development of novel refrigeration cycles for small scale
TRANSCRIPT
Development of novel refrigeration cycles for small scale
LNG processes
A thesis submitted to The University of Manchester for the degree of
Master of Philosophy
in the Faculty of Engineering and Physical Sciences
2016
Héctor Fernando Almeida Trasviña
School of Chemical Engineering and Analytical Science
2
List of Contents
List of Figures ....................................................................................................................... 5
List of Tables ........................................................................................................................ 9
Abstract ............................................................................................................................... 11
Declaration .......................................................................................................................... 12
Copyright Statement .......................................................................................................... 13
Acknowledgements ............................................................................................................. 14
Chapter 1 – Introduction ................................................................................................... 15
1.1 LNG production at small scale .............................................................................. 16
1.2 Challenges in design and optimisation of refrigeration cycles for LNG at
small scale ............................................................................................................. 17
1.3 Motivation and research objectives ....................................................................... 21
1.4 Overview of this Thesis ......................................................................................... 22
Chapter 2 – Technology Background and Literature Review ....................................... 24
2.1 Introduction ........................................................................................................... 24
2.2 Modelling of refrigeration cycles for LNG processes ........................................... 26
2.2.1 Modelling of pure refrigerant cycles .............................................................. 29
2.2.2 Modelling of mixed refrigerant cycles ........................................................... 30
2.2.3 Modelling of cascade cycles .......................................................................... 33
2.3 Research literature on refrigeration cycles for LNG processes ............................. 35
2.3.1 Modelling and design of mixed refrigerant cycles ......................................... 36
2.3.2 Optimisation of mixed refrigerant cycles ....................................................... 42
2.3.3 Design of refrigeration cycles for small scale LNG processes ...................... 47
2.3.4 Design and optimisation of the CryoMan process ......................................... 51
2.3.5 WORK software ............................................................................................. 55
2.3.6 Example of refrigeration cycle simulation in WORK software ..................... 56
2.4 Conclusions ........................................................................................................... 59
Chapter 3 – Development and Design of Novel Refrigeration Cycles ........................... 61
3.1 Introduction ........................................................................................................... 61
3.1.1 Benchmark processes ..................................................................................... 63
3.1.2 Performance evaluation .................................................................................. 64
3.1.3 Operating cost estimation ............................................................................... 64
3.2 Further development of the CryoMan process ...................................................... 64
3
3.2.1 Analysis of the CryoMan process .................................................................. 65
3.2.2 Generation of structural options ..................................................................... 69
3.2.3 Proposal of novel designs .............................................................................. 71
3.3 Process modelling .................................................................................................. 73
3.3.1 Modelling of the Bypass design ..................................................................... 73
3.3.2 Modelling of the Two Flash Levels design .................................................... 80
3.3.3 Modelling of the Mixing After Precooling design ......................................... 83
3.3.4 Example of mixed refrigerant cycle modelling .............................................. 86
3.4 Strategy for the evaluation of the novel designs ................................................... 89
3.4.1 Simulation and analysis ................................................................................. 90
3.4.2 Optimisation of promising designs ................................................................ 90
3.5 Conclusions ........................................................................................................... 91
Chapter 4 – Evaluation of the Novel Refrigeration Cycles ............................................ 93
4.1 Introduction ........................................................................................................... 93
4.2 Determination of the natural gas conditions .......................................................... 93
4.3 Sensitivity analyses: Manipulation of the degrees of freedom in the novel
refrigeration cycles ................................................................................................ 99
4.3.1 Manipulation of the refrigerant composition ............................................... 100
4.4 Evaluation of the Bypass design ......................................................................... 102
4.4.1 Bypass design: Initial simulation ................................................................. 102
4.4.2 Bypass design: Sensitivity studies and discussion ....................................... 103
4.4.3 Bypass design: Manipulation of its degrees of freedom .............................. 110
4.5 Evaluation of the Two Flash Levels design ........................................................ 112
4.5.1 Two Flash Levels design: Initial simulation ................................................ 112
4.5.2 Two Flash Levels design: Sensitivity studies and discussion ...................... 114
4.6 Evaluation of the Mixing After Precooling design.............................................. 120
4.6.1 Mixing After Precooling design: Initial simulation ..................................... 120
4.6.2 Mixing After Precooling design: Sensitivity studies and discussion ........... 121
4.7 Selection of the novel designs for optimisation .................................................. 126
4.8 Conclusions ......................................................................................................... 127
Chapter 5 – Case Study: Optimisation of Novel Refrigeration Cycles ....................... 128
5.1 Introduction ......................................................................................................... 128
5.2 Optimisation criteria ............................................................................................ 129
4
5.3 Problem statement ............................................................................................... 131
5.4 Bypass design: Optimisation and discussion ....................................................... 132
5.5 Two Flash Levels design: Optimisation and discussion ..................................... 136
5.6 Operating costs comparison between the novel refrigeration cycles and
the benchmark processes ..................................................................................... 140
5.7 Conclusions ......................................................................................................... 142
Chapter 6 – Conclusions and Future Work ................................................................... 144
6.1 Conclusions ......................................................................................................... 144
6.2 Future work ......................................................................................................... 147
Appendix 1 – Determination of the Natural Gas Stream Conditions ......................... 149
Appendix 2 – Bypass design and Two Flash Levels design: Optimisation Results .... 159
A2.1 Optimisation results of the Bypass design ....................................................... 159
A2.2 Optimisation results of the Two Flash Levels design ...................................... 162
Appendix 3 – Optimisation of the CryoMan Process with Six Compression Stages . 166
A3.1 Introduction ...................................................................................................... 166
A3.2 Problem formulation ........................................................................................ 167
A3.3 Problem statement............................................................................................ 168
A3.4 Optimisation of the CryoMan process ............................................................. 169
References ......................................................................................................................... 172
5
List of Figures
Figure 1. 1. World energy demand according to the New Policies Scenario:
a) 2015 and b) 2035 [adapted from IEA (2012)]. ................................................ 16
Figure 1. 2. The PRICO cycle (Swenson, 1977). ..................................................................... 18
Figure 1. 3. Complexity of commercial refrigeration cycles increases as efficiency
increases [adapted from (Air Products and Chemicals Inc., 2013)]. .................... 19
Figure 1. 4. The CryoMan configuration (Zheng, 2009). ........................................................ 21
Figure 2. 1. Ideal refrigeration cycle: a) configuration; b) P–H diagram
[adapted from (Smith, 2005b, Ch. 24.6)]. ............................................................. 24
Figure 2. 2. Configuration options to reduce shaft power demand of a simple cycle:
a) multistage expansion; b) intercooling: c) multilevel refrigeration
[adapted from (Smith, 2005b, Ch. 24.6)]. ............................................................ 26
Figure 2. 3. Refrigerant evaporating profile (at constant pressure): a) pure component;
b) mixed refrigerant [adapted from (Radermacher, 1989)]. ................................. 30
Figure 2. 4. Temperature–enthalpy profile calculation for a mixed refrigerant stream
(ΔP = 0 in the evaporator). ................................................................................... 31
Figure 2. 5. PRICO refrigeration cycle (Swenson, 1977). ....................................................... 33
Figure 2. 6. Cascade refrigeration cycle with two temperature levels of refrigeration. ........... 34
Figure 2. 7. Commercial cascade cycles: a) Phillips cascade cycle; b) Propane precooled
mixed refrigerant cycle; c) Dual mixed refrigerant cycle. .................................... 35
Figure 2. 8. Two temperature levels refrigeration cycle using a presaturator
(Shelton and Grossmann, 1986). ........................................................................... 36
Figure 2. 9. Design methodology for low temperature processes presented by Linnhoff
and Dhole (1992)................................................................................................... 38
Figure 2. 10. Generation of the ‘ideal’ cold composite curve (Lee, 2001, Ch. 4). .................. 38
Figure 2. 11. Multistage refrigeration cycle. ............................................................................ 41
Figure 2. 12. a) The PRICO cycle; b) “Pre-flash” design; c) the CryoMan process
[adapted from (Zheng, 2009)]. ............................................................................ 52
Figure 2. 13. Operating conditions of the PRICO cycle for the simulation example
problem. ............................................................................................................. 57
Figure 2. 14. Natural gas temperature–enthalpy data for the simulation example in
WORK software. ................................................................................................. 58
Figure 3. 1. Multistage centrifugal compressor [adapted from (Ludwig, 2001, Ch. 12)]. ....... 62
Figure 3. 2. Multi-stream heat exchanger for five streams [reproduced from (ESDU
International plc, 2006)]. ....................................................................................... 63
Figure 3. 3. Refrigeration cycles presented by Zheng (2009): a) “Pre-flash” design, and
b) the CryoMan process. ...................................................................................... 65
Figure 3. 4. Ternary refrigerant after flash separation: a) composition distribution; and
b) shaft work for compression of 1 kmol·s-1
from 1.2 bar and 30°C to 20 bar. ... 66
Figure 3. 5. Shaft work demand of LP Stream in the “Pre-flash” and in the CryoMan
process for different outlet pressures at constant inlet pressure (1.2 bar). ........... 68
6
Figure 3. 6. Compression trends with actual flow rates (Zheng, 2009): a) LP Streams
at constant inlet pressure (1.2 bar); and b) overall refrigerant streams at
constant outlet pressure (48.3 bar). ...................................................................... 69
Figure 3. 7. A refrigerant stream bypassing the flash unit. ...................................................... 70
Figure 3. 8. Multiple flash separation of the refrigerant stream (liquid from first flash
unit is further expanded). ..................................................................................... 70
Figure 3. 9. Partial mixing of refrigerant stream after precooling in the MSHE. .................... 71
Figure 3. 10. Novel refrigeration cycle 1: Bypass design. ....................................................... 71
Figure 3. 11. Novel refrigeration cycle 2: Two Flash Levels design. ...................................... 72
Figure 3. 12. Novel refrigeration cycle 3: Mixing After Precooling design. ........................... 73
Figure 3. 13. Bypass design: degrees of freedom. ................................................................... 74
Figure 3. 14. Multistage compression. ..................................................................................... 78
Figure 3. 15. Two Flash Levels design: degrees of freedom. .................................................. 81
Figure 3. 16. Mixing After Precooling design: degrees of freedom. ....................................... 84
Figure 3. 17. Natural gas temperature–enthalpy data (from HYSYS, 1 kmol·s-1
) for the
modelling example. ............................................................................................. 87
Figure 4. 1. Temperature–enthalpy profile according to data from Remeljej and Hoadley
(2006) compared to that published by Lee (2001). ............................................... 95
Figure 4. 2. Determination of the natural gas stream conditions through minimisation of
the sum of squared difference of the enthalpy profiles against the data
provided by Lee (2001, Ch. 4). ............................................................................ 96
Figure 4. 3. Temperature–enthalpy profile of the optimised stream compared to the data
provided by Lee (2001). ....................................................................................... 97
Figure 4. 4. The CryoMan process. .......................................................................................... 99
Figure 4. 5. Example of composition manipulation: propane mole percentage is
increased; the mole percentage proportion between the remaining
components remains the same. ........................................................................... 102
Figure 4. 6. The Bypass design showing the new degrees of freedom and initial values. ..... 102
Figure 4. 7. Effect of increasing the Bypass Stream flow rate fraction: a) shaft power
demand and minimum driving force in the MSHE; b) the composite
curves in the MSHE. .......................................................................................... 103
Figure 4. 8. Effect of increasing the pressure level of the Bypass Stream: a) shaft power
demand and minimum driving force in the MSHE; b) P–H diagram. ............... 104
Figure 4. 9. Effect of increasing the compressor discharge pressure: a) power demand
and minimum driving force; b) the vapour fraction of the refrigerant. .............. 105
Figure 4. 10. Effect of increasing the flow rate fraction of liquid from the flash unit (f Liq
):
a) shaft power demand and minimum driving force in the MSHE; b) flow
rates of Stream 1 and Stream 2. ....................................................................... 106
Figure 4. 11. a) Increasing the value of f Liq
increases the heat of vaporisation of Stream 1;
b) heat of vaporisation of the refrigerant is increased as heavy components
in the composition are increased. ..................................................................... 107
Figure 4. 12. Effect of composition on power demand and minimum driving force in
the Bypass design: a) methane; b) ethane; c) propane; d) n-butane;
e) nitrogen ........................................................................................................ 108
7
Figure 4. 13. As the overall refrigerant composition becomes lighter: a) its vapour
fraction increases; and b) its evaporating temperatures decreases. .................. 110
Figure 4. 14. Manipulation of the operating variables in the Bypass design: a) Bypass
Stream flow rate fraction is varied; b) the liquid mixing fraction (f Liq
)
is varied. ........................................................................................................... 111
Figure 4. 15. The Two Flash Levels design showing the new degrees of freedom
and initial values. ............................................................................................. 112
Figure 4. 16. Streams arrangement in the multistage compressor: a) CryoMan
process; b) Two Flash Levels design ............................................................... 114
Figure 4. 17. Effect of increasing the flow rate fraction of liquid fed to the second
flash unit (f 2nd
): a) shaft power demand and minimum driving force
in the MSHE; b) flow rate of Stream 2. ........................................................... 115
Figure 4. 18. Effect of increasing the pressure of the second flash unit (P2nd
):
a) power demand and minimum driving force in the MSHE; b) heat
of vaporisation and flow rate of Stream 3. ....................................................... 116
Figure 4. 19. Effect of increasing the precooling temperature of Stream 3: a) power
demand and minimum driving force in the MSHE; b) evaporating
temperature; c) infeasible heat transfer in the MSHE. ..................................... 117
Figure 4. 20. Effect of composition on power demand and minimum driving force in
the Two Flash Levels design: a) methane; b) ethane; c) propane;
d) n-butane; e) nitrogen. ................................................................................... 118
Figure 4. 21. The Mixing After Precooling design showing the new degrees of freedom
and initial values. ............................................................................................. 121
Figure 4. 22. Effect of increasing flow rate fraction α: a) shaft power demand and
minimum driving force in the MSHE; b) heat of vaporisation of Stream 5. ... 122
Figure 4. 23. a) Evaporating temperature of Stream 5 increases as flow rate fraction α
increases; b) infeasible heat transfer as a result of the increased evaporating
temperature of Stream 5. .................................................................................. 122
Figure 4. 24. Effect of increasing flow rate fraction β on: a) shaft power demand and
minimum driving force in the MSHE; b) the composite curves in the
MSHE. .............................................................................................................. 123
Figure 4. 25. Effect of composition on power demand and minimum driving force in
the Mixing After Precooling design: a) methane; b) ethane; c) propane;
d) n-butane; e) nitrogen. ................................................................................... 124
Figure 5. 1. Objective function progression in the optimisation of the operating variables
of the Bypass design. ......................................................................................... 132
Figure 5. 2. Bypass design. .................................................................................................... 133
Figure 5. 3. Composite curves in the MSHE: a) the Bypass design; b) the CryoMan
process. ............................................................................................................... 135
Figure 5. 4. Objective function progression in the optimisation of the operating variables
of the Two Flash Levels design. ........................................................................ 136
Figure 5. 5. Two Flash Levels design. ................................................................................... 137
Figure 5. 6. Composite curves in the MSHE of the optimised Two Flash Levels design. .... 138
8
Figure A1. 1. Temperature–enthalpy profile of the natural gas stream (Lee, 2001). ............. 150
Figure A1. 2. Temperature–enthalpy profile comparison between the data provided by
Lee (2001) and that according Remeljej and Hoadley (2006). ....................... 151
Figure A1. 3. a) Constructed profile (squares, Starting Point 16) failing to fully cover
the condensing zone; b) condensing enthalpy change increases as
pressure of the stream decreases ..................................................................... 155
Figure A1. 4. Sum of squared errors (SSE) of enthalpy values for each starting point
and for each pressure drop profile assumption. .............................................. 156
Figure A1. 5. Temperature–enthalpy profiles comparison between data published by
Lee (2001) and the stream constructed by optimisation. ................................ 158
Figure A2. 1. Bypass refrigeration cycle. .............................................................................. 159
Figure A2. 2. Optimisation progression of the Bypass design (run 1). .................................. 159
Figure A2. 3. Optimisation progression of the Bypass design (run 2). .................................. 160
Figure A2. 4. Optimisation progression of the Bypass design (run 3). .................................. 161
Figure A2. 5. Two Flash Levels refrigeration cycle. ............................................................. 162
Figure A2. 6. Optimisation progression of the Two Flash Levels design (run 1).................. 163
Figure A2. 7. Optimisation progression of the Two Flash Levels design (run 2).................. 164
Figure A2. 8. Optimisation progression of the Two Flash Levels design (run 3).................. 165
Figure A3. 1. Objective function progression in the optimisation of the operating
variables of the CryoMan process. .................................................................. 169
Figure A3. 2. The CryoMan refrigeration cycle. ................................................................... 170
9
List of Tables
Table 2. 1. Suggested working temperatures for common refrigerants [adapted from
(Lee, 2001, Ch. 1)]. ................................................................................................ 30
Table 2. 2. Cascade refrigeration cycles in the LNG industry [adapted from
(Mokhatab et al., 2014b, Ch. 3.2)]. ........................................................................ 35
Table 2. 3. Models for simulation of refrigeration cycles. ....................................................... 42
Table 2. 4. Mixed refrigerant cycles studied in the open research literature for LNG
production at small scale. ....................................................................................... 50
Table 2. 5. Natural gas and mixed refrigerant compositions [mole %] for the simulation
example in WORK software. ................................................................................. 58
Table 2. 6. Natural gas temperature–enthalpy data for the simulation example in
WORK software. .................................................................................................... 58
Table 2. 7. Comparison of the PRICO cycle simulation between WORK software and
Aspen HYSYS. ...................................................................................................... 59
Table 3. 1. Natural gas and mixed refrigerant composition [mole %] for the modelling
example. ................................................................................................................. 87
Table 3. 2. Natural gas temperature–enthalpy data (from HYSYS, 1 kmol·s-1
) for the
modelling example. ................................................................................................ 87
Table 3. 3. Results comparison of the Two Flash Levels design simulation example. ........... 88
Table 3. 4. Temperature–enthalpy profile of the natural gas stream published by
Lee (2001). ............................................................................................................ 89
Table 4. 1. Natural gas stream data presented by Lee (2001), Del Nogal (2006) and
Zheng (2009). ......................................................................................................... 94
Table 4. 2. Natural gas stream conditions resulted from the optimisation (zero pressure
drop in the MSHE). ................................................................................................ 96
Table 4. 3. Composition of optimised natural gas stream compared to that of Remeljej
and Hoadley (2006). ............................................................................................... 97
Table 4. 4. The CryoMan process (Zheng, 2009): inputs to HYSYS using the natural
gas stream of Table 4.2. ......................................................................................... 99
Table 4. 5. Assumptions in the novel refrigeration cycles to maintain consistency with
Zheng (2009). ....................................................................................................... 100
Table 4. 6. Composition ranges for feasible heat transfer in the Bypass design. ................... 108
Table 4. 7. Operating variables of the Bypass design after the sensitivity analyses. ............. 111
Table 4. 8. Composition ranges for feasible heat transfer in the Two Flash Levels design. . 118
Table 4. 9. Operating variables of the Two Flash Levels design after the sensitivity
analyses. ............................................................................................................... 119
Table 4. 10. Composition ranges for feasible heat transfer in the Mixing After
Precooling design. .............................................................................................. 124
Table 4. 11. Operating variables of the Mixing After Precooling design after the
sensitivity analyses. ............................................................................................ 125
Table 5. 1. Genetic Algorithm parameters for the optimisation of the novel
refrigeration cycles. .............................................................................................. 131
10
Table 5. 2. Temperature–enthalpy profile of the natural gas stream to be liquefied
(Lee, 2001). .......................................................................................................... 132
Table 5. 3. Optimised operating variables of the Bypass design. .......................................... 133
Table 5. 4. Optimised operating variables of the Two Flash Levels design. ......................... 137
Table 5. 5. Operating cost savings comparison (relative to the CryoMan process)
between the novel refrigeration cycles and benchmark processes. ..................... 141
Table A1. 1. Temperature–enthalpy profile of the natural gas stream (Lee, 2001;
Del Nogal, 2006; Zheng, 2009). ....................................................................... 150
Table A1. 2. Natural gas stream profile according to Remeljej and Hoadley (2006)
compared to Lee (2001). ................................................................................... 151
Table A1. 3. Starting points for the natural gas stream optimisation. .................................... 154
Table A1. 4. Optimised values of the vector x* for the natural gas stream (zero
pressure drop in the MSHE). ............................................................................. 157
Table A1. 5. Composition of the optimised natural gas stream and that of Remeljej
and Hoadley (2006). .......................................................................................... 157
Table A1. 6. Temperature–enthalpy data comparison between the optimised stream
and the published data. ...................................................................................... 158
Table A2. 1. Optimised operating variables of the Bypass design (run 1). ........................... 160
Table A2. 2. Optimised operating variables of the Bypass design (run 2). ........................... 161
Table A2. 3. Optimised operating variables of the Bypass design (run 3). ........................... 162
Table A2. 4. Optimised operating variables of the Two Flash Levels design (run 1). .......... 163
Table A2. 5. Optimised operating variables of the Two Flash Levels design (run 2). .......... 164
Table A2. 6. Optimised operating variables of the Two Flash Levels design (run 3). .......... 165
Table A3. 1. Genetic Algorithm parameters for optimisation of the CryoMan process. ....... 168
Table A3. 2. Natural gas stream data (temperature–enthalpy profile). .................................. 169
Table A3. 3. Operating variables of the optimised CryoMan process with six and four
compression stages. ........................................................................................... 170
11
Development of Novel Refrigeration Cycles for Small Scale LNG Processes
Héctor Fernando Almeida Trasviña
The University of Manchester
2016
Abstract – MPhil Thesis
Demand for liquefied natural gas (LNG) is continuously increasing. Commercial
exploitation of small gas reserves is thus becoming accordingly attractive. Natural gas
liquefaction processes are both capital- and energy-intensive. Refrigeration cycles are used
to liquefy the natural gas to temperatures around –160°C; the shaft power demand for
refrigerant compression dominates operating costs. Energy efficiency is usually achieved
in large-scale commercial processes with complex configurations. However, the
complexity of refrigeration cycles in small-scale LNG processes (where production rate is
up to 1 million t per annum) should be low to keep the capital costs relatively low. A trade-
off exists between energy efficiency and capital costs in the design of refrigeration cycles.
