Improved results Final value

p 8.141

p 610.01 1226.98 0.872

p 0.838 1647.16 958

8.141

610.01

1226.98

0.872

0.838

1647.16

958

selected variables. a NFcumulative number of function calls.

function calls, Alternative 1 is the best for Cases 1, 2 and 3; relative to the mathematical optimization, Alternative 2 is better for Cases 2 and 3. It is veried in Table 18 that all nal values of the objective function match the reference value to within a discrepancy below 1.4%. Considering a 90% condence interval, it is estimated that deviations between the nal value of the objective function and the reference value will be less than 2.5% for both alternatives. On physical grounds, it is observed that large variations in the nal values of the RPc are obtained. This is due to the fact that RPc has a relatively small eect near the minimum of the objective function. On the other hand, no variation is obtained for the combustion chamber outlet temperature (T4 ). In fact, high outlet temperatures imply less exergy destruction in the combustor. Because the combustor is always classied as the main component in all iterations, for both Alternatives 1 and 2, no dierence should really be expected for the nal values of T4 .

L.S. Vieira et al. / Energy Conversion and Management 45 (2004) 24952523

2521

Table 18 Final results obtained using the mathematical method (Math.), Alternative 1 (Alt. 1) and Alternative 2 (Alt. 2) applied to Cases 14 Case Method Values of decision variables RPc 1 Initial Math. Alt. 1 Alt. 2 Initial Math Alt. 1 Alt. 2 Initial Math. Alt. 1 Alt. 2 Initial Math. Alt. 1 Alt. 2 8.000 7.000 9.287 8.142 20.000 8.195 9.779 9.779 18.000 8.201 11.387 11.416 12.000 8.201 8.193 7.000 T3 (C) 626.85 634.36 566.04 610.01 525.00 608.02 561.64 560.37 500.00 607.71 548.88 547.39 500.00 609.01 608.46 634.42 T4 (C) 1126.85 1226.98 1226.98 1226.98 1225.00 1226.98 1226.98 1226.98 1200.00 1226.98 1226.98 1226.98 1000.00 1226.98 1226.98 1226.98 gGT 0.800 0.869 0.900 0.873 0.800 0.873 0.900 0.900 0.900 0.873 0.900 0.900 0.700 0.873 0.873 0.869 gAC 0.800 0.842 0.834 0.838 0.800 0.837 0.832 0.833 0.900 0.837 0.825 0.828 0.700 0.838 0.838 0.842 Objective function (US$/h) 1870.62 1650.93 1664.32 1647.16 2132.76 1647.01 1664.58 1664.57 2080.97 1647.01 1669.58 1669.56 13863.39 1647.01 1647.01 1650.78 Number of calls Relative dierence (%) 0.24 1.05 0.01 0.00 1.07 1.07 0.00 1.37 1.37 0.00 0.00 0.23

1509 798 1700 1414 666 1005 3235 688 844 2414 2913 3424

6. Conclusions Thermodynamic process simulation programs are widely used in complex thermal systems projects, but most of them do not incorporate second law optimization techniques. Likewise, available optimization techniques do not usually contemplate integration with simulation programs. In this paper, a new iterative methodology for exergoeconomic improvement of thermal systems integrated with a process simulator is proposed, so as to be applicable to complex systems. For automated implementation of the method, quantitative and qualitative criteria are proposed for an adequate choice of decision variables for system improvement. Two alternatives are identied and implemented: Alternative 1, based on modied structural bond coecients, and Alternative 2, based on sensitivity coecients for two exergoeconomic variables, the relative cost dierence and exergetic eciency. The proposed methodology is applied to the benchmark CGAM cogeneration system, and the results obtained are compared to those reached using the mathematical optimization procedure of the exible polyhedron method. The results of both Alternatives 1 and 2 closely match those of the mathematical optimization, with dierences in the values of the objective function below 1.4%. Relative to the mathematical optimization, the proposed method has the important advantage of not manipulating all decision variables simultaneously, thus leading, in general, to a smaller number of function calls. Comparison between Alternatives 1 and 2 reveals that Alternative 1 exhibits a faster improvement in the early stages of the iteration process and a reduced number of function calls in

2522

L.S. Vieira et al. / Energy Conversion and Management 45 (2004) 24952523

the majority of cases. However, it is impossible to ascertain, in general, which of the alternatives calls the simulator less. Alternative 2 reveals less average deviations from the reference value but requires the use of cost equations in the form of Eq. (13), in which cost is expressed as a function of component product exergy ow and component exergetic eciency. Therefore, Alternative 1 is better suited to situations for which the cost equations have an arbitrary form. The principal dierence between the proposed method and the mathematical method is that in the proposed method, some system components and decision variables are not considered in each iteration, due to the knowledge gained from a preceding exergoeconomic analysis of the system. Despite the fact that not all decision variables are modied at the same time in the improvement process, the nal values of the objective function are very close to the minimum, and sometimes even better than those obtained with the mathematical approach. The proposed method exploits the computational power of a process simulator and is, therefore, suitable for application to complex systems with a large number of decision variables. In fact, the proposed method does not require user intervention and does not enlarge the optimization problem unnecessarily with restrictions and dependent variables to solve the thermodynamic balance equations of the system.

