Format And Type Fonts


 

 CCHHEEMMIICCAALL  EENNGGIINNEEEERRIINNGG  TTRRAANNSSAACCTTIIOONNSS  
 

VOL. 45, 2015 

A publication of 

 
The Italian Association 

of Chemical Engineering 

www.aidic.it/cet 
Guest Editors: Petar Sabev Varbanov, Jiří Jaromír Klemeš, Sharifah Rafidah Wan Alwi, Jun Yow Yong, Xia Liu  

Copyright © 2015, AIDIC Servizi S.r.l., 

ISBN 978-88-95608-36-5; ISSN 2283-9216 DOI: 10.3303/CET1545288 

 

Please cite this article as: Čuček L., Kravanja Z., 2015, Retrofitting of large-scale heat exchanger networks within total sites 
under uncertainty by considering trade-offs between investment and operating cost, Chemical Engineering Transactions, 45, 
1723-1728  DOI:10.3303/CET1545288 

1723 

Retrofitting of Large-Scale Heat Exchanger Networks within 

Total Sites under Uncertainty by Considering Trade-Offs 

between Investment and Operating Cost 

Lidija Čuček, Zdravko Kravanja* 

Faculty of Chemistry and Chemical Engineering, University of Maribor, Smetanova ulica 17, 2000 Maribor 

zdravko.kravanja@um.si 

This contribution presents a developed procedure for retrofitting of large-scale heat exchanger networks 

within industrial plants and Total Sites by considering trade-offs between investment and operating costs. 

Investment cost includes costs for additional heat exchange areas and pipes at Total Site level, and 

operating cost (savings) includes energy savings due to reduced energy consumption and intermediate 

utility production, e.g. production of hot water for district heating. A transhipment-based framework 

TransGen for the energy targeting and retrofitting industrial plants and TSs has been developed for this 

purpose. Retrofitting of heat exchanger networks could be performed for problems of any size – from few 

heat exchange units up to hundreds of them. This procedure could be applied for both nominal designs 

under fixed conditions and under uncertain conditions for flexible designs by considering several 

scenarios. TransGen is illustrated on a simplified demonstration case study of two plants in order to clearly 

represent the features of the developed procedure. 

1. Introduction  

Retrofitting of heat exchanger networks (HENs) within existing industrial plants and Total Sites plays an 

important role regarding energy savings, and emissions and cost reductions. Several mixed-integer 

nonlinear programming (MINLP) models for the synthesis and retrofitting of HENs have been developed so 

far, as well as for entire Total Sites. However, those models are non-linear programming-hard and cannot 

be used for retrofitting of large-scale industrial processes and Total Sites, especially under varying 

conditions, due to the complex combinatorics and non-convexities involved. Therefore, a combined 

Mathematical Programming / Pinch Analysis (MP/PA) approach for targeting and mixed-integer linear 

programming (MILP) for retrofitting HENs have been developed and upgraded for fixed and uncertain 

operating conditions. 

This contribution presents an upgraded version of the recently-developed code TransGen (Čuček and 

Kravanja, 2014) that proposes optimal modifications within industrial plants and Total Sites and provides 

options for producing intermediate utilities such as hot water for district heating from process low-grade 

heat. The TransGen framework is suitable for analysing energy targets and existing HEN designs, and 

especially for proposing optimal modifications for the retrofitting of each plant and Total Sites under 

nominal and uncertain conditions. It has been further upgraded to propose certain number of modifications 

(new heat exchange matches) that are (or could be) feasible in terms of economical, technological, safety 

and other limitations. An additional constraint within the code is added in order to obtain a certain pre-

specified number of proposed modifications (new heat exchange matches). With regards to the set 

objective (e.g. minimising cost, maximising net present value etc.) the most optimal retrofitting 

modifications are then found which now consider also investigations regarding the trade-offs between the 

investment and operating costs by accounting for HEN design at plant and Total Site levels. In addition 

pipeline layout, heat and pressure losses and insulation are included within the TransGen (Nemet et al., 

2015).  



 
1724 

 
The main features of TransGen are thus possibility of obtaining certain pre-defined number of the most 

optimal retrofitting modifications in regards to the trade-offs between investment and operating cost. The 

new features of the framework are illustrated here within a demonstration case study of two units and 

applied also on an existing refinery Total Site (see the contribution by Čuček et al., 2015). From the 

demonstrated case study it can be seen that significant energy and cost savings could be obtained when 

applying the developed procedure. 

