F

MB

IR

h

••••

a

ARRAA

KSDPC

1

istbtdmw[o

h0

ARTICLE IN PRESSG ModelUSION-7473; No. of Pages 4

Fusion Engineering and Design xxx (2014) xxx–xxx

Contents lists available at ScienceDirect

Fusion Engineering and Design

jo ur nal home p age: www.elsev ier .com/ locate / fusengdes

odeling and sizing of the heat exchangers of a new supercritical CO2

rayton power cycle for energy conversion for fusion reactors

.P. Serrano, A. Cantizano, J.I. Linares ∗, B.Y. Moratillaafael Marino Chair on New Energy Technologies, Comillas Pontifical University, Alberto Aguilera, 25, 28015 Madrid, Spain

i g h l i g h t s

We propose a procedure to model the heat exchangers of a S-CO2 Brayton power cycle.Discretization in sub-heat exchangers is performed due to complex behavior of CO2.Different correlations have been tested, verifying them with CFD when necessary.Obtained sizes are agree with usual values of printed circuit heat exchangers.

r t i c l e i n f o

rticle history:eceived 23 August 2013eceived in revised form 21 March 2014ccepted 9 April 2014vailable online xxx

eywords:upercritical CO2 cycleual coolant blanket

a b s t r a c t

TECNO FUS is a research program financed by the Spanish Government to develop technologies relatedto a dual-coolant (He/Pb–Li) breeding blanket design concept including the auxiliary systems for a futurepower reactor (DEMO). One of the main issues of this program is the optimization of heat recovery fromthe reactor and its conversion into electrical power. This paper is focused on the methodology employedfor the design and sizing of all the heat exchangers of the supercritical CO2 Brayton power cycle (S-CO2) proposed by the authors. Due to the large pressure difference between the fluids, and also to theircompactness, Printed Circuit Heat Exchangers (PCHE) are suggested in literature for these type of cycles.Because of the complex behavior of CO2, their design is performed by a numerical discretization into

rinted Circuit Heat ExchangerFD

sub-heat exchangers, thus a higher precision is reached when the thermal properties of the fluids varyalong the heat exchanger. Different empirical correlations for the pressure drop and the Nusselt numberhave been coupled and assessed. The design of the precooler (PC) and the low temperature recuperator(LTR) is also verified by simulations using CFD because of the near-critical behavior of CO2. The size of allof the heat exchangers of the cycle have been assessed.

© 2014 Elsevier B.V. All rights reserved.

. Introduction

Hejzlar et al. [1] have identified the recompression supercrit-cal CO2 Brayton cycle (S-CO2) as a promising power conversionystem for some Generation IV fission reactors (sodium fast reac-or). This type of cycle has been also proposed for fusion reactorsecause, besides of the aspects of high efficiency, CO2 exhibits cer-ain advantages in terms of tritium recovery. Although it can beissociated to CO under heavy neutron irradiation and tritium per-eation into CO2 could react with it, its separation is easier than for

Please cite this article in press as: I.P. Serrano, et al., Modeling and spower cycle for energy conversion for fusion reactors, Fusion Eng. Des

ater [2]. On this basis, a domestic R&D program called TECNO FUS3] was launched in Spain in 2009 to support technological devel-pments related to a specific concept of dual-coolant (He/Pb–Li)

∗ Corresponding author. Tel.: +34 91 542 28 00; fax: +34 91 559 65 69.E-mail address: linares@upcomillas.es (J.I. Linares).

ttp://dx.doi.org/10.1016/j.fusengdes.2014.04.039920-3796/© 2014 Elsevier B.V. All rights reserved.

breeding blanket for fusion reactors based on Model C configurationfor fusion Power Plant Concept (PPCS) [4].