In addition, optimisation of the operating variables of the refrigeration cycle is difficult.
Optimisation aims to find the combination of operating variables (including mixed
refrigerant composition) that minimises the shaft power demand for refrigerant
compression in a process with a given configuration and a given liquefaction duty.
However, because of the relatively large number of degrees of freedom available in the
refrigeration cycle and the complex interactions between the operating variables, the
optimisation becomes challenging.
A limited range of refrigeration cycles is studied in the open research literature for the
production of LNG at small scales. Single mixed refrigerant cycles are commonly studied
because they have ‘simple configurations’, although ‘complexity’ of refrigerant cycle
configurations has not been clearly defined. The PRICO cycle is the simplest commercial
refrigeration cycle for LNG production. The so-called ‘CryoMan’ process, developed at
the University of Manchester, modified the structure of the PRICO cycle and achieved
significant shaft power savings (around 8%), compared to the PRICO cycle.
In this work, further structural modifications to the CryoMan process are proposed,
resulting in three novel refrigeration cycles (namely the ‘Bypass’ design, the ‘Two Flash
Levels’ design and the ‘Mixing After Precooling’ design). Design constraints, related to
the number of refrigerant compression stages and the number of streams in the multi-
stream heat exchanger, are defined in this work to limit the complexity of the novel cycles.
To illustrate the benefits of the structural modifications, the configurations are optimised in
an industrially-relevant case study. Sensitivity studies and optimisation are employed to
explore thoroughly the complex interactions between operating variables; a Genetic
Algorithm is applied, to search the solution space and to avoid local optima. The case
study demonstrates that the structural modifications proposed can bring shaft power
savings of up to 3.2% in the case of the Bypass configuration (equivalent to operating cost
savings of £0.69 million per annum for a natural gas feed of 0.75 million t per annum) with
relatively minor increases in the complexity of the refrigeration cycles.
12
Declaration
No portion of the work referred to in the thesis has been submitted in support of an
application for another degree or qualification of this or any other university or other
institute of learning.
Héctor Fernando Almeida Trasviña
13
Copyright Statement
i. The author of this thesis (including any appendices and/or schedules to this thesis)
owns certain copyright or related rights in it (the “Copyright”) and s/he has given
The University of Manchester certain rights to use such Copyright, including for
administrative purposes.
ii. Copies of this thesis, either in full or in extracts and whether in hard or electronic
copy, may be made only in accordance with the Copyright, Designs and Patents
Act 1988 (as amended) and regulations issued under it or, where appropriate, in
accordance with licensing agreements which the University has from time to time.
This page must form part of any such copies made.
iii. The ownership of certain Copyright, patents, designs, trade marks and other
intellectual property (the “Intellectual Property”) and any reproductions of
copyright works in the thesis, for example graphs and tables (“Reproductions”),
which may be described in this thesis, may not be owned by the author and may be
owned by third parties. Such Intellectual Property and Reproductions cannot and
must not be made available for use without the prior written permission of the
owner(s) of the relevant Intellectual Property and/or Reproductions.
iv. Further information on the conditions under which disclosure, publication and
commercialisation of this thesis, the Copyright and any Intellectual Property
University IP Policy (see
http://documents.manchester.ac.uk/display.aspx?DocID=24420), in any relevant
Thesis restriction declarations deposited in the University Library, The University
Library’s regulations (see http://www.library.manchester.ac.uk/about/regulations/)
and in The University’s policy on Presentation of Theses.
14
Acknowledgements
Firstly, I want to acknowledge my supervisors. I would like to thank very much to Dr.
Megan Jobson for her constant patience, support, supervision and feedback throughout the
completion of my research. Many thanks to Prof. Robin Smith for sharing his knowledge
and for giving me great advices and suggestions for my project. Thanks to both for giving
me the opportunity to study under your supervision.
I would like to thank Steve Doyle for his help during the coding and programming stage in
my project. Thanks to my friends and colleagues at the Centre for Process Integration
(CPI) – the best group! – for their help and good advices (and for all the good moments!).
I also want to acknowledge the Mexican National Council of Science and Technology
(CONACyT) for the financial support provided during my studies.
Thanks to my family and friends for all their support.
Chapter 1 Introduction
15
Chapter 1 – Introduction
Natural gas, along with coal and oil, is one of the major sources of energy all around the
world; natural gas currently supplies 22% of the total energy demand globally (IEA, 2012).
Also, natural gas represents the cleanest form of fuel in terms of carbon dioxide emissions,
compared to coal and oil. When oil and coal are used as fuel, emissions of carbon dioxide
and other pollutants such as nitrogen oxides, which are greenhouse gases related to global
warming, are over 40% greater than those produced when natural gas is employed as the
fuel (Kidnay and Parrish, 2006a, Ch. 1.1). The demand for natural gas is likely to increase
in the next decades as a result of the increasing concern on environmental regulations,
because the natural gas would be less affected by economic penalties for CO2 emissions
compared to coal and oil (IEA, 2012).
Natural gas is usually transported from the gas wells using a pipeline network. However,
when the gas reservoirs are located in remote areas or when long distance transportation is
required (i.e. over 3,500 km), natural gas transportation as liquefied natural gas (LNG) via
cargo ships or trucks becomes more profitable, compared to gas pipelines transportation
(Mokhatab et al., 2014a, Ch. 1.2).
Worldwide LNG production in 2014 – around 330 billion m3 – accounted for nearly 10%
of the total natural gas production, according to a report from BP (BP Global, 2015). In
total, 3,400 billion m3 of natural gas were produced globally during the year 2014. Nearly
1,000 billion m3 of natural gas was traded as exports in the same year, 33% from which
was traded as LNG (BP Global, 2015). Furthermore, over two thirds of the total LNG
produced was exported to the Asian market, especially to Japan and South Korea. Qatar is
the world largest LNG producer as it exported around 103 billion m3 of LNG worldwide
during the year of 2014 – exporting mainly to Japan, South Korea and India –, i.e. over
30% of the total LNG production (BP Global, 2015). Australia is also one of the major
contributors to LNG production – nearly 32 billion m3 exported in the same year – and its
production capacity is expected to increase by around 75 billion m3 per year since seven
LNG projects would be starting operation in the next five years (IEA, 2012).
Unconventional gas resources (e.g. shale gas) are becoming increasingly important for
natural gas production, and potentially for LNG production as well. For example, the sharp
increase of natural gas production in the US during the second half of the past decade,
Chapter 1 Introduction
16
coming from the exploitation of shale gas by hydraulic fracturing the shale rock (process
also known as ‘fracking’), could lead to the US exporting natural gas as LNG to Asian
markets, especially to Japan (IEA, 2012). Production and trade of LNG is expected to
increase in the next decades.
1.1 LNG production at small scale
Natural gas will play an important role as a main source of fuel according to the increasing
trend in energy demand (Kumar et al., 2011) and the demand for natural gas is projected to
increase up to 24% of the total energy share in the next two decades according to the New
Policies Scenario presented by the International Energy Agency (IEA, 2012) – see Figure
1.1. Proven gas reserves considered as technically and economically feasible (187 trillion
m3) would allow natural gas production for 50 years if the production rate is assumed
constant (BP Global, 2015). Remaining gas reserves (including proven reserves and
estimated undiscovered resources) that are technically recoverable are calculated as 790
trillion m3 (IEA, 2012).
However, unprocessed gas wells are estimated to account for nearly half of overall natural
gas reserves around the world; many of these reserves are considered as ‘stranded
reservoirs’ due to the limited access for commercial exploitation (Castillo and Dorao,
2010). Since energy demand is continuously increasing (IEA, 2012), small reserves of
natural gas are gaining attention for commercial exploitation as LNG to help meeting the
global energy demand (small scale production rate is up to 1 million t per annum compared
to large scale in which production rate is above 2.5 million t per annum (Mokhatab et al.,
2014b, Ch. 3.3)).
a) b)
Figure 1. 1. World energy demand according to the New Policies Scenario: a) 2015 and b) 2035 [adapted
from IEA (2012)].
Around 50% of the increase in natural gas production is expected to come from
unconventional gas resources, especially from shale gas (IEA, 2012). However, in Europe
Natural
Gas
22%
Coal
28%
Oil
31%
Nuclear
5% Other
14%
2015
(162.7 million GW·h)
Natural
Gas
24%
Coal
24%
Oil
27%
Nuclear
7%
Other
18%
2035
(200.0 million GW·h)
Chapter 1 Introduction
17
and especially in UK, the development of projects for exploiting shale gas resources is still
uncertain as a result of public opinion on potential social and environmental risks that
‘fracking’ processes may produce, e.g. groundwater contamination and exposure to
particulate matter (Reap, 2015), in spite of the energy security and economic growth that
shale gas would bring compared to coal and oil. Even though public opinion might be
influenced by factors such as lack of knowledge of shale gas as a source of energy
(Whitmarsh et al., 2015), the development of shale gas is still dependent on public
acceptance, and on creation of policies and regulations (IEA, 2012). These concerns on
shale gas development also encourage the exploitation of small gas reserves for production
of LNG at small scale to meet the growing energy demand.
1.2 Challenges in design and optimisation of refrigeration cycles for LNG at small
scale
Since natural gas is mostly methane, which has a normal boiling point of –161°C (112 K),
very low temperatures are needed in order to fully liquefy a natural gas stream. In this
liquefaction process, temperatures down to –161°C are reached by employing refrigeration
cycles that provide the cooling needed by the natural gas stream. In vapour compression
refrigeration cycles, widely employed in the LNG industry (Smith, 2005b, Ch. 24.6), a
refrigerant fluid at a low pressure level continuously removes heat from the process stream
and rejects it to a heat sink (e.g. cooling water) after compression at a high pressure level.
In LNG commercial refrigeration cycles, for example the PRICO cycle (Swenson, 1977)
which is the simplest commercial refrigeration cycle in the LNG industry – shown in
Figure 1.2 –, the refrigerant stream at low pressure removes heat not only from the natural
gas stream but also from the refrigerant stream at high pressure level, in order to precool or
fully condense the refrigerant at high pressure (i.e. self-cooling or self-condensing) and
reach very low temperatures (e.g. –161°C) once expanded. Liquefaction of natural gas is a
highly energy consuming process; the shaft power demand for refrigerant compression in
the refrigeration cycle dominates operating costs (Mokhatab et al., 2014b, Ch. 3.2).
Chapter 1 Introduction
18
Figure 1. 2. The PRICO cycle (Swenson, 1977).
The design of the refrigeration cycles should then be energy-efficient in order to minimise
the shaft power demand for compression and hence operating costs. Any savings in power
demand that can be achieved in the refrigeration cycle significantly reduce the operating
costs. For example, Montanez-Morantes (2015, Ch. 3) presented a case study in which the
operating variables (refrigerant flow rate, pressure levels and compressors rotational speed)
of the precooling stage of a propane precooled mixed refrigerant cycle (commercially
dominant refrigeration cycle in the LNG industry) are optimised in order to minimise the
shaft power demand. Shaft power consumption is reduced by 3% compared to the initial
conditions, and such improvements would represent operating cost savings of nearly £0.9
million per annum by only adjusting the operating variables of the refrigeration cycle.
In commercial liquefaction processes, as the efficiency of the refrigeration cycle increases,
its complexity usually increases as well (as illustrated in Figure 1.3), in order to increase
the number of degrees of freedom in the refrigeration cycles (i.e. the operating variables
that can be manipulated by the designer, such as refrigerant flow rate, composition,
pressure levels) to help reducing the overall energy consumption of the refrigeration cycle
while satisfying the natural gas liquefaction demand. For instance, the propane precooled
mixed refrigerant cycle (0.29 kWh·kg-1
of LNG) is more efficient than the PRICO cycle
(0.40 kWh·kg-1
of LNG) (Castillo and Dorao, 2010) because the propane precooling stage
helps reducing the cooling duty of the mixed refrigerant cycle, which can be adjusted to
liquefy the natural gas stream with a lower overall shaft power consumption compared to
that of the PRICO cycle.
However, the increase in efficiency of the propane precooled mixed refrigerant cycle,
compared to the PRICO cycle, is at the expense of increase in the complexity of its
configuration. The propane cycle comprises four heat exchangers for the precooling stage
Chapter 1 Introduction
19
and its own multistage compressor, which are in addition to the multi-stream heat
exchanger (MSHE) and multistage compressor for the mixed refrigerant cycle. Thus, the
increase in complexity also increases significantly the capital costs of the propane
precooled mixed refrigerant cycle compared to those in the PRICO cycle. Low complexity
and compactness of the refrigeration cycle are especially important in small scale LNG
processes. For example, capital investment has to be low when small projects are intended
for short operating periods (Li and Ju, 2010), in order to maintain profitability.
Additionally, since one of the main purposes of such refrigeration cycles is for offshore
liquefaction, there would be a limited plot area for the plant to be built, and weight can also
be a limiting factor as the LNG plant would be built on a platform or on top of a ship
(Castillo and Dorao, 2010).
Figure 1. 3. Complexity of commercial refrigeration cycles increases as efficiency increases [adapted from
(Air Products and Chemicals Inc., 2013)].
Therefore, an economic trade-off exists in the design of refrigeration cycles for small scale
LNG processes between a refrigeration cycle with high efficiency (for energy savings) and
a cycle with a low-complexity configuration (to keep relatively low capital investment);
the refrigeration cycle should then be designed to take advantage of this economic trade-
off.
Chapter 1 Introduction
20
Another challenge is the modelling of mixed refrigerant cycles because of the non-
isothermal nature of mixed refrigerants. The composition of the mixed refrigerant streams
determines the temperature profile and temperature range in which the refrigerant
evaporates for a fixed evaporating pressure (Mokhatab et al., 2014c, Ch. 5.1). Thus,
although the temperatures of the mixed refrigerant at the inlet and outlet of the MSHE
indicate that heat transfer is feasible (i.e. heat transfer with a minimum temperature
approach), intermediate temperatures may lead to infeasible heat transfer as a result of the
continuously-changing temperature of the refrigerant. Therefore, the composite curves
inside the MSHE need to be checked at intermediate temperatures in order to guarantee
feasible heat transfer. There exist models proposed in the open literature, for example (Del
Nogal et al., 2008), in which a user-specified number of intermediate temperatures are
calculated for the hot composite curve and compared to those of the cold composite at the
same enthalpy values. The composite curves are more detailed as the number of
intermediate temperatures increases but at the expense of increased computational time
that is needed for the calculations.
The optimisation of the operating variables of the refrigeration cycle is also challenging.
The optimisation of the operating variables in mixed refrigerant cycles aims to find a
combination of the operating variables (i.e. values of refrigerant flow rate, composition,
pressure levels, etc.) that yields the lowest total shaft power demand for refrigerant
compression. However, the optimisation is difficult because of a relatively large number of
degrees of freedom and also because complex interactions exist between the operating
variables of the refrigeration cycles; a robust optimisation method is needed in order to
fully explore the possible combinations of the operating variables to achieve the minimum
shaft power.
The results obtained from a deterministic optimisation (known as local optima) are
strongly dependent on the initial conditions selected and thus the optimisation should be
performed from different initial conditions to find the global minimum (Edgar et al., 2001,
Ch. 10.1). Different approaches have been proposed in the research open literature for the
optimisation of LNG refrigeration cycles. For example, in the publication of Del Nogal et
al. (2008), a Genetic Algorithm optimisation technique is used to help avoiding local
optima.
Chapter 1 Introduction
21
1.3 Motivation and research objectives
As mentioned in Section 1.2, shaft power demand in refrigeration cycles for LNG
production has a great impact in process economics. Thus, the design of refrigeration
cycles for small scale LNG production should be aimed at minimising shaft power demand
while keeping low complexity.
Only limited refrigeration cycles have been studied in the open research literature. The
“CryoMan” process (Zheng, 2009, Ch. 3) – shown in Figure 1.4 –, a single mixed
refrigerant cycle developed through structural modifications to the PRICO cycle,
demonstrated significant power savings (nearly 8%) compared to the PRICO cycle. The
structural modifications in the CryoMan process include a flash unit after the partial
condenser; the CryoMan process takes advantage from flashing the mixed refrigerant into
vapour and liquid, and partially mixing these flashed streams to create two refrigerant
streams with different compositions. Each refrigerant stream can be expanded to an
independent pressure level. Thus, the CryoMan process takes advantage of creating new
refrigerant streams from a single mixed refrigerant stream. Yet, the configuration of the
CryoMan process remains with relatively low complexity. However, the ‘complexity’ of
the refrigeration cycles has not been clearly defined in the open research literature.
Figure 1. 4. The CryoMan configuration (Zheng, 2009).
The CryoMan process thus might be further structurally modified to develop novel
refrigeration cycles that bring further shaft power savings while keeping the configuration
with relatively low complexity. Hence, the novel refrigeration cycles would exploit the
trade-off between energy-efficiency of the refrigeration cycle and the complexity of its
design.
Chapter 1 Introduction
22
Therefore, this thesis aims to:
Clearly establish criteria to define ‘complexity’ in a refrigeration cycle design.