2. Methodology 

2.1 Transshipment based framework TransGen 
Framework TransGen (Čuček and Kravanja, 2014) is an extended version of an expanded transshipment 

model by Papoulias and Grossmann (1983). It is based on a combined MP/PA approach and overcomes 

the drawbacks of both approaches. The framework is coded in GAMS (GAMS Development Corporation, 

2013) within mixed-integer linear programming (MILP) form, and provides numerical and graphical outputs 

of the solutions, e.g. by producing Grand Composite Curves and Total Site Profiles. 

The more significant features of the framework are (Čuček and Kravanja, 2014): i) enabling the obtaining 

of existing, target and modified (proposed) HEN designs, ii) handling multiple utilities and production of 

intermediate utilities by including preheating when appropriate, iii) insights from Pinch Technology, iv) data 

independence, v) MILP form and therefore the globally-optimal solutions, vi) internal and Total Site Heat 

Integration under fixed and varying operational conditions, vii) possibility of analysing HENs of all scales, 

from few up to hundreds of heat exchange units. 

However, there were several weaknesses of the previous version (Čuček and Kravanja, 2014), such as 

investment cost, pressure drops and heat losses at the Total Site level, due to retrofit were unconsidered. 

Also, search for modifications was performed iteratively based on utility consumption reduction until there 

was no significant improvement in the subsequent modification. Those weaknesses have now been 

overcome by proposing a three-step procedure, where the potential for improvement is identified in the first 

step, the most optimal retrofitting modifications are obtained in the second step, and more detailed HEN 

design in the third step. 

2.2 Improvements of TransGen framework 
The TransGen framework has been significantly modified and improved in order for it to be possible to 

propose the most optimal HEN’s retrofitting modifications. The most important new feature is that the 

solutions consider trade-offs between investment cost (cost in new heat exchange units’ areas and the 

cost of piping) and savings in energy cost. There are several possibilities regarding the proposed 

retrofitting modifications solutions: i) to propose the solutions for specific unit and/or the entire Total Site, ii) 

to propose all the solutions which have positive impact on the profit, iii) to propose a limited number of the 

retrofitting solutions, e.g. to propose modifications of a certain number of new/modified heat exchange 

matches, and iv) to propose solutions which fall in the limit of the specified payback period or for available 

money for retrofit investment. 

Using an improved TransGen framework it is thus possible to identify under both fixed and uncertain 

conditions i) optimal number of modifications, locations, heat contents and areas of rearranged heat 

exchange units, and ii) matches that are rearranged only partially or completely, and also matches that are 

newly-formed. It is also possible to fix or limit intermediate utility consumption and production such as e.g. 

the production of a certain amount of hot water for district heating, and steam at various pressure levels 

which is used within the plant or Total Site. It also enables eventual utilisation of existing heat exchangers 

in those cases where the entire load is released. The obtained solutions under nominal and flexible 

conditions could be obtained in ‘real’ time even if the number of heat exchange units is large (e.g. they are 

hundreds of them). Several features of the improved TransGen framework will be illustrated within a 

demonstration case study of two process plants (Perry et al., 2008). 

2.3 Three-step procedure for the retrofitting of large-scale HENs 
It is proposed that a three-step procedure is performed for the retrofitting of large-scale HENs. During the 

first and second steps the TransGen framework based on a developed MILP transshipment model is used. 

TransGen could be used for i) obtaining the potential for the improvement within the process plant and/or 

Total Site (target design) in the first step and ii) identification of alternatives, and also to narrow the original 

search space in the second step. The second step also guides optimisation towards obtaining (near) 

global as well as feasible solutions regarding retrofitting. Usually several loops are required in order to 

obtain (near) optimal, verified and feasible results. During the third step detailed investigation of the trade-

offs between investment and operating cost is performed on the reduced space of alternatives, as 

identified during the second step. 



 
1725 

There have been several models developed for this purpose such as e.g., the model by Yee and 

Grossmann (1991) for counter-flow exchangers or the model by Soršak and Kravanja (2004) accounting 

for different heat exchanger types. Third step also enables obtaining positions and temperatures of heat 

exchange units within modified HEN. It should be noted that the main disadvantage of the third step in 

general is that it is impossible to solve large-scale problems, and with this proposed three-step procedure 

this could be overcome. This three-step procedure can namely be applied for any size of plant or Total Site 

and is applicable for those sectors that operate under steady-state and varying conditions. This three-step 

procedure is introduced and presented in more details within contribution by Čuček and Kravanja (2015).  