The S-CO2 cycle has been already analyzed by Angelino [5]. It is aBrayton cycle with CO2 as working fluid and with two recuperators,one at low temperatures (LTR) and another at high temperatures(HTR). Serrano et al. [6] developed a variant of S-CO2 cycle, des-ignated as REC3 (three recuperators), with one more recuperator(BBR) and with an efficiency higher than 46%. Further investigationslead to a new improved one, designated as REC2, in which HTR issuppressed and thus BBR assumes its thermal load. So, this newdesign includes two recuperators, as in classical S-CO2, but withthe same efficiency than in REC3 [7]. Fig. 1 shows the new recuper-ator BBR included in both REC2 and REC3 designs and integrated

izing of the heat exchangers of a new supercritical CO2 Brayton. (2014), http://dx.doi.org/10.1016/j.fusengdes.2014.04.039

in the classical S-CO2 cycle. The REC2 design has been proposed asthe power cycle for TECNO FUS fusion reactor.

The main goal of this paper is to describe the methodology fol-lowed for the design of the different heat exchangers of the REC2

ARTICLE IN PRESSG ModelFUSION-7473; No. of Pages 4

2 I.P. Serrano et al. / Fusion Engineering and Design xxx (2014) xxx–xxx

PC

Main

compressor

Auxiliary

compressor Turbine

LDIV LM

BNK

BBRREC2 & REC3

no in classical layout

LTR

Generator

HDIV

HTR

(no in REC2)

Fv

cCstieiaPhw

t(Pot

drtoCtp

2

tssctaststiMdt(c

Table 1Heat exchangers characteristics.

Fluid Tinlet (◦C) Toutlet (◦C) Power (MW)

LM Li-Pb 700 480 1976CO2 461 586

BNK He 400 300 793CO2 125 395

LDIV He 700 566 328CO2 461 586

HDIV He 800 700 247CO2 586 600

Precooler CO2 59 30 1742H2O 20 30

BBR CO2 477 130 5933CO2 125 472

where Di is the hydraulic diameter of the channel, fi is the Darcy

ig. 1. Supercritical CO2 Brayton cycle including classical layout (without BBR) andariations proposed by authors (REC2 and REC3).

ycle showed in Fig. 1 together with REC3 arrangement. Printedircuit Heat Exchangers (PCHE), manufactured by HEATRIC, areuggested in the literature for its use [8] in this kind of cycles dueo its compactness and its thermal-hydraulic performance [9,8]. Ass explained by Dostal [9], in the case of the shell and tube heatxchangers a high thickness of the tubes is required, thus reduc-ng its compactness. In the case of compact heat exchangers, therere different models, but two types were studied: plate and fin andCHE. In the former, they also required a material thickness tooigh, resulting in large dimensions. Thus the latter option (PCHE)as taken as the most appropriated for this kind of cycles.

Seven heat exchangers are considered, four of them belong tohe heat source (BNK, LDIV, LM and HDIV), two are recuperatorsBBR and LTR) and one is the precooler (PC). All of them are usualCHE except the element LM that has a special design still in devel-pment with the cross-sectional area of the channels increased andhe plates joined by means airfoil shaped fins [10].

Because of the behavior of CO2, a methodology based on theiscretization of the heat exchanger into sub-heat exchangers isequired. Also, three different types of approximations to calculatehe Nusselt number and the pressure losses are used. In the casef the most complex heat exchangers, the LTR and the precooler,FD simulations are required to verify the results. Another goal ofhe paper is to demonstrate how the size of the whole layout of theower cycle is optimized by the choice of PCHE.