Develop novel refrigeration cycles, by structurally modifying the CryoMan
process, that bring shaft power savings compared to the CryoMan process. The
novel configurations are of low complex design according to the previous bullet
point. In order to propose meaningful structural modifications to the refrigeration
cycle, the CryoMan process is analysed in detail to identify key operating variables
that could be exploited.
Model and simulate the novel refrigeration cycles. Models available in the open
research literature are used; the models are consistent with those employed in the
CryoMan process in order to perform meaningful comparisons.
Assess and screen the novel refrigeration cycles. Sensitivity studies are performed
to evaluate the novel refrigeration cycles, using the shaft power demand as the
performance indicator. Only the promising designs are selected for optimisation of
their corresponding operating variables.
Optimise the operating variables (e.g. refrigerant flow rate, composition and
streams pressure levels) of the promising novel refrigeration cycles in an
industrially-relevant case study to liquefy a given natural gas stream. The total shaft
power consumption of the novel refrigeration cycles is compared against both the
CryoMan process and the PRICO cycle, and illustrated on an annual operating
costs basis.
1.4 Overview of this Thesis
This thesis is structured in six chapters and three appendixes. In Chapter 1, the current
status and relevance of natural gas as a source of energy is exposed, as well as the
importance of liquefied natural gas (LNG) processes at small scale (i.e. up to 1 million t
per annum). The main challenges in the design of refrigeration cycles for small scale LNG
processes are discussed. The research objectives of this thesis are presented.
In Chapter 2, refrigeration cycles for LNG processes are introduced as well as modelling
for both single component and mixed refrigerant cycles. The research in the open literature
Chapter 1 Introduction
23
regarding modelling and optimisation of refrigeration cycles for LNG processes, and
design of refrigeration cycles for LNG processes at small scale, is critically reviewed also
in Chapter 2. Additionally, the CryoMan process is introduced in this chapter.
The development of the novel refrigeration cycles is presented in Chapter 3. The criteria to
account for ‘complexity’ in the design of a refrigeration cycle is discussed and clearly
defined. The CryoMan process is analysed; options for structural modifications to the
CryoMan process are identified. The modelling of the novel refrigeration cycles is
presented in detailed.
The novel designs are simulated and assessed in Chapter 4 (see also Appendix 1). The
novel refrigeration cycles are evaluated through sensitivity studies performed in their
corresponding degrees of freedom (e.g. refrigerant composition, streams pressure level) in
order to identify promising designs.
A case study is presented in Chapter 5 (see also Appendixes 2 and 3) to illustrate the
benefits from the structural modifications in the novel refrigeration cycles. The operating
variables of the promising novel refrigeration cycles are optimised in order to fully liquefy
a natural gas stream while minimising the total shaft power consumption. The shaft power
demand of the novel cycles is compared against that of the CryoMan process and the
PRICO cycle. The resulting shaft power savings from the optimised novel cycles are
illustrated on an annual operating cost savings basis.
Finally, conclusions on the development of novel refrigeration cycles for small scale LNG
production, and suggestions for extending this research (future work) are presented in
Chapter 6.
Chapter 2 Technology Background and Literature Review
24
Chapter 2 – Technology Background and Literature Review
2.1 Introduction
A simple refrigeration cycle is composed of four elements: an evaporator, a compressor, a
condenser and an expansion device. Figure 2.1 shows the schematic of this cycle, where an
ideal process is described and no pressure drop is assumed through the evaporator and
condenser. The liquid refrigerant provides cooling to the process stream in the evaporator
at the low-pressure level (PLow) while vaporising, following the path 1 to 2 in Figure 2.1a.
The refrigerant vapour is returned to the high-pressure level (PHigh) in the compressor (path
2 to 3). Compression of superheated vapour (i.e. vapour above its dew temperature at the
inlet pressure) is desired since any presence of liquid might produce damages to the
compression equipment (Lee et al., 2002). The compression increases the temperature of
the refrigerant. Isentropic compression (ΔS = 0), also known as adiabatic compression, is
assumed since no heat losses are considered inside the compressor (Dincer and Kanoglu,
2010, Ch. 3.5). The superheated vapour refrigerant rejects heat in the condenser (path 3 to
4) to an external heat sink, e.g. cooling water or another process stream. The refrigerant is
desuperheated and condensed. Finally, the pressure of the liquid refrigerant is released
from PHigh to PLow through an expansion device, e.g. a throttle valve, and its temperature
decreases accordingly. This is represented by the path 4 to 1. Isentropic (ideal) expansion
is also assumed. The refrigerant is partially vaporised as a consequence of the pressure let
down. Figure 2.1b illustrates the refrigeration cycle in a pressure–enthalpy diagram.
Isentropic compression is represented as the solid line from PLow to PHigh. Isentropic
expansion is the solid line from PHigh to PLow.
Figure 2. 1. Ideal refrigeration cycle: a) configuration; b) P–H diagram [adapted from (Smith, 2005b, Ch.
24.6)].
Chapter 2 Technology Background and Literature Review
25
In practice, neither compression nor expansion is actually ideal and deviations from these
theoretical paths are observed, as shown in dashed lines in the P–H diagram of Figure 2.1b.
Deviation from isentropic compression incurs a higher change of enthalpy of the
refrigerant (dashed line from PLow to PHigh in Figure 2.1b) which increases its temperature
accordingly. Higher duties of the heat sink are therefore required in the condenser to
convert vapour refrigerant into liquid compared to the ideal compression. Non-isentropic
expansion leads to a refrigerant stream with a greater vapour fraction compared to
isentropic expansion, and now isenthalpic expansion can be assumed (Dincer and Kanoglu,
2010, Ch. 3.8) as shown with a vertical dashed line in Figure 2.1b. As a result, there is less
liquid refrigerant for providing cooling; higher flow rates of the refrigerant are then needed
in the evaporator to provide cooling to the process stream compared to the ideal expansion.
Both non-ideal processes increase of the volumetric flow rate of the refrigerant, leading to
an increase of the shaft power required for refrigerant compression.
The simple cycle can be modified to reduce the power consumption. Three examples are
presented in Figure 2.2. Figure 2.2a shows a refrigeration cycle with two expansion levels.
A flash unit (known as economiser) separates vapour and liquid phases of the refrigerant
stream after the first expansion. The liquid phase is further expanded and provides cooling
to the process stream whilst the vapour is sent directly to the compressor. As can be seen
from Figure 2.2a, multistage compression is used; the vapour from the first expansion goes
to the high pressure compression stage. The flowrate in the low pressure compressor is
reduced and thus, savings in shaft power are achieved (Borgnakke and Sonntag, 2009, Ch.
11.12). Intercooling, e.g. with cooling water, can be used after the compression of the low
pressure refrigerant, before mixing with the vapour at intermediate pressure, as illustrated
in Figure 2.2b. The intercooling reduces the volumetric flow rate at the inlet of the high
pressure compressor, leading to shaft power savings (Borgnakke and Sonntag, 2009, Ch.
11.12). Alternatively, a multilevel refrigeration cycle can be used. This is displayed in
Figure 2.2c with a two-level refrigeration cycle, in which a portion of the liquid after the
first expansion can be used to provide cooling to the process stream at a warmer
temperature, whilst the remaining liquid is further expanded and provides cooling at a
colder temperature. Shaft power savings comes from the reduced flow rate in the low
pressure compressor (Smith, 2005b, Ch. 24.6).
Chapter 2 Technology Background and Literature Review
26
Figure 2. 2. Configuration options to reduce shaft power demand of a simple cycle: a) multistage expansion;
b) intercooling: c) multilevel refrigeration [adapted from (Smith, 2005b, Ch. 24.6)].
LNG processes are highly energy-intensive; the operating costs are dominated by the shaft
power requirement for refrigerant compression in the refrigeration cycle (Mokhatab et al.,
2014b, Ch. 3.2). Therefore, minimising the shaft power consumption in the refrigeration
cycle significantly reduces the operating costs. The ‘efficiency’ of the refrigeration cycle
can thus be defined as the shaft power consumption per unit mass flow rate of LNG
produced (i.e. specific shaft power consumption, kWh·kg-1
of LNG) and be used as a
performance indicator for comparison against other refrigeration cycles (Mokhatab et al.,
2014b, Ch. 3.2). A refrigeration cycle with a low specific shaft power consumption value is
more efficient than a refrigeration cycle with a high specific shaft power consumption
value because the former requires less energy input to liquefy a fixed flow rate of natural
gas compared to the latter.
To determine the shaft power consumption of a refrigeration cycle and relative savings
achieved with different configurations and operating conditions, mathematical modelling is
required. The following Section 2.2 addresses the modelling of refrigeration cycles. Pure
component and mixed refrigerant cycles are considered separately in Section 2.2.1 and
Section 2.2.2, respectively, and cascade cycles are also introduced in Section 2.2.3. Where
appropriate, commercial refrigeration cycles for LNG processes are also presented.
2.2 Modelling of refrigeration cycles for LNG processes
Modelling is used to represent a process configuration through mathematical equations
allowing data to be collected from the simulations, including flow rates, temperatures or
Chapter 2 Technology Background and Literature Review
27
energy needs (Smith, 2005a, Ch. 1.3). Thus, different operating variables or process
configurations can be evaluated and screened. To simulate a refrigeration cycle, enthalpy
calculations are required to perform energy balances and entropy calculations are needed
because isentropic compression models are employed. Equations of state can predict
pressure-volume-temperature relations, which allow calculating enthalpy and entropy
values, and vapour-liquid equilibrium (Elliot and Lira, 1999, Ch. 6.5).
The Peng–Robinson equation of state (Equation 2.1) is shown to calculate such volumetric
relations with good accuracy for hydrocarbons and light gases, as a function of their
critical properties – temperature and pressure – and acentric factor (ω). Calculated enthalpy
values and vapour pressures of these components have also been shown in good agreement
with experimental data in a wide range of temperatures and pressures: enthalpy values are
calculated in ranges from –157°C to 10°C and from 14 bar to above 130 bar, vapour
pressures are shown to have a relative error of < 1%, and (Peng and Robinson, 1976).
These components – including C1 to C4 and nitrogen – are common in LNG processes
since they are the major constituents of the natural gas streams to be liquefied, and also the
main components of the refrigerant mixtures employed in the refrigeration cycles. The
Peng–Robinson equation can be thus considered as a reliable equation for calculating
physical and thermodynamic properties of the refrigerant fluid in refrigeration cycles – and
natural gas stream – for LNG processes.
Enthalpy calculations can be performed with the Peng–Robinson equation, when combined
with the thermodynamic relation in Equation 2.2. Vapour-liquid equilibrium calculations
can be carried out when combined with Equation 2.3. For mixtures, the parameters a and b
are calculated with Equations 2.4 and 2.5, according to the corresponding mole fraction of
the ith and jth component in the mixture (Peng and Robinson, 1976).
𝑃 = 𝑅𝑇
𝜐−𝑏−
𝑎(𝑇)
𝜐2+2𝑏𝜐−𝑏2 (2.1a)
𝑎(𝑇) = 0.45724𝑅2𝑇𝑐
2
𝑃𝑐∙ [1 + 𝜅(1 − 𝑇𝑟
1/2)]
2 (2.1b)
𝑏 = 0.07780𝑅𝑇𝑐
𝑃𝑐 (2.1c)
𝜅 = 0.37464 + 1.54226𝜔 − 0.26992𝜔2 (2.1d)
Chapter 2 Technology Background and Literature Review
28
𝐻 − 𝐻∗ = 𝑅𝑇(𝑍 − 1) + ∫ [𝑇 (𝜕𝑃
𝜕𝑇)
𝜐− 𝑃] 𝑑𝜐
𝜐
∞ (2.2)
𝑙𝑛𝑓
𝑃= ∫ (
𝑣
𝑅𝑇−
1
𝑃) 𝑑𝑃
𝑃
0 (2.3)
𝑎 = ∑ ∑ 𝑥𝑖𝑥𝑗(1 − 𝛿𝑖𝑗)√𝑎𝑖𝑎𝑗𝑗𝑖 (2.4)
𝑏 = ∑ 𝑥𝑖𝑏𝑖𝑖 (2.5)
The enthalpy and entropy of a stream of known composition (pure or mixed refrigerant)
can be calculated if its pressure and temperature are known. Alternatively, the temperature
or pressure of a stream can be obtained if any two of the remaining values (temperature,
pressure, enthalpy, entropy) are known.
For example, assuming that the pressure levels in the refrigeration cycle of Figure 2.1 and
the refrigerant composition are known (either pure or mixed refrigerant) for the ideal
refrigeration cycle, the enthalpy of the refrigerant leaving the condenser can be calculated
at PHigh and the temperature of the condenser. Similarly, the temperature of the refrigerant
after the expansion can be obtained with PLow and the same enthalpy value if isenthalpic
expansion is assumed.
The shaft power required for compression in the ideal refrigeration cycle of Figure 2.1
(either pure or mixed refrigerant) can be calculated with an energy balance around the
compressor, according to Equation 2.6 (Dincer and Kanoglu, 2010, Ch. 3.5):
𝑊 = 𝑚 ∙ (𝐻𝐶𝑜𝑚𝑝𝑜𝑢𝑡 − 𝐻𝐶𝑜𝑚𝑝
𝑖𝑛 ) (2.6)
where HCompin
and HCompout
are the enthalpies at the inlet and outlet of the compressor,
respectively. HCompin
can be obtained with PLow and the temperature after the evaporator,
whilst HCompout
can be determined with PHigh and the entropy at the compressor inlet as
isentropic compression is assumed.
Alternatively, the shaft work can be estimated using the following Equation 2.7 (Kyle,
1999b, Ch. 15.1):
𝑊 = 𝛾 ∙ 𝐹𝑖𝑛 ∙ 𝑃𝑖𝑛
𝛾 − 1[1 − (
𝑃𝑜𝑢𝑡
𝑃𝑖𝑛)
𝛾 − 1
𝛾] (2.7)
where W = Compression shaft work [W]
γ = Heat capacity ratio of inlet stream (Cp·Cv-1
)
Fin = Inlet volumetric flow rate [m3·s
-1]
Pin, out = Inlet and outlet pressure [Pa]
Chapter 2 Technology Background and Literature Review
29
In an isentropic compression is assumed an ideal gas in a compression path PVγ = constant,
and no heat is transferred to or from the refrigerant (Kyle, 1999a, Ch. 3.3). As stated in
Section 2.1, however, the compression in non-ideal refrigeration cycles deviates from the
isentropic path. An efficiency term of the compression can then be considered, which can
be included into the model for calculating the shaft power demand, as in Equation 2.8
(Kyle, 1999b, Ch. 15.1):
𝑊 = (𝛾
𝛾−1)
𝐹𝑖𝑛 ∙ 𝑃𝑖𝑛
𝜂𝐼𝑆[1 − (
𝑃𝑜𝑢𝑡
𝑃𝑖𝑛)
𝛾−1
𝛾] (2.8)
ηIS is the isentropic efficiency of the compression, that is the ratio of the shaft work of an
isentropic compression to the actual shaft work needed for compression, and accounts for
losses of power energy including friction in the compressor.
The modelling of the composite curves in the evaporator depends on the type of refrigerant
used in the refrigeration cycle, i.e. pure component or mixed refrigerant. Pure components
evaporate at constant temperature (if a zero pressure drop is assumed) and thus, the
pressure is calculated according to the temperature needed in the evaporator (which should
be lower than the outlet temperature of the process stream by a minimum temperature
difference, to guarantee feasible heat transfer). On the other hand, mixed refrigerants
evaporate over a wide temperature range and their evaporating profile depends on their
corresponding composition (with ΔP = 0); thus, calculation of intermediate temperatures is
required to evaluate feasible heat transfer considering a minimum temperature difference.
In Section 2.2.1, modelling of evaporators with pure refrigerants is briefly described,
whilst modelling for mixed refrigerants is presented in detail in Section 2.2.2. In Section
2.2.3, cascade cycles are introduced.
2.2.1 Modelling of pure refrigerant cycles
In pure refrigerant cycles, cooling is delivered by a refrigerant which is a single
component. When a pure component refrigerant evaporates at a fixed pressure, its
temperature remains constant. Therefore the selection of the refrigerant is influenced by
the final temperature that is needed by the process stream. Table 2.1 shows some common
refrigerants with their suggested working temperatures. For each component, the upper
limit temperature (TMAX) is either ambient or a temperature in which the latent heat of
vaporisation is 50% of that at atmospheric pressure, and the lower limit temperature (TMIN)
is the normal boiling point (Lee, 2001, Ch. 1).
Chapter 2 Technology Background and Literature Review
30
Table 2. 1. Suggested working temperatures for common refrigerants [adapted from (Lee, 2001, Ch. 1)].
Nitrogen Methane Ethane Propane* i-Butane* n-Butane*
TMIN [K] 75 120 187 232 264 274
TMAX [K] 120 163 273 310 310 310
*denotes refrigerants that can reject heat to an ambient utility (e.g. cooling water).
The pressure in the evaporator is calculated so that the temperature of the vaporising
refrigerant (saturation temperature) is lower than the final temperature of the process
stream by a minimum temperature difference (ΔTMIN). The Antoine correlation or an
equation of state (e.g. Peng–Robinson) can be used to calculate the pressures in the
evaporator and in the condenser.
However, from Table 2.1 it can be deducted that the temperature provided by simple
refrigeration cycles using single component refrigerants is limited to around –40°C (232
K) since only propane, i-butane and n-butane can reject heat to an ambient utility, e.g.
cooling water. Lower temperatures (e.g. –161°C for LNG processes) can be achieved by
pure component refrigeration cycles when cascade configurations are used (Borgnakke and
Sonntag, 2009, Ch. 11.12).
2.2.2 Modelling of mixed refrigerant cycles
In mixed refrigerant cycles, the refrigerant is a mixture: light hydrocarbons (C1 to C4) and
nitrogen are commonly used in LNG processes (Mokhatab et al., 2014b, Ch. 3.2). Unlike
with single components, the temperature of mixed refrigerants continuously changes as the
evaporation takes place (see Figure 2.3). The composition of the refrigerant determines the
profile and the temperature range in which the refrigerant evaporates for a fixed
evaporating pressure (Radermacher, 1989).
Figure 2. 3. Refrigerant evaporating profile (at constant pressure): a) pure component; b) mixed refrigerant
[adapted from (Radermacher, 1989)].
Chapter 2 Technology Background and Literature Review
31
The composition of the refrigerant is selected to have close temperature approaches to the
ΔTMIN in the composite curves, which represent shaft power consumption reduction. Since
the temperature of both the process stream and the mixed refrigerant are continuously
changing, feasible heat exchange (i.e. heat transfer between hot and cold streams with a
minimum temperature difference) needs to be checked at intermediate temperatures and
included into the modelling of the evaporator. In order to evaluate the feasibility of heat
transfer at intermediate temperatures, the overall enthalpy change in the mixed refrigerant
stream can be divided into a user-specified number of intervals, as shown in Figure 2.4,
and the temperature for each interval can be then calculated with an equation of state (e.g.
Peng–Robinson) using each enthalpy value, the refrigerant composition and the pressure of
the evaporator (zero pressure drop assumed). Similarly, intermediate temperatures of the
process stream can be calculated at the same enthalpy values.
Figure 2. 4. Temperature–enthalpy profile calculation for a mixed refrigerant stream (ΔP = 0 in the
evaporator).