Acknowledgements The authors would like to gratefully acknowledge the support of CNPq (Grant 500086/2003-6, 521002/97-4) and CAPES (Grant PICDT/UFES 23068.003437/98-85).

References
[1] Wall G. Exergy, ecology and democracyconcepts of a vital society. In: Szargut J et al., editors. Proceedings of the International Conference on Energy Systems and Ecology (ENSEC 93). Cracow, Poland, July 59; 1993. p. 11121. [2] Rosen MA, Dincer I. Exergy as the conuence of energy, environment and sustainable development. Exergy, Int J 2001;1(1):313. [3] Tsatsaronis G. Thermoeconomic analysis and optimization of energy systems. Progress Energy Combust Sci 1993;19(3):22757. [4] Frangopoulos CA. Thermoeconomic functional analysis: a method for optimal design or improvement of complex thermal systems. Ph.D. Thesis, Georgia Institute of Technology, Georgia; 1983. [5] von Spakovsky MR. A practical generalized analysis approach to the optimal thermoeconomic design and improvement of real-world thermal systems. Ph.D. Thesis, Georgia Institute of Technology, Georgia; 1986. [6] Lozano MA, Valero A. Theory of the exergetic cost. Energy, Int J 1993;18(9):93960. [7] Tsatsaronis G. Exergoeconomics: is it only a new name? Chem Eng Technol 1996;19:1639. [8] Hua B, Chen QL, Wang P. A new exergoeconomic approach for analysis and optimization of energy systems. Energy 1997;22(11):10718. [9] Kim S-M, Oh S-D, Kwon Y-H, Kwak H-Y. Exergoeconomic analysis of thermal systems. Energy 1998;23(5):393 406. [10] Zhang G, Hua B, Chen Q. Exergoeconomic methodology for analysis and optimization of process systems. Comput Chem Eng 2000;24:6138. [11] Kwon Y-H, Kwak H-Y, Oh S-D. Exergoeconomic analysis of gas turbine cogeneration systems. Exergy, Int J 2001;1(1):3140. [12] Sciubba E. Beyond thermoeconomics? The concept of extended exergy accounting and its application to the analysis and design of thermal systems. Exergy, Int J 2001;1(2):6884.

L.S. Vieira et al. / Energy Conversion and Management 45 (2004) 24952523

2523

[13] Rosen MA, Dincer I. Exergy-costenergy-mass analysis of thermal systems and processes. Energy Conversion Manage 2003;44:163351. [14] El-Sayed YM, Gaggioli RA. A critical review of second law costing methods I: background and algebraic procedures. ASME J Energy Resour Technol 1989;111:17. [15] Gaggioli RA, El-Sayed YM. A critical review of second law costing methodsII: calculus procedures. ASME J Energy Resour Technol 1989;111:815. [16] Lazzaretto A, Tsatsaronis G. On the calculation of eciencies and costs in thermal systems. In: Aceves SM et al., editors. Proceedings of the ASME Advanced Energy Systems Division 1999, AES-Vol. 39. New York: ASME; 1999. p. 42130. [17] Cziesla F, Tsatsaronis G. Iterative exergoeconomic evaluation and improvement of thermal power plants using fuzzy inference systems. Energy Conver Manage 2002;43:153748. [18] Tsatsaronis G, Park M-H. On avoidable and unavoidable exergy destructions and investment costs in thermal systems. Energy Conver Manage 2002;43:125970. [19] Evans RB. Thermoeconomic isolation and essergy analysis. Energy, Int J 1980;5(8-9):80521. [20] El-Sayed YM. A second-law-based optimization: part 1methodology. ASME J Eng Gas Turbines Power 1996;118:6937. [21] Benelmir R, Feidt M. A comparative synthesis of exergo-economic optimization: the IEEB method. In: Duncan AB et al., editors. Proceedings of the ASME Advanced Energy Systems Division, AES-Vol. 36. New York: ASME; 1996. p. 44555. [22] Jaluria Y. Design and optimization of thermal systems. New York: McGraw-Hill; 1997. [23] Kotas TJ. The exergy method of thermal plant analysis. Florida: Krieger Publishing Company; 1995. [24] Tsatsaronis G, guest editor. Special Issue, Invited Papers on Exergoeconomics. Energy Int J 1994;19(3):279381. [25] Bejan A, Tsatsaronis G, Moran M. Thermal design and optimization. New York: John Wiley and Sons, Inc; 1996. [26] Lazzaretto A, Tsatsaronis G. On the quest for objective equations in exergy costing. In: Ramalingam ML et al., editors. Proceedings of the ASME Advanced Energy Systems Division, AES-Vol. 37. New York: ASME; 1997. p. 197210. [27] Himmelblau DM. Applied nonlinear programming. New York: McGraw-Hill; 1972. [28] SimTech. IPSEpro user documentationVersion 3.1. Austria; 2000. [29] StatSoft, Inc. STATISTICA for Windows computer program manual. Tulsa, OK; 1996.

Вам также может понравиться