. Methodology

Table 1 shows the total heat load, the temperature ranges andhe fluids of every each heat exchanger. The design process con-ists of dividing each heat exchanger into smaller elements, i.e.,ub-heat exchangers. Thus, the behavior of each fluid along thehannels can be obtained. The method starts introducing the inletemperature and pressure of the fluids in each sub-heat exchangernd follows an iterative process until the heat balance for everyub-heat exchanger is accurate. Also the pressure drop defined inhe development of the power cycle has to be satisfied, as is pre-ented by Dostal [9]. Three different empirical correlations to obtainhe Nusselt number and the pressure drop in each element arentroduced. These were extracted from the studies performed by

oisseytsev [11], Gnielinsky [9,12] and Dittus–Boelter [12]. Theesign process is programed and solved with Engineering Equa-

Please cite this article in press as: I.P. Serrano, et al., Modeling and spower cycle for energy conversion for fusion reactors, Fusion Eng. Des

ion Solver (EES) [13]. For the evaluation of the Nusselt numberNu), Dostal [9] recommended the use of the Gnielinski empiricalorrelation for Printed Circuit Heat Exchangers, based on [12]. Thus,

LTR CO2 130.4 59 1571CO2 51 118

for the different Reynolds number (Re) ranges, on the i-th elementcan be stated:

Nui = 4.089 (Rei < 2300) (1)

Nui = 4.089 + Nu5000 − 4.0895000 − 2300

(Rei − 2300)(2300 ≤ Rei < 5000)(2)

Nui = (fi/8)(Rei − 1000) · Pri

1 + 12.7 · (Pr2/3i

− 1) ·√

fi/8(Rei ≥ 5000) (3)

where f is the Darcy friction factor and Pr is de Prandtl number. Also,the Dittus–Boelter correlation is taken into account. This method isill-advised by Hesselgreaves [12], but there are other authors, likeMoisseytsev et al. [11] that use it in their studies. So:

Nui = 0.023 · Re4/5i

· Prni (Rei > 10, 000) (4)

where n is 0.4 for heating and 0.33 for cooling. For Rei ≤ 10, 000, theGnielinski empirical correlation should be used, Eqs. 1–3. Moisseyt-sev et al. [11] proposed another empirical correlation which is onlyused for CO2 (for helium, the Gnielinski or Dittus–Boelter correla-tions are used). So the Nusselt number is calculated as a functionof the Colburn factor (j), as follows:

Nui = ji · Rei · Pr1/3i

(5)

with

ji = 4.1Rei

(1 + Rei + 50

1000

)(Rei < 2300) (6)

ji = 0.6 · 0.1341 · Re−0.3319i (Rei ≥ 2300) (7)

For liquid metal the correlations proposed by Dostal et al. [9]have been assumed:

Nui = 5 + 0.025 · (Rei · Pri)0.8 (8)

Finally, evaluating the overall heat transfer coefficient (U) ofeach element, with a wetted perimeter (P), the length (L) of eachsub-heat exchanger is obtained by

Qi = Ui · Li · Pi · (Thot,i − Tcold,i) (9)

The pressure drop (�pi) is calculated as

�pi = fi

(Li

Di

)(c2

i

2

)(10)

izing of the heat exchangers of a new supercritical CO2 Brayton. (2014), http://dx.doi.org/10.1016/j.fusengdes.2014.04.039

friction factor, calculated as in Dostal et al. [9] and Moisseytsevet al. [11] correlations, and ci is the mean velocity of the fluid. Thetotal pressure drop of the heat exchanger must be 0.4 bar, which is

ARTICLE IN PRESSG ModelFUSION-7473; No. of Pages 4

I.P. Serrano et al. / Fusion Engineering and Design xxx (2014) xxx–xxx 3

tirush

cCodgi

oo[cHe1Ccalaudp

3

fweispthatfCt

Table 2CO2 heat transfer coefficients for HDIV, LDIV, LM and BBR.

Heat exchanger Dittus–Boelter(W/m2 K)

Gnielinski(W/m2 K)

Moisseytsev(W/m2 K)

HDIV 8567.4 7697.8 8142.2LDIV 5627.0 5036.4 5381.4LM 4046.7 3545.0 3450.8BBR hot stream 1082.0 1035.6 1307.2BBR cold stream 1020.8 1014.4 1320.0

transfer coefficient in CO2 side suffers a peak when CO2 reaches thecritical point. This occurs due to the considerable changes in thefluid properties. As can be seen the values obtained are very highwhich favor the heat transfer between both streams (Fig. 4).