The difference between the calculated temperatures of the mixed refrigerant stream and the
process stream, at the ith interval (ΔTi), is compared against the ΔTMIN value selected in
order to check feasible heat transfer. If all the ΔTi’s are greater than or equal to the value of
ΔTMIN then the design is feasible (in terms of heat transfer). If any temperature difference
is less than the specified minimum temperature difference, however, the design is
infeasible and a degree of freedom would need to be manipulated, such as increasing the
flow rate of the mixed refrigerant, increasing the pressure ratio (i.e. ratio of outlet to inlet
compression pressure) or changing the composition of the mixed refrigerant (Lee, 2001,
Ch. 4). After a degree of freedom is manipulated, the calculation of the temperature–
enthalpy of the mixed refrigerant stream would need to be performed again in order to
compare its evaporating temperature, at each ith interval, against the corresponding
condensing temperature of the process stream and thus, checking that heat transfer is
Chapter 2 Technology Background and Literature Review
32
feasible (i.e. a positive minimum temperature difference value between the condensing
temperature of the process stream and the evaporating temperature of the mixed
refrigerant).
Very low temperatures are needed in LNG processes (e.g. –161°C); the mixed refrigerant
can be cooled by an external utility (i.e. another refrigeration cycle) or self-cooled to
achieve such low temperatures. In self-cooling, the low pressure refrigerant provides
cooling to the process stream and also to the high pressure refrigerant (as shown in Figure
2.5), allowing the mixed refrigerant to achieve a colder temperature before the expansion
compared to the case without self-cooling. After the expansion, the mixed refrigerant can
thus achieve a very low temperature (e.g. –161°C). In addition, the mixed refrigerant
stream might be fully condensed during self-cooling and, once expanded, the liquid
fraction of the refrigerant would be greater, compared to without self-cooling, which helps
reducing the refrigerant flow rate needed for providing cooling.
Multi-stream heat exchangers (MSHE) are commonly used in the LNG industry to
accommodate several streams in the same equipment unit, e.g. when refrigerant self-
cooling is employed. To model an MSHE as the evaporator in which a mixed refrigerant is
self-cooled, the calculation of the T–H curves is needed for both cold and hot composite
curves each time the degrees of freedom of the refrigeration cycle are manipulated since
the conditions of the refrigerant (pressure and temperature) and its composition will affect
both composite curves.
The simplest commercially established mixed refrigerant cycle for LNG processes is the
PRICO (Poly Refrigerated Integrated Cycle Operation) process, developed and patented by
Black & Veatch (Swenson, 1977). Figure 2.5 shows the configuration of the PRICO
process. It consists only of a single mixed refrigerant stream with two pressure levels. The
refrigerant is a mixture that can be composed of C1 to C5 and N2 (Swenson, 1977). The
high pressure refrigerant and the natural gas stream are condensed by the low pressure
refrigerant stream in a single MSHE unit.
Chapter 2 Technology Background and Literature Review
33
Figure 2. 5. PRICO refrigeration cycle (Swenson, 1977).
This process shows low efficiency (0.40 kWh·kg-1
of LNG) when compared to more
complex cycles such as the APCI C3MR cycle, which is a cascade cycle (0.29 kWh·kg-1
of
LNG) (Castillo and Dorao, 2010).
2.2.3 Modelling of cascade cycles
As mentioned in Section 2.2.1, cascade cycles are employed when very low temperatures
are required (e.g. –161°C in the LNG industry). In cascade cycles, two or more cycles
operate at different temperature levels. The refrigerant cycle at the coldest temperature
liquefies the natural gas stream, and rejects heat to a warmer refrigeration cycle; the
refrigeration cycle at the warmest temperature rejects heat to an ambient heat sink (e.g.
cooling water). For example, Figure 2.6 presents a cascade cycle consisting of two simple
refrigeration cycles. Cycle 1 in Figure 2.6 provides cooling to the process stream and
rejects heat to Cycle 2; Cycle 2 rejects heat to an ambient utility (e.g. cooling water). Each
cycle has an independent refrigerant and, therefore, an independent compressor. The
refrigerant in each cycle can be either pure component or mixed refrigerant.
MSHE units can also be employed in cascade cycles. Cascade cycles in the LNG industry
use MSHEs for precooling both the natural gas stream as well as the lower temperature
refrigerant. The calculation of the composite curves in MSHEs as evaporators for cascade
cycles can be performed as detailed in Section 2.2.1 and Section 2.2.2, depending if pure or
mixed refrigerants are used.
Chapter 2 Technology Background and Literature Review
34
Figure 2. 6. Cascade refrigeration cycle with two temperature levels of refrigeration.
In cascade cycles, each refrigeration cycle can use a single component refrigerant or a
mixed refrigerant; alternatively, the cascade cycle can combine a pure refrigerant cycle and
a mixed refrigerant cycle. Table 2.2 compares the refrigerants and the relative efficiency of
three commercial cascade cycles applied to LNG processes; Figure 2.7 shows a simplified
schematic of each cascade process presented in Table 2.2. The Phillips optimized cascade
cycle, illustrated in Figure 2.7a, consists of three pure refrigerant cycles (propane, ethylene
and methane) for the liquefaction of the natural gas stream (Hughes, 1971). The propane
precooled mixed refrigerant cycle (C3MR), developed by Air Products and Chemicals Inc.
(Figure 2.7b), uses a propane cycle and a mixed refrigerant cycle; the propane cycle
precools the natural gas stream and the mixed refrigerant, whereas the mixed refrigerant
fully liquefies the natural gas stream (Gaumer and Newton, 1973). The dual mixed
refrigerant (DMR), developed by Shell (Figure 2.7c), consists of two mixed refrigerant
cycles in cascade; in the first mixed refrigerant cycle, both the natural gas and the second
mixed refrigerant are precooled, whilst in the second mixed refrigerant cycle the natural
gas is fully liquefied (Grootjans et al., 2002).
In Table 2.2, the Phillips optimized cascade cycle shows a higher efficiency than the C3MR
and the DMR cycle, but at the expense of higher complexity and associated initial capital
investment (Mokhatab et al., 2014b, Ch. 3.2). The APCI C3MR cycle is the dominant
refrigeration technology in the LNG industry as over 70% of the installed plants in the
world use this cascade cycle (Mortazavi et al., 2012).
Chapter 2 Technology Background and Literature Review
35
Table 2. 2. Cascade refrigeration cycles in the LNG industry [adapted from (Mokhatab et al., 2014b, Ch.
3.2)].
No. Cycles
Precooling
Stage Liquefaction Stage
Efficiency relative to
cascade cycle
Phillips optimized
cascade cycle 3
Propane (C3)
Ethylene (C2=)
Methane (C1) 1.00
Propane precooled
mixed refrigerant 2 Propane (C3) MR (C1 to C3 + N2) 1.15
Dual mixed
refrigerant 2 MR (C2 + C3) MR (C1 to C3 + N2) 1.05
Figure 2. 7. Commercial cascade cycles: a) Phillips cascade cycle; b) Propane precooled mixed refrigerant
cycle; c) Dual mixed refrigerant cycle.
2.3 Research literature on refrigeration cycles for LNG processes
The research literature is reviewed and organised as follows: in Section 2.3.1
methodologies for systematic design of refrigeration cycles are presented and the
modelling approach is analysed. Most of these design methodologies include a case study
or an example problem in which a refrigeration cycle is designed and optimised using their
corresponding modelling approach. These case studies and the optimisation methods
selected in each methodology are discussed in Section 2.3.2. In Section 2.3.3 a review on
the design of refrigeration cycles for small scale LNG processes is presented. Section 2.3.4
provides an introduction to the “CryoMan” process (Zheng, 2009, Ch. 3), which is an
single mixed refrigerant cycle developed by structural modifications to the PRICO cycle.
In Section 2.3.5 WORK software is presented and tested to simulate refrigeration cycles.
Chapter 2 Technology Background and Literature Review
36
WORK is in-house software developed in the Centre for Process Integration at the
University of Manchester for modelling, simulation and optimisation of refrigeration
cycles (see Section 2.3.5). WORK software is used to simulate and optimise the CryoMan
process (Zheng, 2009, Ch. 3).
2.3.1 Modelling and design of mixed refrigerant cycles
The use of mathematical programming and thermodynamic-based approaches has been
proposed to design refrigeration cycles systematically. For instance, Shelton and
Grossmann (1986) developed a mathematic model for the systematic design of pure
component multilevel refrigeration cycles. The model is based on a superstructure, which
is a conceptual configuration of a refrigeration cycle with multiple discrete temperature
levels. Refrigeration at two different temperature levels are connected using a presaturator
in which the liquid refrigerant from the upper temperature level is contacted with the
vapour compressed from the lower temperature level (see Figure 2.8). According to
Shelton and Grossmann (1986), the use of presaturators yields only linear relations
between power demand, and evaporating and condensing duties in the refrigeration levels
(because the refrigerant vapour is always saturated at the inlet of the compressor;
otherwise, the degree of superheating is dependent on the duty in the intercooler). To
evaluate the refrigeration cycle with different combinations of temperature levels, binary
variables are introduced to activate or deactivate a temperature level. The resulting
mathematical model is a Mixed Integer Linear Programming (MILP) problem.
Figure 2. 8. Two temperature levels refrigeration cycle using a presaturator (Shelton and Grossmann, 1986).
Chapter 2 Technology Background and Literature Review
37
The methodology presented by Shelton and Grossmann (1986) considers only refrigeration
cycles using pure component refrigerants. Further, the selection of the refrigerant is not
explicitly included into the design methodology. To create the superstructure for the
multilevel refrigeration cycle, the number and temperatures of the levels are chosen by the
designer, so the final results are likely to depend on the temperatures selected. Because of
the linear relations assumed, the model only applies for refrigeration cycles using
presaturators. The shaft power demand in the refrigeration cycles is calculated using
correlations from the pressure–enthalpy diagram, between the heat of vaporisation of the
selected pure refrigerant and the refrigeration cycle working temperatures (Shelton and
Grossmann, 1985). An isentropic efficiency factor is considered in the correlations for
calculating the shaft power demand; however, the correlations only hold when assuming
compression of saturated vapour of single component refrigerants.
Linnhoff and Dhole (1992) proposed an approach to design pure component refrigeration
cycles integrated with the heat exchanger network (HEN) for sub-ambient processes.
Conventional composite curves (T–H profiles) shows the heat loads in the process but
ignore energy in the form of power; thus Linnhoff and Dhole (1992) replaced the
temperature axis of the T–H diagram in the conventional pinch analysis with the Carnot
factor 𝜂𝐶 = (1 −𝑇0
𝑇), and allowed for including shaft work in the diagram. In the Carnot
factor, T0 is the ambient temperature and T is the temperature at which energy (heat and
shaft work) is available. The area between the resulting composite curves (σTHEN) is
demonstrated to be proportional to the shaft work input in the refrigeration cycle. Once a
refrigeration cycle is modelled and represented in the T–H diagram, changes in the
refrigeration cycle (such as number of refrigeration levels or the temperature of each level)
can be suggested to reduce the area between the composite curves (as shown in Figure 2.9)
in order to reduce the total shaft power demand. The main strength of the methodology is
to investigate potential shaft power savings with different configurations without the need
of modelling each possible design but rather calculating the area between the composite
curves and comparing it to a base case design. Linnhoff and Dhole (1992) presented a case
study for the design of a refrigeration cycle and a HEN in an ethylene recovery process
(discussed in Section 2.3.2).
Chapter 2 Technology Background and Literature Review
38
Figure 2. 9. Design methodology for low temperature processes presented by Linnhoff and Dhole (1992).
The methodology, however, is limited to refrigeration cycles using pure components and
the selection of the refrigerant is not included. The choice of a different configuration of
the refrigeration cycle (e.g. including an additional refrigeration level, changing the
temperature level of an existing refrigeration level) does not follows an optimisation-based
approach but is rather guided only by judgement.
Lee et al. (2002) proposed a design methodology for mixed refrigerant cycles for LNG
processes in which the composition of the refrigerant is optimised. The PRICO single
mixed refrigerant cycle is used in the design methodology. Initial conditions (refrigerant
composition, condensing and evaporating pressure, and refrigerant flow rate) are first
selected and the hot composite curve is generated. The hot composite curve is created by
combining the T–H profiles of the natural gas stream and the self-cooled refrigerant, as in
the conventional pinch analysis. For each stream, the T–H profile is obtained by dividing
its corresponding overall enthalpy change into a user-specified number of intervals for
which the temperature is calculated according to the corresponding stream pressure and
composition. An ‘ideal’ cold composite is generated as a duplicate of the hot composite but
colder by a minimum temperature difference (ΔTMIN), as illustrated in Figure 2.10. In
Lee’s work, a minimum temperature difference of 5°C is assumed.
Figure 2. 10. Generation of the ‘ideal’ cold composite curve (Lee, 2001, Ch. 4).
Chapter 2 Technology Background and Literature Review
39
The composition of the mixed refrigerant is optimised as a non-linear programming (NLP)
problem using a deterministic algorithm, and three options of objective function are
defined: i) minimisation of the largest negative difference between the evaporating
temperatures of the ‘ideal’ cold composite curve and those of the cold refrigerant stream
(because in the PRICO cycle there is only one cold refrigerant stream in the MSHE, the
evaporating profile of the cold refrigerant stream is equivalent to the evaporating profile of
the cold composite curve); ii) minimisation of the sum of the negative differences between
the evaporating temperatures of the ‘ideal’ cold composite curve and those of the cold
refrigerant stream; and iii) minimisation of shaft power demand for refrigerant
compression. The cold refrigerant composition is only constrained to have evaporating
temperatures less than those of the hot composite at each interval, in order to have feasible
heat transfer. That is, the evaporating temperatures of the cold composite curve should be
colder than those of the hot composite at each interval, but not necessarily colder by a
minimum temperature difference.
The composite curves are updated as the refrigerant composition changes. When the
refrigerant composition optimisation is completed, either the refrigerant flow rate or the
pressure levels in the refrigeration cycle is decreased (by judgement or optimisation,
although no details are provided regarding the mathematical formulation for the
optimisation) in order to reduce the shaft power demand for refrigerant compression. This
decrease of the refrigerant flow rate, or pressure levels, changes the composite curves.
Thus, the optimisation of the refrigerant composition is performed again. This procedure is
repeated until no further minimisation of the shaft power can be obtained from reducing
the refrigerant flow rate, or reducing the pressure levels, and with feasible heat transfer.
The proposed methodology allows manipulating the operating variables of the mixed
refrigerant cycle, including the refrigerant composition, using an optimisation approach.
Also, the model provides an approach to consider the non-isothermal evaporation of mixed
refrigerants by evaluating intermediate temperatures of the cold refrigerant stream for
feasible heat transfer in the MSHE. However, the operating variables of the refrigeration
cycle (refrigerant composition, refrigerant flow rate and pressure levels) are not
manipulated simultaneously; that is, interactions between the operating variables are not
considered, leading to results that are likely to be non-optimal. Moreover, even though the
refrigerant composition is optimised to minimise the differences between the evaporating
temperature of cold refrigerant and that of the ‘ideal’ cold composite curve (created to be
Chapter 2 Technology Background and Literature Review
40
colder than the hot composite curve by a minimum temperature difference), feasible heat
transfer is defined in the model only as a positive temperature difference at each interval
between the hot composite and the cold composite curves, but not a positive difference by
a minimum temperature approach. Furthermore, no details of the compressor model
employed are provided or discussed. Single stage compression is assumed.
Vaidyaraman and Maranas (2002) implemented a mathematical approach to model and
design systematically cascade refrigeration cycles using mixed refrigerants with multistage
refrigeration. Unlike conventional cascade refrigeration cycles in which each different
refrigerant stream has its own compression system, the multistage refrigeration cycles
proposed by Vaidyaraman and Maranas (2002) (see Figure 2.11) consist of a single mixed
refrigerant stream with only one compression system. When partially condensed and
flashed, the mixed refrigerant creates two streams (vapour and liquid) with different
compositions. The number of levels in the cascade refrigeration cycle and the number of
refrigeration stages in each refrigeration level are user-specified. It is assumed that the hot
refrigerant stream at each MSHE leaves as saturated liquid, and the cold refrigerant stream
at each MSHE leaves as saturated vapour.
Vaidyaraman and Maranas (2002) presented a case study for the design of a cascade cycle
with multistage refrigeration to cool a methane-rich process stream down to –58 °C (215
K). The design variables of the refrigeration cycle include pressure levels, refrigerant
composition and quality of the refrigerant after each partial condensation.
The model employed by Vaidyaraman and Maranas (2002) allows exploiting the
composition of mixed refrigerants, to create two streams with different compositions after
partial condensing and flashing a single mixed refrigerant. The model is thus useful for
designing single- and multiple-stage mixed refrigerant cycles, including cascade cycles.
However, in the model, the constraint of minimum driving force for feasible heat transfer
(ΔTMIN) between hot and cold streams is only applied at the inlet and outlet of the heat
exchangers and thus, feasible heat transfer at intermediate temperatures inside each MSHE
is not guaranteed. Additionally, as with Lee et al. (2002), the model considers only single
stage compression of the refrigerant, and also ideal (isentropic) compression is assumed.
Similarly, Del Nogal et al. (2008) presented a systematic methodology for the design and
optimisation of refrigeration cycles using mixed refrigerants for processes such as the
liquefaction of natural gas. Multistage refrigeration (Figure 2.11) and cascade cycles with
Chapter 2 Technology Background and Literature Review
41
multistage refrigeration are considered in the proposed methodology. For the case of
multistage refrigeration cycles, it is considered that the process stream is liquefied as it is
fed through a series of MSHEs, as in the designs proposed by Vaidyaraman and Maranas
(2002), so the cooling duty of the process stream is divided into the selected number of
refrigeration stages (user-defined). The high pressure refrigerant is partially condensed and
flashed to generate a vapour and a liquid stream of different composition. However, this
time the liquid can be subcooled in the subsequent heat exchanger stage. Additionally, the
cold refrigerant leaving any heat exchanger can be superheated (to avoid wetness at the
inlet of the compression system).
Figure 2. 11. Multistage refrigeration cycle.
Compared to the methodology presented by Vaidyaraman and Maranas (2002), the
compression is no longer considered isentropic as the isentropic efficiency is introduced.
Multistage compression with intercooling is also adopted in the methodology developed by
Del Nogal et al. (2008). The maximum ratio of outlet to inlet compression pressures (PRAT)
is set to 5, which is an industrial common practice according to the authors. Additionally,
feasible heat transfer checks between the hot and cold streams are included in the
modelling of the MSHEs. Once both composite curves are constructed, a user-specified
number of intermediate temperatures are compared between the hot and cold composite
curves. The design has feasible heat transfer when a positive difference of temperatures is
equal to or greater than a minimum temperature approach (ΔTMIN), which is specified by
the designer.
Table 2.3 summarises the modelling approaches reviewed in the open literature for the
design of refrigeration cycles highlighting the main assumptions (e.g. consideration of
feasible heat transfer at intermediate temperatures).
Chapter 2 Technology Background and Literature Review
42
Table 2. 3. Models for simulation of refrigeration cycles.
Refrigeration cycle
Isentropic
efficiency
Multistage
compression
Feasible heat
transfer
Shelton and Grossmann (1986) Pure refrigerant N/A
Linnhoff and Dhole (1992) Pure refrigerant N/A
Lee et al. (2002) SMR - Partially covered
Vaidyaraman and Maranas (2002) Mixed refrigerants
and Cascade cycles
Only at the ends of
the heat exchangers
Del Nogal et al. (2008) Mixed refrigerants
and Cascade cycles Fully covered
- no details provided
SMR = Single Mixed Refrigerant
N/A = Not applicable
2.3.2 Optimisation of mixed refrigerant cycles
When optimisation is applied to the design or the operating variables of a refrigeration
cycle, commonly the objective function is to minimise the total shaft power for refrigerant
compression as it represents the major energy-consuming operation in the LNG plant and
dominates operating costs (Mokhatab et al., 2014b, Ch. 3.2). The conditions obtained as
optimal depend on the modelling assumptions as well as on the optimisation technique that
is selected. This section reviews the case studies from the design methodologies presented
in Section 2.3.1, and the optimisation methods selected in each case study are discussed.