It should be noted that in this contribution the improved version of TransGen - in terms of obtaining the 

most optimal retrofitting modifications from operating and investment viewpoint - is introduced and 

illustrated on an illustration case study. In the contribution by Čuček and Kravanja (2015) the novel three-

step methodology is introduced which combined this improved TransGen framework and the framework for 

the detailed analysis of modified HENs. 

3. Illustration case study 

A simple illustration case study of two units A and B (Perry et al., 2008) under uncertain conditions will 

show the applicability of the improved TransGen framework. 

3.1 Description of the illustration case study 
The data for process units were taken from Perry et al. (2008) and had been slightly modified to account 

for fluctuations of temperature and flows. Operating within three representative conditions was assumed 

(three cases). The HENs for the existing designs of units A and B were assumed as shown in Figures 1 – 

3 for cases 1 – 3. Also, the temperatures and heat duties of the heat exchangers, heaters and coolers are 

shown. 

Several assumptions were made for retrofitting such as units A and B being located 1.5 km apart, 2 MW of 

hot water for district heating would be produced in all cases, the price of hot water would be 25 €/MWh, all 

streams except the low and medium pressure steam (LPS and MPS) would be liquids, the amortisation 

period would be of 5 y and the discount rate 8 %, heat loss would be 10 % when applying Total Site Heat 

Integration, and the minimum heat transfer for reallocations for each case would be at least 0.5 MW. The 

data for investment cost of heat exchange area and pipes were taken from Nemet et al. (2015). For 

simplicity it was also assumed that the pressures of the process streams would be at 1 bar. 

The MILP model consists of approximately 7,200 single equations, 37,500 single variables, and 300 binary 

variables, and was solved within a few seconds with GAMS 23.6 (GAMS Development Corporation, 2013) 

using a CPLEX solver. 

170  C
H1

C1 C3

143  C 80  C

25  C

100  C

30  C

53.3  C

65  C

1.5 MW 3.5 MW

1.75 MW

150  C
H2

C2

134.6  C 55  C

70  C

100  C

1.05 MW 5.43 MW

170  C

170  C

15  C

19  C

200  C
H3

C7 C4

116  C 80  C

75  C

150  C

20  C

80  C

7 MW 3 MW

150  C
H4

C5

120.4  C 40  C

100  C

104.3  C

2.15 MW 4.35 MW

15  C

19  C

120  C

7.85 MW

170  C

170  C

106.65  C

100  C

1 MW

85  C

0.5 MW

110  C

0.5 MW

170  C

170  C

99.78  C

60  C

C6

1 2

3

4 5

6 7

8

9

10 11

12

13

 

Figure 1: HEN for units A and B for case 1 



 
1726 

 

167  C
H1

C1 C3

144.48  C 84.98  C

25  C

109  C

41  C

65  C

1.36 MW 3.58 MW 0.66 MW

154  C
H2

C2

139  C 62  C

67  C

98  C

0.92 MW 4.72 MW

19  C

15  C

15  C

19  C

196  C
H3

C7 C4

112.52  C 74  C

79  C

154  C

20  C

80.1  C

6.54 MW 3 MW

165  C
H4

C5

121  C 39  C

101  C

107.5  C

3.2 MW 4.2 MW

15  C

19  C

126  C

9.1 MW

170  C

170  C

106.45  C

101  C

1 MW

86  C

0.66 MW

103  C

0.45 MW

170  C

170  C

97.32  C

61  C

C6
1 2

3

4 5

6 7

8

9

10 11

12

13

74  C

 

Figure 2: HEN for units A and B for case 2 

172  C
H1

C1 C3

152.11  C 90.59  C

18  C

89  C

36  C

61  C

1.39 MW 4.3 MW 0.67 MW

145  C
H2

C2

134.2  C 57  C

74  C

104  C

0.83 MW 5.95 MW

19  C

15  C

15  C

19  C

202  C
H3

C7 C4

98.36  C 75  C

60  C

147  C

21  C

58.3  C

7.75 MW 1.75 MW

147  C
H4

C5

109  C 37  C

89  C

93  C

2.58 MW 3.22 MW

15  C

19  C

121  C

14.47 MW

170  C

170  C

89.19  C

87  C

1.35 MW

69  C

0.33 MW

95  C

0.52 MW

170  C

170  C

84.4  C

53  C

C6
1 2

3

4 5

6 7

8

9

10 11

12

13

81  C

 