Fig. 2. Heat transfer coefficient BNK.

he value used in the calculation of the power cycle [7] and usedn [2] and [14] for the design of classical S-CO2 cycles for fusioneactors. The number of channels within the PCHE must be modifiedntil the pressure drop reaches that value. Therefore, when both thepecified heat transfer and the total pressure drop are satisfied, theeat exchanger is dimensioned.

Due to the variation of CO2 properties, especially nearby its criti-al point (74 bar, 30 ◦C), the LTR and the precooler are modeled withFD implemented in ANSYS-Fluent v12.0. Properties as a functionf temperature have been approximated by different polynomialsepending on the range of temperatures. They have been pro-ramed and introduced, by means of user defined functions (udf),n ANSYS-Fluent.

The Printed Circuit Heat Exchanger (PCHE) has a similar numberf cold and hot channels. The finite element model only simulatesne hot and cold channel, and the rest is supposed to be similar9,15]. It is assumed that the channels are straight to minimize theomputational cost. The dimensions are those recommended byEATRIC in order to obtain the best performances (channel diam-ter of 2 mm, plate thickness of 1.5 mm and a channel pitch of.5 mm [15]). Because of computational cost, only the part whereO2 properties change abruptly has been simulated in the pre-ooler. Wall y+ is used as guidance for the grid configuration, using

wall layer with a total thickness of 209.6 �m. The SST k − ω turbu-ence model, recommended by the Nuclear Energy Agency (NEA),s a more advanced approach, for this kind of geometries [16] issed. Mass flow and temperature are introduced as boundary con-itions at the inlets and the external walls have been modeled aseriodic boundary conditions [17].

. Results

The sizing procedure with EES allows the comparison of the dif-erent empirical correlations. The heat transfer coefficient of theorking fluid (CO2) is analyzed along the length of every heat

xchanger and is also graphically represented when its variations appreciable. In the BNK the heat transfer coefficient in CO2 sideuffers a sharp drop at the cold inlet because of the change in CO2roperties (density and thermal conductivity) within that range ofemperatures. After that initial drop, they are stabilized and so theeat transfer coefficient (Fig. 2). This also means that the temper-ture drop between both streams decreases at higher rate during

Please cite this article in press as: I.P. Serrano, et al., Modeling and spower cycle for energy conversion for fusion reactors, Fusion Eng. Des

his initial section. However, for HDIV, LDIV and LM, the heat trans-er coefficient remains nearly constant because of the stability ofO2 properties. The values are shown in Table 2. In the LTR the heatransfer coefficient, on the hot side, increases in a similar way for

Fig. 3. Heat transfer coefficient LTR.

each correlation through the length of the heat exchanger as canbe seen in Fig. 3. The Moisseytsev correlation in cold side behavesdifferently, increasing the heat convection coefficient at an approx-imate constant rate, from 2122 W/m2 K at the inlet to 2459 W/m2 Kat the outlet (Fig. 3). The lower number of channels calculatedwith this correlation produces a higher Reynolds number differ-ence between the correlations. This influences the Nusselt numberand therefore the heat transfer coefficient (at the hot side the Nus-selt number obtained with each correlation is similar). There is alsoa visible change at position 0.25 m with the Dittus–Boelter corre-lation, because at this point the Reynolds number decreases from10,000 and then the Gnielinsky correlation is used.

In the BBR the heat transfer coefficient is also nearly constant.The values obtained are shown in Table 2. Finally, in the PC the heat

izing of the heat exchangers of a new supercritical CO2 Brayton. (2014), http://dx.doi.org/10.1016/j.fusengdes.2014.04.039

Fig. 4. Heat transfer coefficient PC (CO2 side).