Shelton and Grossmann (1986) presented a case study to design a refrigeration cycle that
operates between 240 K and 320 K. A minimum temperature approach (ΔTMIN) of 10 K is
considered in the evaporators and condensers. Also, is assumed a temperature level each
10 K for the refrigeration cycle. The MILP problem is solved with branch and bound
optimisation algorithm. With this method, different combinations of temperature levels are
evaluated according to the activated binary variables. For each combination of temperature
levels, the work coefficient, defined as the ratio of shaft work input to heat removed from
the process stream, is evaluated. The objective function is either to minimise the total
utility cost, the capital cost (only compressors are evaluated) or the total annualised cost.
The three objective functions are tested in the case study. Results showed that minimising
the capital cost resulted in the simplest refrigeration cycle configuration. When the
objective function is to minimise the total utility cost, the optimal refrigeration cycle yields
the highest capital costs because of the increased number of temperature levels (and
corresponding compressors, although the number is not explicit) compared to the optimal
design when minimising capital cost or total annualised cost. Minimising the total
annualised cost is a trade-off between utility and capital costs.
Chapter 2 Technology Background and Literature Review
43
As discussed by Shelton and Grossmann (1986), the main disadvantage of the MILP
problem formulation is the computational time needed according to the number of
temperature levels selected. On the one hand, if a large number of temperatures are
explored then a high computational effort is required to solve the problem. On the other
hand, a simpler problem formulation would not represent any computational inconvenience
but the temperature levels would not be thoroughly explored to find an optimum solution
to the problem.
For the modelling approach developed by Linnhoff and Dhole (1992), an example problem
is studied to design a refrigeration cycle integrated with the HEN in an ethylene production
process. Refrigeration is provided using propylene and ethylene at multiple temperature
levels. A base case is established using a single level of propylene (at 229 K) and three
levels of ethylene (at 212.5 K, 189 K and 173 K). The shaft power consumption is
calculated as 12.9 MW using modelling and simulation (models are not explicit in the
paper). Their approach is then used to calculate the shaft power input required for a
refrigeration cycle consisting of three levels of propylene (at 282 K, 257 K and 238 K) and
three levels of ethylene (at 209 K, 189 K and 173 K). Savings in shaft power are predicted
to be 3.83 MW. The proposed design is simulated (models are not explicit in the paper),
and the calculated shaft power savings are in good agreement (3.76 MW) as compared to
the predicted value using their approach.
However, since changes in the configuration of the refrigeration cycle are based on
judgement, the methodology to design the integrated refrigeration cycles is not systematic.
Thus, no optimisation technique is adopted by Linnhoff and Dhole (1992) in their work.
Three case studies are presented by Lee et al. (2002) for the design of a PRICO process
using their systematic design methodology for refrigeration cycles based on refrigerant
composition optimisation. The composition optimisation is formulated as an NLP problem.
In the first case study, the objective function of the optimisation is to minimise the
difference of the evaporating temperatures between the cold refrigerant stream and the
‘ideal’ cold composite curve (a duplicate of the hot composite curve but colder by a
minimum temperature difference, 5°C). The refrigerant can be composed of C1 – C4 and
nitrogen. The operating conditions are specified (flow rate, pressure levels and
composition), and the composition is optimised while the pressure levels and the
refrigerant flow rate are held constant. After the initial composition optimisation, either the
Chapter 2 Technology Background and Literature Review
44
pressure levels or the flow rate are decreased to reduce the compression shaft power
demand. Then, the composition optimisation is reinitiated to minimise possible
temperature crossovers between the composite curves, resulting from the modifications in
the refrigerant flow rate or pressure levels. This iterative process is continued until shaft
power savings are no longer possible. Results showed that shaft power savings of 21% are
achieved with the PRICO process designed with the presented methodology, compared to
the commercial PRICO process (1.49 MJ·kg-1
of LNG).
In the second case study, the composition optimisation is performed in two stages. In the
first stage, the refrigerant composition is optimised to minimise the sum of negative
differences between the evaporating temperatures of the ‘ideal’ cold composite curve and
those of the cold refrigerant stream, whilst in the second stage the composition is optimised
only to reduce the shaft power consumption. According to Lee et al. (2002), the second
stage begins when the refrigerant flow rate is decreased enough to avoid wetness at the
inlet of the compressor. The total shaft power is reduced 3.6% further in the second case
study, leading to a shaft power consumption of 1.13 MJ·kg-1
of LNG. In the third case
study, the effect of the ΔTMIN (from 3 K to 8 K) in the shaft power consumption is
investigated. As expected, the power demand is accordingly increased as the minimum
temperature approach is increased (from 1.14 to 1.32 MJ·kg-1
of LNG).
Although significant shaft power savings are achieved compared to the commercial PRICO
process, the degrees of freedom are not manipulated simultaneously during the
optimisation. Instead, the pressure levels, or the refrigerant flow rate, is adjusted (by
judgement or optimisation, although no details of the mathematical formulation are
provided for the optimisation) only after each composition optimisation. Thus, the degrees
of freedom are not fully exploited. Moreover, no criterion is explicitly discussed for the
manipulation of the flow rate and the pressure levels. Additionally, during the second stage
of the second case study, the optimisation is likely to lead to temperature crossover as the
ΔTMIN is no longer being considered.
Another shortcoming in the methodology followed by Lee et al. (2002) is that the results
obtained from an NLP optimisation strongly depend on the initial conditions selected
(Edgar et al., 2001, Ch. 10.1). Therefore, results presented by Lee et al. (2002) are likely to
be trapped in local optima. Optimisation approaches that helps avoiding local optima
Chapter 2 Technology Background and Literature Review
45
include multistart-point procedures and stochastic models, such as Genetic Algorithm and
Scatter Search (Edgar et al., 2001, Ch. 10).
The methodology for systematic design of cascade refrigeration cycles presented by
Vaidyaraman and Maranas (2002) is evaluated with an example problem to cool down a
methane-rich process stream from 20°C to –58°C. A minimum temperature difference of
2.5°C for feasible heat transfer is assumed. The optimisation is formulated as an NLP
problem, where the objective is to minimise the compression shaft power demand. The
degrees of freedom of the optimisation include the condensing and evaporating pressures,
the share of the process stream cooling duty in each MSHE, the vapour fraction of the
refrigerant after each partial condensation and the composition of the refrigerant. The
components of the refrigerant are ethane, propane and n-butane. The problem is solved for
various configurations: a cascade cycle of two refrigeration levels with different number of
refrigeration stages in each level. The results are expressed as the coefficient of
performance, COP, defined by Vaidyaraman and Maranas (2002) as the ratio of shaft work
input to heat rejected (thus, low COP values represent low values of shaft power demand
for a fixed heat load). The minimum COP is found with a configuration consisting in two
refrigeration stages in the upper cycle and three refrigeration stages in the lower cycle
(COP = 0.3957). According to Vaidyaraman and Maranas (2002), the optimum cascade
cycle design solution (COP = 0.3957) is a trade-off between a close temperature approach
in the composite curves (because of multiple streams with different compositions coming
from successive partial condensation) and an increasing flow rate, as the number of
refrigeration stages increases.
Unlike in the case studies presented by Lee et al. (2002), each configuration is optimised
by Vaidyaraman and Maranas (2002) from different starting points in order to avoid local
optima in the NLP problem. However, the number of optimisations per configuration is not
stated and the criteria for the selection of the different starting points are not discussed
either.
Del Nogal et al. (2008) selected two previously published case studies to illustrate their
proposed design methodology. Firstly, the case study presented by Lee et al. (2002) is
considered to design a PRICO process to liquefy a natural gas stream. The refrigerant
mixture is composed of C1 – C4 and nitrogen. Secondly, the case study of Vaidyaraman
and Maranas (2002) is studied for the design of cascade refrigeration cycles. Ethane,
Chapter 2 Technology Background and Literature Review
46
propane and n-butane are available to select the composition of the refrigerant. In both
cases, the degrees of freedom in the refrigeration cycles are manipulated simultaneously. A
Genetic Algorithm optimisation method is used to help avoid local optima.
In the Genetic Algorithm optimisation method, a user-specified number of initial
conditions (initial values of a vector that contains all the degrees of freedom), called
population, are generated randomly and thus, the optimisation is performed from different
starting points to help avoid local optima. That is, a set with a user-specified number of
vectors is generated (each vector is called member of the population). As the optimisation
progresses, one or more values of each member are changed (mutation) or replaced with
the value of the same variable from another member (crossover) (Edgar et al., 2001, Ch.
10.5), according to a probability factor ranging from 0 to 1. The new members (offspring)
are simulated and ranked according to the shaft power demand, in order to update the
population with those members with best performance (lowest shaft power consumption).
A penalty function is used to reject members that violate constraints (e.g. leading to
infeasible heat transfer inside the MSHE). An iteration (called generation) of the
optimisation is completed after all the members of the population are evaluated and the
population is updated. The optimisation is terminated when the maximum number of
generations (also user-defined) is reached. The member that yields the lowest specific
shaft power demand after the optimisation terminates, is considered as the optimum
solution.
Compared to deterministic optimisation methods (e.g. Sequential Quadratic Programming),
in which the result obtained from the optimisation depends on the initial conditions
selected, Genetic Algorithm explores the combination of the values of the variables more
thoroughly. Thus, Genetic Algorithm helps to avoid local optima, although the results
obtained with Genetic Algorithm are not guarantee to be optimal (Edgar et al., 2001, Ch.
10.1).
In the first case study presented by Del Nogal et al. (2008), the conditions obtained by Lee
et al. (2002) are specified and resulted in a ΔTMIN of 1.2°C between the composite curves.
Using this minimum temperature approach, to compare with the results found by Lee et al.
(2002), 8% in shaft work savings are achieved with the simultaneous optimisation of the
degrees of freedom. In the second case study, the same designs presented by Vaidyaraman
and Maranas (2002) are reproduced (considering the same minimum temperature approach
Chapter 2 Technology Background and Literature Review
47
of 2.5°C) and ΔTMIN violations are observed at intermediate temperatures inside the
MSHEs. The same refrigeration cycle designs are optimised with the Genetic Algorithm
method, and using Del Nogal’s modelling approach. The COP values increased up to 15%
for some of the configurations when compared to those in Vaidyaraman and Maranas
(2002). This time, however, the minimum temperature difference of 2.5°C between the
composite curves is guaranteed at intermediate temperatures inside the MSHEs.
In summary, optimisation of refrigeration cycles is commonly applied to minimise the total
shaft power for refrigerant compression (and therefore, to minimise operating costs).
Moreover, interactions between operating variables in the refrigeration cycles (e.g.
refrigerant composition, pressure levels, etc.) are exploited by simultaneous manipulation
of the degrees of freedom. Further, optimisation of the operating variables of refrigeration
cycles using stochastic methods (e.g. Genetic Algorithm) helps avoiding local optima
compared to NLP algorithms (e.g. Sequential Quadratic Programming).
2.3.3 Design of refrigeration cycles for small scale LNG processes
The research of refrigeration cycles for LNG processes at small scale, i.e. LNG production
up to 1 million t per annum according to Mokhatab et al. (2014b, Ch. 3.3), can be classified
in two groups: firstly, technology selection (especially between single mixed refrigerant
cycles and nitrogen expander cycles) based on performance indicators such as the
compression shaft power input; the operating conditions of the refrigeration cycles are not
optimised. Secondly, optimisation of the operating variables (e.g. refrigerant composition,
flow rate, etc.) of a refrigeration cycle to minimise the shaft power consumption is the aim
of research. The publications regarding technology selection are first reviewed.
Exergy is defined as a measure of the maximum work that can be obtained from a stream
when it reaches equilibrium with the surroundings (Querol et al., 2013, Ch. 2.1). Exergy
can be thus used to measure the minimum work input that needs to be supplied to a stream
to reach the specified conditions of temperature and pressure. Remeljej and Hoadley
(2006) published an exergy study of refrigeration cycles for LNG processes at small scale.
The LNG production rate is 0.700 million t per annum. A single mixed refrigerant cycle
(the PRICO cycle) and a nitrogen expander cycle are selected for the evaluation as well as
two open loop systems (“New LNG Scheme” and “GCL”) in which part of the feed gas is
used as the refrigerant mixture. The aim is to compare the exergetic ratio (i.e. the ratio of
actual work input to theoretical minimum work) of the refrigeration cycles for the same
Chapter 2 Technology Background and Literature Review
48
liquefaction task. The PRICO cycle showed an exergetic ratio of 2.8, which represented the
lowest shaft work consumption (0.3440 kWh·kg-1
of LNG) of the four refrigeration cycles.
The nitrogen expander cycle showed an exergetic ratio of 3.5 (shaft work is 0.4119
kWh·kg-1
of LNG), i.e. the shaft work is higher by 20% compared to the PRICO cycle.
Cao et al. (2006) evaluated a single mixed refrigerant cycle (with two refrigeration stages)
and a N2–CH4 expander cycle (with two refrigeration stages) for LNG production at small
scale (690 t per annum of LNG) by comparing the shaft power consumption of each
refrigeration cycle. Liquefaction is performed in two refrigeration stages (two heat
exchangers) and the natural gas stream is cooled down to a final temperature of –152°C.
The results indicated that the N2–CH4 cycle is about 24% less energy-consuming (1.3141
kWh·kg-1
of LNG) compared to the single mixed refrigerant cycle (1.7244 kWh·kg-1
of
LNG).
Yin et al. (2008) also compared the power consumption of a single mixed refrigerant cycle
and a nitrogen expander cycle. Both refrigeration cycles consisted of four stages of cooling
to liquefy the natural gas stream (down to –146°C). The LNG production rate is 0.006
million t per annum. The specific shaft work demand is estimated as 0.6715 kWh·kg-1
of
LNG for the nitrogen expander cycle and 0.3042 kWh·kg-1
of LNG for the single mixed
refrigerant cycle (i.e. 55% lower than that of the nitrogen expander cycle).
Li and Ju (2010) assessed the shaft power consumption of a nitrogen expander cycle, a
single mixed refrigerant cycle and a C3MR cycle for offshore LNG production according
to natural gas conditions (composition, working temperature and pressure) in South China
Sea. The LNG production rate is 0.061 million t per annum. The liquefaction is achieved
with four cooling stages in the C3MR cycle, three cooling stages in the single mixed
refrigerant cycle and three cooling stages in the nitrogen expander cycle. The nitrogen
cycle power consumption (0.5064 kWh·kg-1
of LNG) is 52% and 68% higher than that of
the single mixed refrigerant cycle (0.3330 kWh·kg-1
of LNG) and that of the C3MR cycle
(0.3013 kWh·kg-1
of LNG), respectively.
Castillo and Dorao (2010) proposed an economic model for refrigeration technology
selection in which the area for building the liquefaction plant is included in the capital
costs as installation costs. The LNG production rate is assumed as 0.500 million t per
annum. Mixed refrigerant technology (the specific mixed refrigerant cycle is not stated) is
compared against nitrogen expander technology (the specific cycle configuration is not
Chapter 2 Technology Background and Literature Review
49
specified). The shaft power consumption of the mixed refrigerant technology is assumed as
0.35 kWh·kg-1
of LNG and that of the nitrogen technology is assumed as 0.70 kWh·kg-1
of
LNG. The area needed for installation is estimated as a linear function of the LNG
production rate. An analysis for three different installation areas of the mixed refrigerant
technology (data not shown explicitly), showed that the installation area of the mixed
refrigerant significantly impacts on overall profit (including revenues, operating costs and
capital costs), as profit of the mixed refrigerant technology decreased up to 5 times
compared to that of the nitrogen expander. However, the authors suggested that more
accurate data should be used to correlate the installation area to the LNG production rate.
Shirazi and Mowla (2010) performed an optimisation of a PRICO cycle for small scale
LNG production (0.028 million t per annum). The case study addressed is similar to that
studied by Lee et al. (2002), and discussed in Section 2.3.1 and Section 2.3.2, but the
composition of the natural gas stream is different. Two-stage compression is assumed with
intercooling at 30 °C. The minimum temperature approach in the MSHE is 1.5°C. Genetic
Algorithm is adopted as the optimisation method to avoid local optima. The degrees of
freedom are manipulated simultaneously. The objective is to minimise the compression
shaft power. Shaft power savings of 3% (0.3034 kWh·kg-1
of LNG) are achieved compared
to the results published by Lee et al. (2002) (0.3130 kWh·kg-1
of LNG). The authors
concluded that the implementation of multistage compression with intercooling, compared
to single compression stage without intercooling in Lee (2001), led to the power savings.
He and Ju (2014b) presented a nitrogen expander cycle enhanced with a precooling stage
for LNG production at small scale (LNG production rate is 160 t per annum); either a
propane cycle or a cycle in which the refrigerant is a mixture of difluoromethane (CH2F2)
and pentafluoroethane (C2HF5), is used as the precooling stage. The refrigeration cycles,
including a conventional nitrogen expander cycle (as the base case), the nitrogen cycle
precooled with a propane cycle, and the nitrogen cycle precooled with the mixture of
difluoromethane and pentafluoroethane, are modelled in HYSYS. The NLP optimiser
within HYSYS is employed to optimise their corresponding operating variables.
Optimisations are performed for the conventional nitrogen cycle (base case) and for the
two precooled refrigeration cycles, to evaluate the precooling stage influence on the overall
shaft power demand (objective function). Compared to the base case, which consumes
0.6032 kWh·kg-1
of LNG, the propane precooling stage reduced the shaft power
consumption by 20% (0.4824 kWh·kg-1
of LNG) whilst the precooling stage with the
Chapter 2 Technology Background and Literature Review
50
mixture of difluoromethane and pentafluoroethane showed shaft power savings of 23%
(0.4660 kWh·kg-1
of LNG).
Hwang et al. (2013) considered the optimisation of a cascade of two mixed refrigerant
cycles (dual mixed refrigerant cycle) for the production of 380 t per annum of LNG. A
hybrid optimisation approach is used, consisting of a Genetic Algorithm method with SQP
(Sequential Quadratic Programming). For both mixed refrigerant cycles of the cascade, the
refrigerant flow rate, refrigerant composition, streams pressure level and refrigerant
precooling temperatures are optimised. The objective function is to minimise the shaft
work power for refrigerant compression. The optimum shaft power consumption of the
dual mixed refrigerant cycle is 0.2721 kWh·kg-1
of LNG.
He and Ju (2014a) presented a single mixed refrigerant cycle for LNG production at small
scale (0.009 million t per annum). The refrigeration cycle consists of two refrigeration
stages. Two-stage compression with intercooling is adopted. The refrigeration cycle is
modelled in HYSYS and optimised employing a Genetic Algorithm method. The
refrigerant flow rate, refrigerant composition, streams pressure levels and compressor
discharge pressure are the optimisation variables. The objective function is to minimise the
shaft power demand for refrigerant compression. The shaft power consumption of the
optimised refrigeration cycle is 0.3158 kWh·kg-1
of LNG.
From this review it is clear that single mixed refrigerant cycles are commonly studied for
LNG production at small scale. Table 2.4 summarises the mixed refrigerant cycles that are
being studied in the open research literature for production of LNG at small scale (i.e. up to
1 million t per annum), and also provides the efficiency reported for each mixed refrigerant
cycle (i.e. the specific shaft power consumption).
Table 2. 4. Mixed refrigerant cycles studied in the open research literature for LNG production at small scale.