Figure 3: HEN for units A and B for case 3 

3.2 Results and discussion 
The demonstration case study was solved using the improved TransGen framework. Analysis was 

performed in order to obtain existing, targeted and modified designs. Analysis regarding modifications was 

performed for a certain number of new heat exchange matches (five heat exchange matches), and also for 

optimum design, when all the profitable solutions were obtained in regard to the trade-off between 

investment and operating costs. The main results regarding existing and target designs are shown in Table 

1, and the main results regarding modified designs with 5 modifications and all profitable modifications are 

shown in Table 2. Both modified designs are compared to an existing design. 

When comparing the existing and target designs it can be seen that there is a large potential for 

improvement. Energy cost and energy consumption could be significantly reduced. 

 



 
1727 

Table 1: Main results regarding existing and target designs 

  Existing design Target design 

Average energy cost 3.4 M€/y 0.9 M€/y 

Average energy consumption  21.4 MW 4.8 MW 

Hot utility consumption   

Case 1 10.1 MW 1.7 MW 

Case 2 9.6 MW 3.7 MW 

Case 3 15.4 MW 6.7 MW 

Cold utility consumption   

Case 1 9.8 MW 0.1 MW 

Case 2 9.6 MW 2 MW 

Case 3 9.8 MW 0.2 MW 

 

Table 2:  Main results regarding modified design with 5 modifications, and optimal design 

  Modified design with 5 

modifications  

Modified optimal design with all 

profitable modifications 

Number of modifications 5 13 

Average profit 1.23 M€/y 1.55 M€/y 

Average energy savings  7.96 M€/y 12.08 MW 

Retrofit area investment 0.78 M€ 1.72 M€ 

Pipe investment 0.84 M€ 1.09 M€ 

Payback time 0.99 y 1.25 y 

Modified heat exchange units 

(fraction of rearranged heat 

exchange unit’s energy) 

2 (29 – 40 %), 5 (87 – 96 %),  

6 (30 – 43 %), 9 (32 – 43 %),  

13 (27 – 60 %) 

1 (98 – 100 %), 2 (75 – 97 %),  

4 (100 %), 5 (100 %), 6 (62 – 75 %),  

7 (51 – 74 %), 8 (31 – 46 %) 

9 (40 – 79 %), 13 (72 – 79 %)  

New heat exchange matches at 

process level 

H5 – hot water (2 MW) H2 – C1 (1.33 – 1.5 MW),  

H5 – C2 (2.7 – 3.92 MW),  

H5 – C4 (0.83 – 1.05 MW 

New heat exchange matches at 

Total Site level 

H2 – C9 (1 – 1.7 MW),  

H5 – C6 (2.2 – 3.7 MW),  

H13 – C2 (1.1 – 1.9 MW) 

H1 – C6 (1.33 – 1.5 MW),  

H2 – C6 (1.37 – 2.53 MW),  

H4 – C6 (0.83 – 1.05 MW), 

H5 – C8 (1 – 1.19 MW), 

H7 – C6 (0.55 – 1.28 MW), 

H13 – C7 (1.28 – 1.78 MW), 

H6 – C9 (4.26 – 5.27 MW),  

H8 – C9 (0.9 – 1.07 MW), 

H7 – hot water (0.5 – 0.74 MW),  

H13 – hot water (1.26 – 1.5 MW)  

Hot utility consumption (absolute 

and relative reduction) 
  

Case 1 6.7 MW (3.4 MW, 33 %) 3.9 MW (6.2 MW, 61 %) 

Case 2 6.7 MW (2.9 MW, 30 %) 4.3 MW (5.2 MW, 55 %) 

Case 3 10.6 MW (4.9 MW, 31.6 %) 9.5 MW (5.9 MW, 38 %) 

Cold utility consumption (absolute 

and relative reduction) 

  

Case 1 3.9 MW (5.9 MW, 60 %) 1.1 MW (8.7 MW, 89 %) 

Case 2 4.3 MW (5.3 MW, 55 %) 1.9 MW (7.7 MW, 81 %) 

Case 3 2.2 MW (7.6 MW, 77 %) 1.3 MW (8.5 MW, 86 %) 

 

By performing retrofitting modifications quite significant profit due to Total Site Heat Integration is obtained, 

and also energy consumption is substantially reduced. Most modifications are proposed for 

implementations at the Total Site level, and due to hot water production. It can be seen that all the results 

were obtained within ranges, as the conditions vary. It can also be seen that the average profit did not 

improved much when performing a large number of modifications. Usually with just a few modifications 

significant profit could be obtained. Also the payback period was shorter when performing lower number of 



 
1728 

 
modifications. Finally, from Table 2 it can be seen that in most cases the hot or cold streams’ energies 

were partially reallocated and only a few of the existing matches were completely eliminated (heat 

exchange units 4 and 5, and heat exchange unit 1 in some cases). 