ARTICLE ING ModelFUSION-7473; No. of Pages 4

4 I.P. Serrano et al. / Fusion Engineering

Table 3Heat exchangers sizes.

Helium/carbon dioxideBNK LDIV HDIV

Volume (m3) 8.8 1.6 2.3Heat load to volume (MW/m3) 90.1 205.6 107.4

Carbon dioxide/carbon dioxideBBR LTR

Volume (m3) 1272.8 340.6Heat load to volume (MW/m3) 4.7 4.6

Liquid/carbon dioxide

btcepliwbiHaag(sttsfsgo

4

slpbutifaa

sa

[

[

[

[

[

[

[

LM PCVolume (m3) 43.7 26.1Heat load to volume (MW/m3) 45.1 66.7

Sizes of each heat exchanger are given in Table 3. Volumes haveeen calculated using the Gnielinski correlation because of the bet-er results achieved in the LTR and the precooler, between eachorrelation and the CFD results. The LM is a novel proposed heatxchanger developed for fusion reactors [10] so it cannot be com-ared with the rest of heat exchanger. The ratio of the total heat

oad to the volume depends on the thermal properties of the flu-ds and the thermal approach. So, as recuperators (BBR and LTR)

ork with CO2 to CO2 they exhibits the lowest ratio; water hasetter thermal properties than helium, but thermal approach at PC

s higher than in BNK, so their ratios are close. Finally, LDIV andDIV work with helium and large thermal approach so their ratiosre the highest. In [2] a sizing study of a S-CO2 power cycle for

helium cooled (with low temperature divertor) fusion reactor isiven. The ratios of heat transfer rates to volumes in recuperators229 m3 for 1757 MW at LTR and 369 m3 for 3326 MW at HTR) areimilar with the ones resulting at present paper (taking into accounthey use the exact zig–zag channel distribution of PCHE). However,hey employ a shell and tube heat exchanger for the blanket heatource, requiring 4213 m3 for 3000 MW. This entails to 4948 m3

or all heat exchangers against 1696 m3 at present paper. So, theelection of PCHE for all the heat exchangers (except for the LM)enerates a large reduction in total volume, related with the costf the components [2].

. Conclusions

The different empirical correlations tested in this work giveimilar results, specially the Moisseytsev correlation and the corre-ation developed by Gnielinski. CFD simulations for the LTR and therecooler demonstrate that the three correlations give good results,eing the Gnielinski correlation the most approximated to the sim-lations. Hesselgreaves et al. [12] deduced that Dittus–Boelter is nothe best correlation for compact heat exchangers, as it is confirmedn these simulations. The Moisseytsev correlation was obtainedor a case where the CO2 is at different conditions of pressurend temperature. However, the results obtained are approximated,

Please cite this article in press as: I.P. Serrano, et al., Modeling and spower cycle for energy conversion for fusion reactors, Fusion Eng. Des

lthough they could not very considered appropriate.The ratio of total heat load to volume of the PCHE is not con-

tant but it depends on the fluids involved in the heat exchangernd the thermal approach. These ratios have been compared with

[

PRESS and Design xxx (2014) xxx–xxx

reported values from other authors, finding a good agreement. Theexception is the LM heat exchanger, which is proposed as a newheat exchanger and is still in development.

Acknowledgment

TECNO FUS is funded by the Ministry for Science and Innovationof the Spanish Government through CONSOLIDER-INGENIO 2010Programme (CSD2008 079).

Appendix A. Supplementary Data

Supplementary data associated with this article can be found,in the online version, at http://dx.doi.org/10.1016/j.fusengdes.2014.04.039.

References

[1] P. Hejzlar, V. Dostal, M.J. Driscoll, A supercritical CO2 cycle – a promising powerconversion system for generation IV reactors, in: ICAPP’06, Reno, USA, June 4–8,2006.