Author(s) Mixed refrigerant
cycle
LNG production
[million t per annum]
Shaft power consumption
[kWh·kg-1
of LNG]
Remeljej and Hoadley (2006) PRICO cycle 0.7000 0.3440
Cao et al. (2006) 2 MSHE stages 0.0007 1.7284
Yin et al. (2008) 4 MSHE stages 0.0061 0.3042
Li and Ju (2010) 3 MSHE stages 0.0610 0.3330
Castillo and Dorao (2010) MR Technology* 0.5000 0.3500
Shirazi and Mowla (2010) PRICO cycle 0.0280 0.3035
He and Ju (2014a) 2 MSHE stages 0.0095 0.4175
*Not specified
Chapter 2 Technology Background and Literature Review
51
Moreover, from Table 2.4, is evident that limited refrigerant cycle configurations are being
studied in the open research literature for LNG production at small scale. The following
Section 2.3.4 describes the work of Zheng (2009), on the development of single mixed
refrigerant cycles for LNG processes, especially the “CryoMan” refrigeration cycle, which
is developed based on configurational modifications to the PRICO cycle.
2.3.4 Design and optimisation of the CryoMan process
The so-called “CryoMan” process is a single mixed refrigerant cycle for LNG processes.
The CryoMan process is developed by Zheng (2009) and later patented by the University
of Manchester (Kim and Zheng, 2011).
Zheng (2009) first presented a refrigeration cycle design, also resulting from structural
modifying the PRICO cycle, that consists on a flash unit after the condenser, so the vapour
and liquid phases of the partially condensed single mixed refrigerant are separated (this
refrigeration cycle is called “Pre-flash” design). The CryoMan process is developed by
further structurally modifying the “Pre-flash” design. Thus, the PRICO refrigeration cycle
is first explained, followed by the “Pre-flash” design and the CryoMan process.
In the PRICO cycle (Figure 2.12a), the mixed refrigerant is compressed to the compressor
discharge pressure and partially condensed using an ambient utility (e.g. cooling water).
The refrigerant stream is fed to the MSHE where is fully condensed. The refrigerant stream
is throttled to the low pressure level and returned to the MSHE where is vaporised to
provide cooling both to the natural gas stream and the high-pressure refrigerant stream.
The vaporised refrigerant stream is fed to the compressor where is recompressed to the
compressor discharge pressure, completing the refrigeration cycle. The composition of the
refrigerant stream strongly impacts on the compression shaft power demand of the
refrigeration cycle.
Regarding the “Pre-flash” design (Figure 2.12b), the structural modifications include a
flash unit attached after the partial condenser. Once the mixed refrigerant stream is
compressed to the compressor discharge pressure and partially condensed with the ambient
utility (e.g. cooling water), the refrigerant stream is fed to the flash unit where the vapour
and liquid phases are separated. The vapour and liquid streams have different composition.
Thus, two refrigerant streams with different composition are created from the single mixed
refrigerant stream. These two refrigerant streams are precooled in the MSHE to
Chapter 2 Technology Background and Literature Review
52
independent temperature levels, and are expanded to independent pressure levels using
throttle valves. The two refrigerant streams thus vaporise at different temperature levels to
provide the cooling duty needed in the MSHE for the liquefaction of the natural gas stream
and the self-cooling of the high-pressure refrigerant streams. Each refrigerant stream is fed
to the compressor at its corresponding pressure level.
In the CryoMan process (Figure 2.12c), similar to the “Pre-flash” design, the structural
modifications include a flash unit after the partial condenser; additionally, the product
streams from the flash unit are split and partially re-mixed to create the two actual
refrigerant streams. Compared to the “Pre-flash” design in which the composition and flow
rate of the refrigerant streams are those directly obtained from the flash separation, in the
CryoMan process, partially mixing of the vapour and liquid streams leaving the flash unit
allows manipulating the composition and flow rate of the two refrigerant streams. The two
refrigerant streams in the CryoMan process are also precooled to independent temperature
levels in the MSHE, throttled to independent pressure levels and vaporised at different
temperature levels (in the MSHE) to provide the cooling needed for liquefying the natural
gas stream and self-cooling the high-pressure refrigerant streams. The vaporised refrigerant
streams are fed to the compressor at their corresponding pressure levels. LP Stream is
compressed from the lowest pressure level; HP Stream is fed to the compressor at its
corresponding pressure level, where is mixed with LP Stream. The re-mixed refrigerant
stream is further compressed to the compressor discharge pressure.
Figure 2. 12. a) The PRICO cycle; b) “Pre-flash” design; c) the CryoMan process [adapted from (Zheng,
2009)].
Chapter 2 Technology Background and Literature Review
53
Thus, the degrees of freedom (i.e. the operating variables manipulated by the designer) in
the CryoMan process are the refrigerant composition, the compressor discharge pressure,
the flow rate fraction of the vapour stream leaving the flash unit fed to LP Stream, the flow
rate fraction of the liquid stream leaving the flash unit fed to LP Stream, the precooling
temperature, pressure level and MSHE outlet temperature of both LP Stream and HP
Stream.
In the model used to simulate the CryoMan process, the hot and cold composite curves in
the MSHE are calculated; enthalpy values are calculated for a predefined number of
intermediate temperatures, as the composition of the refrigerant streams and their
corresponding pressure levels are known (zero pressure drop in the MSHE is assumed).
Thus, feasible heat transfer at intermediate temperatures is checked by comparing a
predefined number of intermediate temperature differences between the hot and cold
composite curves against the ΔTMIN value (user-defined). Physical and thermodynamic
properties of the refrigerant streams (e.g. temperatures, enthalpies) are calculated with the
Peng–Robinson equation of state in Aspen Properties.
The composition of the two refrigerant streams, resulting from the partial mixing of liquid
and vapour streams, are calculated by mass balance of each component of the refrigerant
mixture. The overall refrigerant stream is composed of hydrocarbons C1 to C4 and
nitrogen.
Multistage compression model is adopted for calculating the shaft power demand of the
compressor; intercooling with an ambient heat sink (e.g. cooling water) after each
compression stage is assumed. Zero pressure drop of the refrigerant stream is assumed in
the intercoolers and the refrigerant stream is assumed to be cooled down to 30°C. The shaft
power demand for each compression stage is calculated with an energy balance over the
compressor, with the enthalpy values of the refrigerant stream at the inlet and outlet of
each compression stage under isentropic compression assumption. The enthalpy values are
calculated using Peng–Robinson equation of state, and an isentropic efficiency of 80% is
assumed.
To obtain the temperature of the overall refrigerant stream at the inlet of the intermediate
compression stage in which LP Stream and HP Stream are mixed, the enthalpy after
mixing is calculated with an energy balance over the mixing point. The enthalpy of HP
Stream is calculated at its MSHE outlet temperature (refrigerant composition and pressure
Chapter 2 Technology Background and Literature Review
54
level are known); the enthalpy of LP Stream is calculated at the intermediate pressure level
and the outlet temperature from the previous compression stage (refrigerant composition is
known).
Regarding the optimisation of the CryoMan process, the shaft power consumption is the
performance indicator. Therefore, the minimisation of the total compression shaft power
demand (the sum of the compression power demand of each compression stage) is defined
as the objective function. In order to optimise CryoMan process, i.e. to find the set of
values of its degrees of freedom that yield the minimum compression power input, a
Genetic Algorithm method is selected as the optimisation algorithm. The Genetic
Algorithm is implemented in WORK software (described in Section 2.3.5).
The case study first published by Lee et al. (2002) (see Section 2.3.2) is considered in
order to evaluate the performance (i.e. total shaft power consumption of the refrigeration
cycle) of the PRICO cycle, the “Pre-flash” design and the CryoMan process, when their
corresponding degrees of freedom are optimised to fully liquefy a natural gas stream.
According to Zheng (2009), the shaft power demand for the PRICO cycle is 28.27 MW;
the “Pre-flash” design achieved shaft power consumption of 26.60 MW, whereas the shaft
power demand in the CryoMan process is 26.05 MW. That is, the “Pre-flash” design and
the CryoMan process yielded shaft power savings of 6% and 8%, respectively, compared
to the PRICO cycle.
As discussed by Zheng (2009), the “Pre-flash” design benefits from an additional
refrigerant stream, compared to the PRICO cycle, created from the structural modification
applied (i.e. a flash unit attached). The additional refrigerant stream has independent
precooling temperature level and evaporating pressure level; the “Pre-flash” design also
benefits from refrigerant streams with different compositions that help reducing the total
shaft power demand. In the case of the CryoMan process, the structural changes
implemented in the configuration allows manipulating the flow rate and composition of the
refrigerant streams in order to minimise the shaft power consumption, by partially mixing
the vapour and liquid streams leaving the flash unit. Furthermore, the configuration of both
the “Pre-flash” design and the CryoMan process remained with relatively low complexity
(Zheng, 2009).
According to Zheng (2009), the benefits (shaft power savings compared to the PRICO
cycle) resulting from both the “Pre-flash” design and the CryoMan process is that the
Chapter 2 Technology Background and Literature Review
55
different composition of the refrigerant streams helped reducing the area between the
composite curves in the MSHE because of their corresponding evaporating profiles (but
keeping a minimum temperature difference between the composite curves for feasible heat
transfer). However, no numerical data of the area between the composite curves is
provided in order to compare and quantify the area reduction between the PRICO cycle,
the “Pre-flash” design and the CryoMan process. A comparison and analysis between the
composition of the refrigerant streams of the “Pre-flash” design and the CryoMan process
is not provided, in order to assess the impact of the refrigerant composition in the
performance of the refrigeration cycle. Additionally, although the configuration of both
refrigeration cycles (the “Pre-flash” design and the CryoMan process) is considered as a
low complex design, the ‘complexity’ of the refrigeration cycle design is not quantified or
clearly defined.
Thus, by analysing and taking advantage of key degrees of freedom in the configuration of
the CryoMan process, novel refrigeration cycles could be developed by further structurally
modifying the CryoMan process in order to bring further shaft power savings and hence,
operating cost savings. Moreover, because the novel refrigeration cycles developed from
structural modifications would be for LNG production at small scale, they should maintain
a design with low complexity in order to keep capital costs low. The ‘complexity’ of the
refrigeration cycle design should then be clearly defined.
2.3.5 WORK software
WORK software is in-house software of the Centre for Process Integration at the
University of Manchester. WORK software capabilities include modelling, simulation and
optimisation of refrigeration cycles, based on the research carried out in the Centre by Lee
(2001), Del Nogal (2006), and Zheng (2009).
Simple refrigeration cycles and cascade refrigeration cycles can be simulated in WORK
software. To check feasible heat transfer at intermediate temperatures inside the MSHEs,
the minimum temperature approach and the number of intermediate temperature
differences between the composite curves that are compared, are user-defined. Multistage
refrigerant compression with intercooling can also be simulated; the isentropic
compression efficiency, the maximum pressure ratio in a compression stage and the
temperature of the refrigerant stream after intercooling are user-supplied. Pure component
refrigerants as well as mixed refrigerants can also be simulated; physical and
Chapter 2 Technology Background and Literature Review
56
thermodynamic properties (e.g. temperatures, enthalpies) of the refrigerants can be
calculated through the use of coded equations of state (i.e. Peng–Robinson and Soave–
Redlich–Kwong) or, alternatively, by interfacing with other commercial software
packages, such as Aspen HYSYS or Aspen Properties.
Optimisation of operating variables of refrigeration cycles in WORK can be performed
using an NLP algorithm (Sequential Quadratic Programming) or with stochastic algorithms
(i.e. Genetic Algorithm and Simulated Annealing). Further, WORK is capable of
optimising the composition of a mixed refrigerant, given its constituent components. The
objective function can be the minimisation of the total shaft power consumption for
refrigerant compression.
The process stream (the stream that needs to be cooled or liquefied) is represented in
WORK using only its T–H profile. That is, composition, pressure levels and flow rate of
the process stream are not required. So, for example, the cooling/condensing profile (T–H
profile) of any process stream can be obtained from a simulation in Aspen HYSYS and
input directly in WORK software as the T–H profile of the process stream. However, a
process stream that is represented in WORK cannot be directly input in Aspen HYSYS as
the T–H profile only, because the full conditions of the process stream (composition, flow
rate, inlet and outlet pressures, and pressure drop profile in the heat exchanger) are
required in Aspen HYSYS.
However, despite the difference in the method for specifying the process stream that needs
to be cooled or liquefied, the results obtained from simulation of refrigeration cycles in
WORK software are in good agreement compared to those obtained from Aspen HYSYS
(see Section 2.3.6). Moreover, compared to Aspen HYSYS, WORK software is capable of
optimising the composition of mixed refrigerants. Furthermore, WORK can perform
stochastic optimisations (using either Genetic Algorithm or Simulated Annealing) to help
avoiding local optima.
2.3.6 Example of refrigeration cycle simulation in WORK software
In order to validate the simulations performed in WORK software (in the absence of
experimental data), an example problem is next presented and the results are compared to
those obtained from the simulation in Aspen HYSYS for the same example. Thus, the
results of Aspen HYSYS are assumed to represent real operational data accurately.
Chapter 2 Technology Background and Literature Review
57
The problem is taken from Morosuk et al. (2015), and consists of liquefying a natural gas
stream using a PRICO cycle. The flow rate of the natural gas stream is 50 kg·s-1
whilst that
of the mixed refrigerant stream is 475 kg·s-1
. The compositions of the natural gas and the
refrigerant mixture are displayed in Table 2.5. The natural gas stream enters the MSHE at
67 bar and 38°C and leaves as LNG at 64 bar and –159°C. The high-pressure mixed
refrigerant enters the MSHE at 22 bar and 30°C, and leaves at 19 bar and –159°C. The
precooled mixed refrigerant is then throttled to the low-pressure level (6 bar, –163°C). The
cold (low-pressure) refrigerant stream provides the cooling needed by the natural gas
stream and the high-pressure refrigerant stream, and leaves the MSHE at 3 bar. The
vaporised cold refrigerant is compressed to the compressor discharge pressure in two
compression stages; the intercooler and the condenser cool the refrigerant down to 30°C.
Figure 2.13 shows the schematic of the example problem.
Figure 2. 13. Operating conditions of the PRICO cycle for the simulation example problem.
According to the example problem, there is a pressure drop of 3 bar for each stream inside
the MSHE, but because the pressure drop profile is not stated, is assumed as linearly
dependent on the heat that each stream rejects or absorbs, in order to simulate the pressure
drop profile in Aspen HYSYS.
Physical and thermodynamic properties of the refrigerant (e.g. temperatures, enthalpies)
are calculated using Peng–Robinson. In WORK software, physical and thermodynamic
properties of the refrigerant stream are calculated using Peng–Robinson equation by
interfacing with Aspen HYSYS. It is assumed that the minimum temperature difference
between the composite curves in the MSHE is 3°C for feasible heat transfer; also,
isentropic efficiency for refrigerant compression is assumed as 80%.
Chapter 2 Technology Background and Literature Review
58
Table 2. 5. Natural gas and mixed refrigerant compositions [mole %] for the simulation example in WORK
software.
C1 C2 C3 C4 N2
Natural Gas 0.88 0.08 0.02 - 0.02
Mixed Refrigerant 0.30 0.30 - 0.25 0.15
Once the natural gas stream is simulated in Aspen HYSYS, its T–H profile is generated
and is fed to the simulation in WORK software (see Table 2.6 and Figure 2.14). Also,
because the flow rate of the mixed refrigerant has to be supplied to WORK on a molar
basis, the reported mass flow rate of the refrigerant is converted to molar flow rate
according to the composition in Table 2.5 and then fed to WORK software.
Table 2. 6. Natural gas temperature–enthalpy data for the simulation example in WORK software.
Segment Supply Temperature [K] Target Temperature [K] ΔH [kW] CP [kW·K-1
]
1.1 311.15 290.84 2701.3 133.0
1.2 290.84 270.53 2835.3 139.6
1.3 270.53 252.92 2691.0 152.8
1.4 252.92 237.91 2691.0 179.2
1.5 237.91 226.55 2691.0 236.8
1.6 226.55 219.12 2691.0 362.5
1.7 219.12 213.93 2691.0 518.3
1.8 213.93 208.74 2466.1 474.9
1.9 208.74 200.95 2723.1 349.5
1.10 200.95 190.70 2723.1 265.8
1.11 190.70 178.27 2723.1 219.1
1.12 178.27 164.05 2723.1 191.5
1.13 164.05 149.83 2483.8 174.6
1.14 149.83 133.55 2663.6 163.6
1.15 133.55 114.15 3022.0 155.8
Figure 2. 14. Natural gas temperature–enthalpy data for the simulation example in WORK software.
Chapter 2 Technology Background and Literature Review
59
Table 2.7 compares the results obtained with WORK against those obtained with Aspen
HYSYS; there is an error of less than 0.1% in the calculation of the shaft power demand
for compression. The minimum driving force for heat exchange (ΔTMIN) is different
compared to that in HYSYS by only 0.1°C. The calculations of the refrigeration cycle
performed in WORK are in good agreement compared to those obtained from the
simulation in Aspen HYSYS, and are thus considered successfully validated.
Table 2. 7. Comparison of the PRICO cycle simulation between WORK software and Aspen HYSYS.
WORK HYSYS Difference
Input Data
Refrigerant flow rate [kg·s-1
] 475a 475
Refrigerant high-pressure level [bar] 22 22
Refrigerant low-pressure level [bar] 6 6
Refrigerant precooling temperature [°C] –159 –159
Maximum compression ratio 3 3
Output Data
Number of compression stages 2 2 -
ΔTMIN [°C] 4.0 4.1 0.1°C
Shaft power [MW] 91.15 91.09 < 0.1 % aRefrigerant flow rate fed in WORK software as 14.6 kmol·s
-1
2.4 Conclusions
The literature review revealed that single mixed refrigerant cycles are commonly studied in
the open research for the production of LNG at small scale (up to 1 million t per annum).
Moreover, the literature review also revealed that there is only limited refrigeration cycle
configurations studied.
The CryoMan process (Zheng, 2009) is a single mixed refrigerant cycle that is developed
by structurally modifying the PRICO cycle (which is the simplest commercial refrigeration
cycles for LNG processes). Significant shaft power savings (nearly 8%) are achieved by
the CryoMan process compared to the PRICO cycle (which required 28.27 MW of shaft
power). Additionally, the design of the CryoMan process remained with relatively low
complexity.
Novel refrigeration cycles could be developed by analysing and further modifying the
structure of the CryoMan process, in order to bring shaft power savings and, therefore,
operating cost savings. The design of the novel refrigeration cycles should also remain
with low complexity to help keeping low capital costs, as they are intended for LNG
production at small scale. Thus, the ‘complexity’ of the refrigeration cycles has to be
clearly defined in order to develop novel designs with low complexity.
Chapter 2 Technology Background and Literature Review
60
Additionally, there are useful models in the open research literature for simulating mixed
refrigerant cycles. Because of the non-isothermal evaporation and condensation of mixed
refrigerants, feasible heat transfer needs to be checked inside the MSHE by calculating
intermediate temperatures of the composite curves. Introducing an isentropic efficiency for
refrigerant compression allows considering energy losses in compression (e.g. because of
friction in the compressor); modelling multistage compression with intercooling is useful
to represent commonly-practised industrial scenarios compared to single stage refrigerant
compression without intercooling.
Furthermore, optimisation of the operating variables of refrigeration cycles for LNG
processes, in the open research literature, is commonly aimed at minimising the total shaft
power demand for refrigerant compression since shaft power demand dominates operating
cost of refrigeration cycles. Simultaneous manipulation of the degrees of freedom in the
refrigeration cycles (e.g. refrigerant composition, pressure levels, etc.) allows exploiting
complex interactions between the operating variables. Stochastic optimisation methods
(e.g. Genetic Algorithm) helped avoiding local optima.