4. Conclusions and future work 

This contribution described the upgraded version of MP/PA framework TransGen which has been included 

in the three-step procedure (Čuček and Kravanja, 2015). The TransGen can be applied for retrofitting of 

HENs of any scale at process and Total Site levels under uncertainty by considering trade-offs between 

investment and operating costs. This contribution was focused on proposing a certain pre-specified 

number of retrofitting modifications (new heat exchange matches) in regards to the trade-off between 

investment and operating cost. A small demonstration case study covering three representative cases was 

solved in order to show the features of the improved framework. Software tool TransGen now enables to 

obtain a certain number of the most optimal new heat exchange matches as shown in Table 2. It should be 

noted that the previous version of TransGen presented in Čuček and Kravanja (2014) enabled to obtain 

the results which consider a certain number of modifications from existing heat exchange matches and 

only in terms of operating cost. Also, the results presented in Čuček and Kravanja (2014) considered 

deterministic designs, whilst now the results are presented for both deterministic and flexible designs. 

From the results from improved version of TransGen it was shown that significant energy and economic 

savings could be obtained.  

The improved TransGen is also applied to a large-scale HEN within an existing refinery (Čuček et al., 

2015). The third and more detailed step employs the model by Soršak et al. (2004) considering different 

types of heat exchangers, and was upgraded in terms of pipeline cost (Nemet et al., 2015) and 

investigation of the proposed HEN modifications under varying conditions. 

Acknowledgments 

The authors acknowledge the financial support from the EC FP7 project ENER/FP7/296003/EFENIS 

‘Efficient Energy Integrated Solutions for Manufacturing Industries – EFENIS’, and from the Slovenian 

Research Agency (programs P2-0032 and P2-0377). The authors are also grateful to EFENIS partners, 

especially to Dr Valter Mantelli, Prof Jiří J. Klemeš, Prof Petar S. Varbanov, DDr Andreja Nemet and Mr 

Jun Yow Yong. 

References 

Čuček L., Kravanja Z., 2014, Efficient Transshipment-Based Framework for Energy Targeting and 

Retrofitting Industrial Total Sites, Chemical Engineering Transactions, 39, 1813-1818, doi: 

10.3303/CET1439303. 

Čuček L., Kravanja Z., 2015, A Procedure for the Retrofitting of Large-Scale Heat Exchanger Networks for 

Fixed and Flexible Designs, Chemical Engineering Transactions, 45, 31–36, DOI: 

10.3303/CET1545006. 

Čuček L., Mantelli V., Yong J.Y., Varbanov P.S., Klemeš J.J., Kravanja Z., A Procedure for the Retrofitting 

of Large-Scale Heat Exchanger Networks for Fixed and Flexible Designs applied to Existing Refinery 

Total Site, Chemical Engineering Transactions, 45, 109–114, DOI: 10.3303/CET1545019. 

GAMS Development Corporation, 2013, GAMS — A User’s Guide, Washington, DC, USA. 

Nemet A., Klemeš J.J., Kravanja Z., 2015, Designing a Total Site for an entire lifetime under fluctuating 

utility prices, Computers & Chemical Engineering, 72, 159-182. 

Papoulias S. A., Grossmann I. E., 1983, A structural optimization approach in process synthesis – II: Heat 

recovery networks, Computers & Chemical Engineering, 7(6), 707-721. 

Perry S., Klemeš J.J., Bulatov I., 2008, Integrating waste and renewable energy to reduce the carbon 

footprint of locally integrated energy sectors, Energy, 33(10), 1489-1497. 

Soršak A., Kravanja Z., 2004, MINLP retrofit of heat exchanger networks comprising different exchanger 

types, Computers & Chemical Engineering, 28(1–2), 235-251. 

Yee T. F., Grossmann I. E., 1991, A screening and optimization approach for the retrofit of heat-exchanger 

networks, Industrial & Engineering Chemistry Research, 30(1), 146-162.