[2] S. Ishiyama, Y. Muto, Y. Kato, S. Nishio, T. Hayashi, Y. Nomoto, Study of steam,helium and supercritical CO2 turbine power generations in prototype fusionpower reactor, Progr. Nucl. Energy 50 (2008) 325–332.

[3] J. Sánchez, L.A. Sedano, E.R. Hodgson, M. Victoria, J. Sanz, M. Chuimenti, et al.,CONSOLIDER TECNO FUS: a nuclear fusion technology programme develop-ing dual functional helium/Pb15.7Li breeding blanket systems engineering andassociated underlying technologies, in: CO2 Symposium on Fusion Engineering,San Diego, USA, May 31–June 5, 2009.

[4] P. Norajitra, L. Bühler, U. Fischer, S. Gordeev, S. Malang, G. Reimann, Concep-tual design of the dual-coolant blanket in the frame of the EU power plantconceptual study, Fusion Eng. Des. 69 (2003) 669–673.

[5] G. Angelino, Perspectives for the liquid phase compression gas turbine, J. Eng.Power Trans. ASME 89 (1967) 229–237.

[6] I.P. Serrano, J.I. Linares, A. Cantizano, B.Y. Moratilla, A novel supercritical CO2

power cycle for energy conversion in fusion power plants, Fusion Sci. Technol.64 (2013) 483–487.

[7] I.P. Serrano, J.I. Linares, A. Cantizano, B.Y. Moratilla, 11th International Sympo-sium on Fusion Nuclear Technology, Fusion Eng. Des. (2013), http://dx.doi.org/10.1016/j.fusengdes.2014.03.083 (in press).

[8] K. Nikitin, Y. Kato, L. Ngo, Printed circuit heat exchanger thermal-hydraulicperformance in supercritical CO2 experimental loop, Int. J. Refrigeration 29(2006) 807–814.

[9] V. Dostal, A supercritical carbon dioxide cycle for next generation nuclear reac-tors, Massachusetts Institute of Technology, USA, 2004, Ph.D. Thesis.

10] I. Fernández, L. Sedano, Design analysis of a lead–lithium/supercritical CO2

printed circuit heat exchanger for primary power recovery, Fusion Eng. Des.88 (2013) 2427–2430.

11] A. Moisseytsev, J.J. Sienicki, D.H. Cho, M.R. Thomas, Comparison of heatexchanger modeling with data from CO2-to-CO2 printed circuit heat exchangerperformance tests, in: ICAPP’10, San Diego, USA, June 13–17, 2010.

12] J.E. Hesselgreaves, Compact Heat Exchangers: Selection, Design and Operation,Pergamon Press, Oxford, 2001.

13] S.A. Klein, Development and integration of an equation-solving program forengineering thermodynamics courses, Comput. Appl. Eng. 1 (1993) 265–275.

14] M. Medrano, D. Puente, E. Arenaza, B. Herrazti, A. Paule, B. Branas, et al., Powerconversión cycles study for He-cooled reactor concepts for DEMO, Fusion Eng.Des. 82 (2007) 2689–2695.

15] I.P. Serrano, A. Cantizano, J.I. Linares, B.Y. Moratilla, I. Fernndez-Berceruelo,L. Sedano, Numerical modeling and design of supercritical CO2 pre-cooler forfusion nuclear reactors, Fusion Eng. Des. 87 (2012) 1329–1332.

16] J. Mahaffy, B. Chung, F. Dubois, F. Ducros, E. Graffard, M. Heitsch, Best practice

izing of the heat exchangers of a new supercritical CO2 Brayton. (2014), http://dx.doi.org/10.1016/j.fusengdes.2014.04.039

guidelines for the use of CFD in nuclear reactor safety applications, in:NEA/CSNI, 2007.

17] I.H. Kim, H.C. No, J.I. Lee, B.G. Jeon, Thermal hydraulic performance analysis ofthe printed circuit heat exchanger using a helium test facility and CFD simula-tions, Nucl. Eng. Des. 239 (2009) 2399–2408.