The accuracy of WORK software for the simulation of mixed refrigerant cycles, compared
to commercial software (i.e. Aspen HYSYS), was illustrated with an example problem.
The results obtained from WORK software were in good agreement compared to those
from the simulation in the commercial software. The calculation of the total shaft power
demand showed a relative difference of less than 0.1%, whereas the minimum temperature
approach was calculated with a difference of only 0.1°C between the two simulations.
Chapter 3 Development and Design of Novel Refrigeration Cycles
61
Chapter 3 – Development and Design of Novel Refrigeration
Cycles
3.1 Introduction
An important aim in the development of novel refrigeration cycles for small scale LNG
processes (i.e. LNG production is up to 1 million t per annum), is to keep the designs with
low complexity. Low complexity of the refrigeration cycles is especially important in LNG
processes at small scale, because of the capital costs associated with the design. For
instance, capital costs have to be low when small LNG projects are intended to operate for
short periods of time (Li and Ju, 2010) in order to keep profitability. Additionally, when
the refrigeration cycles are for offshore LNG production, the LNG plant would be built in a
platform or on top of a ship, and so the area or weight of the plant would be constrained
(Castillo and Dorao, 2010).
On the other hand, high energy-efficiency is important as operating costs dominates
economics in refrigeration cycles for LNG processes and shaft power consumption
represents the major energy-consuming stage of the refrigeration cycle (Mokhatab et al.,
2014b, Ch. 3.2). Large scale commercial refrigeration processes (e.g. propane precooled
mixed refrigerant cycle and dual mixed refrigerant cycle) yield high efficiencies by
introducing more equipment in order to increase the number of degrees of freedom in the
refrigeration cycle. The resulting operating variables are adjusted to reduce the overall
power consumption. However, this also results in large and complex configurations. For
example, the propane precooled mixed refrigerant cycle is reported to have a greater
efficiency (0.29 kWh·kg-1
of LNG) compared to the PRICO cycle (0.40 kWh·kg-1
of LNG)
(Castillo and Dorao, 2010); the multilevel propane precooling cycle reduces the heat load
of the natural gas stream in the mixed refrigerant cycle, and the operating variables of the
mixed refrigerant cycle (e.g. refrigerant composition, pressure levels, etc.) can be adjusted
to liquefy the natural gas stream with a lower overall energy consumption compared to the
PRICO cycle. However, the complexity and the capital costs of the propane precooled
mixed refrigerant cycle significantly increase since four heat exchangers and the multistage
compressor are needed for the multilevel propane precooling stage, which is in addition to
the mixed refrigerant cycle that includes the multi-stream heat exchanger (MSHE) as well
as its own compressor.
Chapter 3 Development and Design of Novel Refrigeration Cycles
62
Therefore, an economic trade-off exists between a refrigeration cycle with high energy-
efficiency (low operating costs) and a cycle with a low complexity configuration (low
capital investment). In order to limit of the complexity of a refrigeration cycle, equipment
design constraints are defined in this work for the design of the novel refrigeration cycles.
Compressors represent the most capital-intensive equipment in the refrigeration cycle and
also the most energy-consuming (Mokhatab et al., 2014b, Ch. 3.2). Centrifugal
compressors are commonly used in the LNG industry (Hanlon, 2001, Ch. 3.11). These
compressors can be designed with multiple compression stages in a single casing (usually
up to 8 compression stages) as well as with nozzles at intermediate pressures, for example,
to send the partially compressed refrigerant to an intercooler or to admit side streams at
intermediate pressure levels (see Figure 3.1) (Ludwig, 2001, Ch. 12). Each compression
stage is usually designed for a maximum pressure ratio of 3 (Hanlon, 2001, Ch. 3.5). Thus,
in this work, as a limit for complexity, only one centrifugal multistage compressor is
considered, and is constrained to a single casing with up to 8 compression stages, in which
the maximum pressure ratio of each individual compression stage is 3.
Figure 3. 1. Multistage centrifugal compressor [adapted from (Ludwig, 2001, Ch. 12)].
Regarding the main heat exchanger, plate-fin heat exchangers are one type of MSHEs that
are commonly used in the LNG industry. The plate-fin heat exchangers can be designed to
accommodate up to 12 streams in a single piece of equipment (Hesselgreaves, 2001, Ch.
2), which is defined in this work as the design constraint for the limit for complexity in the
MSHE. Figure 3.2 shows an example of an MSHE in which 5 streams are simultaneously
involved in heat exchange.
Chapter 3 Development and Design of Novel Refrigeration Cycles
63
Figure 3. 2. Multi-stream heat exchanger for five streams [reproduced from (ESDU International plc, 2006)].
Thus, a single casing centrifugal compressor with a maximum of 8 compression stages,
and a maximum pressure ratio of 3 in each compression stage, plus a single MSHE with a
maximum of 12 streams, are the criteria defined as the boundary for complexity. In this
thesis, beyond this boundary, any refrigeration cycle is considered as a complex design.
The design constraints defined in this thesis as the limits for complexity would help
keeping the capital costs of the novel refrigeration cycles relatively low.
Capital cost estimation is difficult because of scarce data available in the open literature.
For instance, even though the methodology presented in (ESDU International plc, 2006)
allows obtaining a general estimate of the cost of a MSHE, based on the volume of the unit
and the number of streams involved in heat exchange, cost correlations are limited to
MSHE units with a maximum of six streams. Additionally, correlation models for
equipment cost estimation based on energy consumption (e.g. for compressors) do not
always reflect the complexity of the equipment. For example, the shaft power consumption
– and hence cost – of a multistage compressor would be reduced as the number of
compression stages (with their corresponding intercooling stages) is increased; however,
the complexity of the multistage compressor design would be increased and its cost would
be expected to increase accordingly. Capital costs estimation is thus not considered in this
thesis.
3.1.1 Benchmark processes
As mentioned in Section 2.3.3, single mixed refrigerant cycles are commonly considered
for LNG processes at small scale. Amongst the single mixed refrigerant cycles, the PRICO
cycle is the leading refrigeration technology for small and medium scale liquefied natural
Chapter 3 Development and Design of Novel Refrigeration Cycles
64
gas production (Mokhatab et al., 2014b, Ch. 3.3). Also, the CryoMan process is used as the
starting point for the development of novel refrigeration cycles. Therefore, in order to
evaluate the performance of the novel refrigeration cycles, the results obtained (i.e. shaft
power consumption, operating costs) are compared against these two benchmark processes,
i.e. the PRICO cycle and the CryoMan process.
3.1.2 Performance evaluation
The shaft power requirement for refrigerant compression of each novel design is firstly
evaluated (as a preliminary assessment) through sensitivity analyses, and compared to the
shaft power demanded in the CryoMan process. Simulation and sensitivity studies are
performed on each degree of freedom (e.g. refrigerant composition, pressure levels, etc.) of
the novel designs using Aspen HYSYS v8.2 (Aspen Technology Inc., 2013). The novel
refrigeration cycles that show shaft power savings in the preliminary assessment, compared
to the CryoMan process, are optimised in WORK software by means of a Genetic
Algorithm method. The results obtained from the optimisation are compared and evaluated
against the benchmark processes.
3.1.3 Operating cost estimation
As mentioned in Section 3.1, refrigerant compression is the most energy-consuming stage
in the refrigeration cycle and thus, is used to estimate the operating costs of the novel
refrigeration cycles. The operating costs are calculated on an annual basis and the unit
power energy cost is assumed as £0.0955·kWh-1
for extra-large scale industrial consumers
(U.K. Department of Energy & Climate Change, 2015). Thus, the operating costs of the
novel refrigeration cycles are estimated from the total shaft power demand by the
compressor annually and the unit power energy cost.
3.2 Further development of the CryoMan process
As mentioned in Section 3.1.1, the CryoMan process is used as the starting point design for
developing novel refrigeration cycles based on structural modifications. In order to make
meaningful structural modifications, the CryoMan process is first analysed in detail
(Section 3.2.1) to identify and take advantage of key degrees of freedom in its design. As a
result of the analysis, possible structural modifications that can bring shaft power savings
in the liquefaction process are then presented in Section 3.2.2. Finally, in Section 3.2.3, the
configuration modifications are proposed and the novel refrigeration cycles are designed.
Chapter 3 Development and Design of Novel Refrigeration Cycles
65
3.2.1 Analysis of the CryoMan process
Zheng (2009) presented two refrigeration cycle designs, resulting from structural
modifications to the PRICO cycle, in which two independent refrigerant streams are
obtained, i.e. their corresponding inlet and outlet temperatures of the MSHE, pressure
levels and composition are different from each other (see Figure 3.3). One of those
refrigeration cycles consists of a flash unit after the partial condenser (named “Pre-flash”
design; see Figure 3.3a). Thus, the vapour and liquid phases (LP Stream and HP Stream in
Figure 3.3a, respectively) are generated and used as the refrigerant streams. The two
streams have different composition. Nearly 6% in shaft power savings are achieved with
this configuration (0.3075 kWh·kg-1
of LNG), compared to the PRICO cycle (0.3268
kWh·kg-1
of LNG), for the natural gas liquefaction case study first presented by Lee (2001)
(see Section 2.3.4).
In the CryoMan process (Figure 3.3b), the vapour and liquid phases from the flash unit are
partially mixed (isobaric mixing is assumed), which allows manipulating the composition
of the actual refrigerant streams (LP Stream and HP Stream). The CryoMan process
showed shaft power consumption of 0.3011 kWh·kg-1
of LNG, i.e. savings in shaft power
demand of nearly 8% compared to the PRICO cycle.
Figure 3. 3. Refrigeration cycles presented by Zheng (2009): a) “Pre-flash” design, and b) the CryoMan
process.
To illustrate how the composition of the refrigerant streams (LP Stream and HP Stream)
changes when the overall refrigerant stream is flashed and partially mixed, consider, as an
example, the flash separation of a ternary refrigerant (mixture of 50% ethane, 30% propane
and 20% n-butane) at 20 bar and 30°C. The resulting compositions are shown in Figure
3.4a. The square represents the composition of the liquid phase resulting from the flash
unit, whereas the diamond is the composition of the vapour phase, and the triangle stands
Chapter 3 Development and Design of Novel Refrigeration Cycles
66
for the composition of the refrigerant that is being flashed. The circles represent the
compositions obtained when the vapour and liquid phases are partially mixed at different
proportions (proportions are shown in Figure 3.4b as Case 1 to Case 7). As the proportions
of vapour and liquid partially mixed increase, the resulting composition moves towards the
triangle (the overall refrigerant composition) because the overall refrigerant is obtained
when all the vapour and liquid is mixed.
Figure 3.4b shows the shaft power demand for compression of 1 kmol·s-1
of the refrigerant
mixtures in Figure 3.4a from 1.2 bar and 30°C to 20 bar. A refrigerant flow rate on a molar
basis (1 kmol·s-1
assumed) would allow comparing the compression shaft power demand
of streams with the same number of molecules, and so changes in shaft power can be
related to the differences in composition (i.e. different proportions of the components) of
the streams.
According to Figure 3.4b, as the composition moves towards that of the vapour phase, the
shaft power required for compression increases. The vapour phase is highly composed of
the ‘light components’ in the refrigerant mixture, i.e. molecules with relatively small size
(72% of the vapour stream is ethane compared to only 6% of n-butane). In the liquid
stream, on the other hand, 41% is ethane compared to 26% of n-butane. More power is
required to compress and increase the pressure of a refrigerant stream as the mole fraction
of the ‘light components’ increases. Thus, the compression power required is sensitive to
the composition of the refrigerant mixture and increases as the mole fraction of ethane (the
smallest molecule in the mixture compared to propane and n-butane) increases. This
ternary mixture example can be extended to a refrigerant with five or more components.
Figure 3. 4. Ternary refrigerant after flash separation: a) composition distribution; and b) shaft work for
compression of 1 kmol·s-1
from 1.2 bar and 30°C to 20 bar.
Chapter 3 Development and Design of Novel Refrigeration Cycles
67
In the “Pre-flash” design (Figure 3.3a), the streams leaving the flash unit are used as the
refrigerant streams. Zheng’s results suggested that different compositions in the two
refrigerant streams bring benefits since shaft power savings of 6% are achieved in the
refrigeration cycle, compared to the PRICO cycle. In the CryoMan process, the flash unit
product streams are partially mixed. This partial mixing of vapour and liquid phases
yielded a further 2% savings in shaft power, i.e. a total of 8% savings in shaft power
demand compared to the PRICO cycle.
Thus, according to the example illustrated with Figure 3.4, partial mixing of the vapour
and liquid phases helps reducing the mole fraction of light components (i.e. small
molecules) in LP Stream in the CryoMan process, resulting in a refrigerant stream with a
composition that requires less power for compression compared to that of LP Stream in the
“Pre-flash” design.
The compositions of the streams obtained with the “Pre-flash” design and with the
CryoMan process are compared (see Figure 3.5) in terms of the shaft power required for
compression of 1 kmol·s-1
of refrigerant in LP Stream. The shaft work is calculated for
different outlet pressures, at a constant inlet pressure (1.2 bar, which is the pressure level of
LP Stream in both designs), to create a compression power trend for the “Pre-flash” design
(diamonds) and for the CryoMan process (squares). As can be seen from Figure 3.5, the
partial mixing of streams leaving the flash unit yielded a composition that requires less
power, per unit flow rate, to be compressed at the same outlet pressure, compared to LP
Stream in the “Pre-flash” design. This is in agreement with the previous example of the
ternary mixture in which as the composition moves away from that of the flash unit vapour
stream (i.e. as the mole fraction of the light components decreases), the compression power
decreases.
Chapter 3 Development and Design of Novel Refrigeration Cycles
68
Figure 3. 5. Shaft work demand of LP Stream in the “Pre-flash” and in the CryoMan process for different
outlet pressures at constant inlet pressure (1.2 bar).
However, if the actual flow rates of LP Stream and HP Stream are considered, then the
compression power trend of LP Stream in the CryoMan process is higher compared to that
in the “Pre-flash” design (see Figure 3.6a). This is also true for the compression of the
overall refrigerant from their corresponding intermediate pressure level to the compressor
discharge pressure (48.3 bar, which is the same in both designs) (Figure 3.6b). Since the
CryoMan process achieved greater shaft power savings than those achieved with the “Pre-
flash” design, the increased compression power trends of the CryoMan process thus
suggests that a trade-off exists between the shaft power savings achieved with intermediate
compositions compared to those obtained from simple flash separation, and the increase in
flow rate.
LP Stream in the CryoMan process is compressed to 9.6 bar, whereas in the “Pre-flash”
design the corresponding stream is compressed to 6.9 bar. This leads to a total power
consumption of 3 MW higher than that in the “Pre-flash” design. Nevertheless, the
compression of the overall refrigerant stream (when LP Stream and HP Stream are mixed
in the compressor) yields total shaft power savings of 3.6 MW in the CryoMan process,
compared to the “Pre-flash” design, since the compression starts from a higher inlet
pressure (9.6 bar) compared to that in the “Pre-flash” design (6.9 bar) . Therefore, the
overall resulting shaft power demand is reduced by 0.6 MW in the CryoMan process
compared to the “Pre-flash” design.
7
8
9
10
8 10 12 14 16
Sh
aft
Po
wer
[M
J·k
mo
l-1]
Outlet Pressure [bar]
PreFlash = 1 kmol/s
CryoMan = 1 kmol/s
Chapter 3 Development and Design of Novel Refrigeration Cycles
69
Figure 3. 6. Compression trends with actual flow rates (Zheng, 2009): a) LP Streams at constant inlet
pressure (1.2 bar); and b) overall refrigerant streams at constant outlet pressure (48.3 bar).
Thus, the partial mixing of vapour and liquid not only benefits from having intermediate
compositions with less compression power requirement per unit flow rate compared to the
vapour stream obtained from the flash unit, but also exploits a trade-off existing between
the flow rate of each stream and the compression level at which each stream is compressed.
Therefore, the design of the novel refrigeration cycles should be aimed at exploiting these
trade-offs in complex interactions. Creating different refrigerant streams can result in
compositions that require less power for compression, compared to only flash product
streams. This new compositions show a trade-off with the flow rate of each stream. On the
other hand, there is also a trade-off between the flow rate of each refrigerant stream and the
pressure levels at which the stream is expanded and compressed.
3.2.2 Generation of structural options
Different modifications of the structure of the refrigeration cycle can be performed to
exploit the trade-offs identified and discussed in Section 3.2.1, in order to reduce the shaft
power demand. For example, since refrigerant streams with different compositions can
bring shaft power savings, one option is to introduce a bypass stream before the flash unit
(see Figure 3.7). This stream would have an intermediate composition (the overall
composition) compared to those obtained with the flash unit, with or without partial
mixing. The trade-off between stream flow rate and pressure level could also be exploited
since the bypass stream would help reducing the flow rate of the stream at the lowest
pressure level (LP Stream).
6.9 bar
9.6 bar
8
9
10
11
12
13
14
15
5 7 9 11 13
Sh
aft
Po
wer
[M
W]
Outlet Pressure [bar]
PreFlash = 1.64 kmol/sCryoMan = 1.79 kmol/s
Power
Savings:
-3 MW
Inlet pressure = 1.2
6.9 bar
9.6 bar
10
12
14
16
18
20
22
6 8 10 12
Sh
aft
Po
wer
[M
W]
Inlet Pressure [bar]
PreFlash = 3.08 kmol/sCryoMan = 3.21 kmol/s
Power
Savings:
+3.6 MW
Outlet pressure = 48.3
Chapter 3 Development and Design of Novel Refrigeration Cycles
70
Figure 3. 7. A refrigerant stream bypassing the flash unit.
Another modification to create intermediate compositions is to flash repeatedly the main
refrigerant stream to obtain multiple streams with different compositions. Each flash
separation would require either an additional partial condensation of the vapour from the
previous flash unit or an expansion of the liquid stream prior to entering the MSHE (see
Figure 3.8). Furthermore, if each of the multiple streams generated has an independent
pressure level, the trade-off between the streams flow rate and their corresponding pressure
level could be exploited as the flow rate in the low pressure compression stages might be
reduced.
Figure 3. 8. Multiple flash separation of the refrigerant stream (liquid from first flash unit is further
expanded).
The above structural modification concepts (a bypass stream and multiple flash
separations) focus on modifying the refrigeration cycle before the refrigerant enters the
MSHE. Another option is partial mixing the refrigerant streams after precooled in the
MSHE and expanded to their corresponding pressure levels. That is, a flow rate fraction α
from LP Stream can be mixed with a flow rate fraction β from HP Stream to create a new
stream with intermediate composition (see Figure 3.9). Because the partial mixing of the
refrigerant streams is after expansion, the new refrigerant stream would be constrained to
have the same pressure level of the stream at the lowest pressure level. However, by
creating a new refrigerant stream with intermediate composition, the trade-off between
refrigerant flow rate and composition would be exploited in order to help reducing the
overall refrigerant flow rate and potentially bring shaft power savings.
Chapter 3 Development and Design of Novel Refrigeration Cycles
71
Figure 3. 9. Partial mixing of refrigerant stream after precooling in the MSHE.
3.2.3 Proposal of novel designs
Based on the conceptual modifications that can be implemented in the CryoMan process,
as discussed in Section 3.2.2, the configuration of the novel refrigeration cycles can be
then designed.
Firstly, the Bypass design is shown in Figure 3.10. In this design, before the main
refrigerant stream is flashed into vapour and liquid, a portion of the stream bypasses the
flash unit. This additional third stream has an intermediate composition compared to those
obtained from the flash unit. Moreover, the bypass stream has its own pressure level.
Figure 3. 10. Novel refrigeration cycle 1: Bypass design.
To create several streams with different compositions, a refrigeration cycle design with a
double flash separation is introduced (Two Flash Levels design; see Figure 3.11). The
liquid phase from the first flash unit is divided in three streams: two are partially mixed
with the vapour from the same flash unit (as in the conventional CryoMan process),
whereas the remainder is expanded to a second pressure level and further flashed in a
second flash unit. The resulting vapour and liquid phases from the second flash unit are
Chapter 3 Development and Design of Novel Refrigeration Cycles
72
allowed to be partially mixed with each other. In addition, each stream has its own
independent pressure level.
Figure 3. 11. Novel refrigeration cycle 2: Two Flash Levels design.
Finally, in a third refrigeration cycle – Mixing After Precooling design in Figure 3.12 –, a
new cold stream is created from the streams partially mixed after the flash unit. However,
the new stream is created after the two partially mixed streams have been precooled in the
MSHE and expanded to their corresponding pressure levels. According to Figure 3.12, a
flow rate fraction α from Stream 1 is mixed with a flow rate fraction β from Stream 2, to
create Stream 5. The resulting new stream has an intermediate composition and would take
advantage of the trade-off between composition and refrigerant flow rate to potentially
bring shaft power savings. Also, as discussed in Section 3.2.2, the new refrigerant stream
(Stream 5) would be constrained to have the same pressure level as Stream 4, i.e. the
lowest pressure level.
Chapter 3 Development and Design of Novel Refrigeration Cycles
73
Figure 3. 12. Novel refrigeration cycle 3: Mixing After Precooling design.
3.3 Process modelling
In order to simulate and evaluate the performance of each novel refrigeration cycle,
process modelling is required. The Bypass design model is described in Section 3.3.1,
whereas the Two Flash Levels design is in Section 3.3.2 and the Mixing After Precooling
design model is in Section 3.3.3. In all three cases, the starting point of the refrigeration
cycle simulation is at the outlet of the condenser; it is firstly assumed a refrigerant flow
rate of 1 kmol·s-1
.
The models of the novel refrigeration cycles are implemented in WORK software. Physical
and thermodynamic properties of the refrigerant (such as temperatures, enthalpies, phase
equilibrium, etc.) are calculated using the Peng–Robinson equation of state by interfacing
with Aspen HYSYS v8.2. These calculations are expressed in this thesis as ‘Relations’,
using the notation of Example Relation E3.1, in which the specific enthalpy of a refrigerant
stream is calculated at temperature level 1 (TLevel 1) and pressure level 2 (PLevel 2), provided
that the refrigerant composition (XRef
) is known.
ℎ𝑅𝑒𝑓 → 𝑋𝑅𝑒𝑓 , 𝑇𝐿𝑒𝑣𝑒𝑙 1, 𝑃𝐿𝑒𝑣𝑒𝑙 2 (E3.1)
3.3.1 Modelling of the Bypass design
Figure 3.13 shows the Bypass design refrigeration cycle with its degrees of freedom (i.e.
the variables that are independently manipulated by the designer). Once a set of values for
the degrees of freedom are specified, the molar flow rate of the bypass stream is obtained
with Equation 3.1, where FMR
is the molar flow rate of the overall refrigerant stream, and α
is the flow rate fraction that bypasses the flash unit.
Chapter 3 Development and Design of Novel Refrigeration Cycles
74
𝐹𝐵𝑦𝑝𝑎𝑠𝑠 = 𝛼 ∙ 𝐹𝑀𝑅 (3.1)
where F Bypass
= Molar flow rate of Bypass Stream
F MR
= Molar flow rate of the overall refrigerant stream
α = Flow rate fraction of the overall refrigerant stream that bypasses the flash unit
Figure 3. 13. Bypass design: degrees of freedom.
The vapour fraction of the mixed refrigerant (with known composition, XMR
) is obtained
with Relation 3.2 at the condenser temperature and compressor discharge pressure, which
are user-specified. The remainder of the refrigerant is flashed and the composition of each
of the phases is calculated using Relation 3.3 and Relation 3.4. The flow rate of each flash
product stream is calculated with Equation 3.5 and Equation 3.6.
𝑉𝐹𝑀𝑅 → 𝑋𝑀𝑅 , 𝑇𝐶𝑜𝑛𝑑, 𝑃𝐻𝑖𝑔ℎ (3.2)
𝑋𝑉𝑎𝑝 → 𝑋𝑀𝑅 , 𝑇𝐶𝑜𝑛𝑑, 𝑃𝐻𝑖𝑔ℎ (3.3)
𝑋𝐿𝑖𝑞 → 𝑋𝑀𝑅 , 𝑇𝐶𝑜𝑛𝑑, 𝑃𝐻𝑖𝑔ℎ (3.4)
𝐹𝑉𝑎𝑝 = 𝑉𝐹𝑀𝑅 ∙ (𝐹𝑀𝑅 − 𝐹𝐵𝑦𝑝𝑎𝑠𝑠) (3.5)
𝐹𝐿𝑖𝑞 = (1 − 𝑉𝐹𝑀𝑅) ∙ (𝐹𝑀𝑅 − 𝐹𝐵𝑦𝑝𝑎𝑠𝑠) (3.6)
where VF MR
= Vapour fraction of the overall refrigerant stream
X MR
= Vector for the composition of the overall refrigerant stream
X Vap,Liq
= Vector for the composition of the vapour and liquid streams leaving the flash unit
TCond = Temperature of the refrigerant stream after the condenser
PHigh = Compressor discharge pressure
F Vap
= Molar flow rate of the vapour stream leaving the flash unit
F Liq
= Molar flow rate of the liquid stream leaving the flash unit
Chapter 3 Development and Design of Novel Refrigeration Cycles
75
After the partial mixing of vapour and liquid (isobaric mixing is assumed), the composition
of Stream 1 and Stream 2 is calculated with a mass balance for each component (xi)
according to the flow rate fraction of vapour (f Vap
) and liquid (f Liq
) to be mixed in each
stream (Equations 3.7 and 3.8). f Vap
is the flow rate fraction of the vapour stream leaving
the flash unit, that is mixed with f Liq
to create Stream 1; f Liq
is the flow rate fraction of the
liquid stream leaving the flash unit, that is mixed with f Vap
to create Stream 1. The flow
rate of Stream 1 and Stream 2 are obtained with Equations 3.9 and 3.10. The composition
of Bypass Stream is the same as the overall refrigerant stream composition.
𝑥𝑖𝑆1 = [
(𝑉𝐹𝑀𝑅∙𝑓𝑉𝑎𝑝∙𝑥𝑖𝑉𝑎𝑝
) + ((1−𝑉𝐹𝑀𝑅)∙𝑓𝐿𝑖𝑞∙𝑥𝑖𝐿𝑖𝑞
)
(𝑉𝐹𝑀𝑅∙𝑓𝑉𝑎𝑝) + ((1−𝑉𝐹𝑀𝑅)∙𝑓𝐿𝑖𝑞)]; 𝑥𝑖
𝑆1 ∈ 𝑋𝑆1 (3.7)
𝑥𝑖𝑆2 = [
(𝑉𝐹𝑀𝑅∙(1−𝑓𝑉𝑎𝑝)∙𝑥𝑖𝑉𝑎𝑝
) + ((1−𝑉𝐹𝑀𝑅)∙(1−𝑓𝐿𝑖𝑞)∙𝑥𝑖𝐿𝑖𝑞
)
(𝑉𝐹𝑀𝑅∙(1−𝑓𝑉𝑎𝑝)) + ((1−𝑉𝐹𝑀𝑅)∙(1−𝑓𝐿𝑖𝑞))]; 𝑥𝑖
𝑆2 ∈ 𝑋𝑆2 (3.8)
𝐹𝑆1 = [(𝐹𝑉𝑎𝑝 ∙ 𝑓𝑉𝑎𝑝) + (𝐹𝐿𝑖𝑞 ∙ 𝑓𝐿𝑖𝑞)] (3.9)
𝐹𝑆2 = [(𝐹𝑉𝑎𝑝 ∙ (1 − 𝑓𝑉𝑎𝑝)) + (𝐹𝐿𝑖𝑞 ∙ (1 − 𝑓𝐿𝑖𝑞))] (3.10)
where xiS1,S2
= Mole fraction of the ith component in Stream 1 and Stream 2
xiVap,Liq
= Mole fraction of the ith component in the vapour and liquid streams from the flash unit
X S1,S2
= Vector for the composition of Stream 1 and Stream 2
f Vap
= Flow rate fraction of the vapour leaving the flash unit, mixed with f Liq
to create Stream 1
f Liq
= Flow rate fraction of the liquid leaving the flash unit, mixed with f Vap
to create Stream 1
F S1,S2
= Molar flow rate of Stream 1 and Stream 2
S1,S2 = Stream 1, Stream 2
All the three hot streams are precooled to their user-specified temperature and expanded to
their corresponding pressure level. The specific enthalpy after the precooling of each hot
stream (hAP) is thus obtained with Relation 3.11 according to their corresponding
composition. The superscripts ‘Ref’, ‘Ref1’, ‘Ref2’ and ‘Ref3’ are employed in this section
to refer to generic refrigerant streams. The temperature after the expansion of each stream
is determined with Relation 3.12 under the assumption of isenthalpic expansion (hAP is
therefore employed for calculating the temperature after expansion). The specific enthalpy
of the cold streams at the outlet of the MSHE is then calculated using Relation 3.13
according to their user-specified outlet temperatures. The amount of heat transferred by
each hot stream and absorbed by each cold stream in the MSHE is then determined with
Equations 3.14 and 3.15, respectively (zero pressure drop is assumed). The overall
refrigerant flow rate is then obtained from an energy balance around the MSHE according
Chapter 3 Development and Design of Novel Refrigeration Cycles
76
to Equation 3.16, considering the heat load of the natural gas to be liquefied, which is user-
specified.
ℎ𝐴𝑃 → 𝑋𝑅𝑒𝑓 , 𝑇𝐴𝑃, 𝑃𝐻𝑖𝑔ℎ (3.11)
𝑇𝐴𝐸 → 𝑋𝑅𝑒𝑓 , 𝑃𝐴𝐸 , ℎ𝐴𝑃 (3.12)
ℎ𝑜𝑢𝑡 → 𝑋𝑅𝑒𝑓, 𝑇𝑜𝑢𝑡, 𝑃𝐴𝐸 (3.13)
∆ℎℎ𝑜𝑡 = (ℎ𝑖𝑛 − ℎ𝐴𝑃) (3.14)
∆ℎ𝑐𝑜𝑙𝑑 = (ℎ𝑜𝑢𝑡 − ℎ𝐴𝑃) (3.15)
𝐹𝑀𝑅 = [𝑄𝑁𝐺
(∆ℎ𝑐𝑜𝑙𝑑𝑆1 +∆ℎ𝑐𝑜𝑙𝑑
𝑆2 +∆ℎ𝑐𝑜𝑙𝑑𝐵𝑦𝑝𝑎𝑠𝑠
) − (∆ℎℎ𝑜𝑡𝑆1 +∆ℎℎ𝑜𝑡
𝑆2 +∆ℎℎ𝑜𝑡𝐵𝑦𝑝𝑎𝑠𝑠
)] (3.16)
where hAP = Specific enthalpy of a hot refrigerant stream after precooling in the MSHE
hin = Specific enthalpy of a hot refrigerant stream prior to entering the MSHE
TAP = Temperature of a hot refrigerant stream after precooling in the MSHE
TAE = Temperature of a cold refrigerant stream after expanded with the throttle valve
PAE = Pressure level of a cold refrigerant stream after expended with the throttle valve
Tout = Temperature of a cold refrigerant stream at the outlet of the MSHE
hout = Specific enthalpy of a cold refrigerant stream at the outlet of the MSHE
QNG
= Heat load of the natural gas that is liquefied
Heat transfer feasibility is then checked inside the MSHE based on a minimum
temperature difference between the composite curves. The T–H profile for each refrigerant
stream is first generated. A user-specified number of intermediate temperatures is selected,
and these intermediate temperatures are evenly distributed in the range between the inlet
and outlet temperatures of each stream, according to Equation 3.17. The corresponding
enthalpy is calculated for each intermediate temperature, using Relation 3.18, where the
specific enthalpy is first determined and is then multiplied by the stream flow rate,
provided that the composition of the stream is known. The pressure of the stream remains
constant as zero pressure drop is assumed.
𝑇𝑖𝑛𝑡𝑅𝑒𝑓
= 𝑇𝑖𝑛𝑙𝑒𝑡 + [(𝑇𝑜𝑢𝑡𝑙𝑒𝑡−𝑇𝑖𝑛𝑙𝑒𝑡)
𝑁𝑖𝑛𝑡∙ 𝑖𝑛𝑡]; 𝑖𝑛𝑡 = 1, 2, 3, … 𝑁𝑖𝑛𝑡 (3.17)
𝐻𝑖𝑛𝑡𝑅𝑒𝑓
→ [𝑋𝑅𝑒𝑓 , 𝑇𝑖𝑛𝑡𝑅𝑒𝑓
, 𝑃𝑅𝑒𝑓] ∙ 𝐹𝑅𝑒𝑓 (3.18)
where Tint = Temperature of the refrigerant stream at the corresponding interval
Hint = Enthalpy of the refrigerant stream at Tint
Nint = Number of intermediate temperatures for calculating the T–H profile of the refrigerant stream
Chapter 3 Development and Design of Novel Refrigeration Cycles
77
The hot composite curve is generated by summing the enthalpy differences of the
individual hot streams (natural gas and hot refrigerants self-cooling) for each intermediate
temperature interval. Similarly, the cold composite curve is generated by summing the
enthalpy differences of the individual cold streams (evaporating refrigerant streams) in
each intermediate temperature interval.
A user-defined number of intermediate temperatures of the hot composite curve are then
compared against those of the cold composite curve at the same enthalpy values. The
resulting temperature difference is compared to the specified minimum temperature
difference for feasible heat transfer (ΔTMIN), according to Equation 3.19. Feasible heat
transfer is only considered when the temperature difference between the hot and cold
composite curves is greater than or equal to the specified minimum temperature difference,
in all the intervals evaluated.
∆𝑇𝑀𝐼𝑁 ≤ (𝑇𝑖𝑛𝑡𝐻𝑂𝑇 − 𝑇𝑖𝑛𝑡
𝐶𝑂𝐿𝐷) (3.19)
After the cold refrigerant streams have absorbed heat in the MSHE, they are recompressed
to the compressor discharge pressure. A multistage compression model is adopted for the
refrigerant compression (Figure 3.14). Each of the three refrigerant streams has a different
pressure level and, therefore, enters the multistage compressor at a different compression
stage accordingly. The refrigerant stream at the lowest pressure level enters the low-
pressure stage of the compressor, where it is partially compressed to an intermediate
pressure. Then, the stream is cooled down with an intercooler and mixed with the second
refrigerant stream (at the same pressure). The resulting stream is further compressed,
cooled in an intercooler and mixed isobarically with the remaining stream. Finally, the
total refrigerant is compressed to the compressor discharge pressure and cooled down in
the condenser. A zero pressure drop is assumed in the intercoolers, and it is also assumed
that the refrigerant is cooled to the same temperature as in the condenser (TCond).
Chapter 3 Development and Design of Novel Refrigeration Cycles
78
Figure 3. 14. Multistage compression.
Using Equation 3.20, the integer number of compression stages is determined iteratively in
order to have a pressure ratio equal to or less than 3. The pressure ratio is first calculated
assuming a single compression stage; if the resulting pressure ratio is greater than 3, then
the number of compression stages is increased by one, and the pressure ratio is calculated
again. The outlet pressure of each compression stage is calculated by Equation 3.21.
𝑃𝑅𝐴𝑇 = √𝑃𝑜𝑢𝑡
𝑃𝑖𝑛
𝑆𝑡𝑔𝑠 𝑃𝑅𝐴𝑇 ≤ 3 (3.20)
𝑃𝑆𝑡𝑔,𝑜𝑢𝑡𝑅𝑒𝑓1
= 𝑃𝑅𝐴𝑇 ∙ 𝑃𝑆𝑡𝑔,𝑖𝑛𝑅𝑒𝑓1
(3.21)
where PRAT = Ratio of outlet to inlet pressure of each compression stage
Stgs = Number of compression stages
PStg = Pressure of the refrigerant stream at the inlet and outlet of the compression stage
For each compression stage, the shaft power required is calculated using an isentropic
model. The specific entropy and enthalpy values of the stream at the inlet of the
compression stage are calculated using Relation 3.22 at the corresponding stage inlet
temperature and inlet pressure. Since isentropic compression is assumed (sStg,in = sStg,out),
the outlet specific enthalpy and outlet temperature can be obtained from Relation 3.23 at
the compression stage outlet pressure. The isentropic efficiency (ηIS, user-specified) is then
introduced to account for compression inefficiencies, and the shaft power is calculated
with an energy balance around the compressor, considering the corresponding refrigerant
flow rate, using Equation 3.24.
𝑠𝑆𝑡𝑔,𝑖𝑛𝑅𝑒𝑓1
, ℎ𝑆𝑡𝑔,𝑖𝑛𝑅𝑒𝑓1
→ 𝑋𝑅𝑒𝑓1, 𝑇𝑆𝑡𝑔,𝑖𝑛𝑅𝑒𝑓1
, 𝑃𝑆𝑡𝑔,𝑖𝑛𝑅𝑒𝑓1
(3.22)
ℎ𝑆𝑡𝑔,𝑜𝑢𝑡𝑅𝑒𝑓1
, 𝑇𝑆𝑡𝑔,𝑜𝑢𝑡𝑅𝑒𝑓1
→ 𝑋𝑅𝑒𝑓1, 𝑃𝑆𝑡𝑔,𝑜𝑢𝑡𝑅𝑒𝑓1
, 𝑠𝑆𝑡𝑔,𝑖𝑛𝑅𝑒𝑓1
(3.23)
Chapter 3 Development and Design of Novel Refrigeration Cycles
79
𝑊𝑆𝑡𝑔 = 𝐹𝑅𝑒𝑓1 ∙ (ℎ𝑆𝑡𝑔,𝑜𝑢𝑡
𝑅𝑒𝑓1−ℎ𝑆𝑡𝑔,𝑖𝑛
𝑅𝑒𝑓1
𝜂𝐼𝑆) (3.24)
where sStg,in = Specific entropy of the refrigerant stream at the inlet of the compression stage
hStg,in = Specific enthalpy of the refrigerant stream at the inlet of the compression stage
hStg,out = Specific enthalpy of the refrigerant stream at the outlet of the compression stage
TStg,in = Temperature of the refrigerant stream at the inlet of the compression stage
TStg,out = Temperature of the refrigerant stream at the outlet of the compression stage
ηIS = Compression isentropic efficiency
WStg = Shaft power demand of the compression stage
After a refrigerant stream has been compressed, it is cooled down by an intercooler, which
is in between two compression stages (see Figure 3.14). The duty in the intercooler is
calculated from the change in enthalpy of the refrigerant stream that is cooled from its inlet
temperature to TCond. The inlet and outlet enthalpies of the refrigerant in the intercooler are
obtained with Relations 3.25 and 3.26, respectively. The duty in the intercooler is
calculated by Equation 3.27.
ℎ𝑐𝑜𝑜𝑙,𝑖𝑛𝑅𝑒𝑓1
→ 𝑋𝑅𝑒𝑓1, 𝑇𝑆𝑡𝑔,𝑜𝑢𝑡𝑅𝑒𝑓1
, 𝑃𝑆𝑡𝑔,𝑜𝑢𝑡𝑅𝑒𝑓1
(3.25)
ℎ𝑐𝑜𝑜𝑙,𝑜𝑢𝑡𝑅𝑒𝑓1
→ 𝑋𝑅𝑒𝑓1, 𝑇𝐶𝑜𝑛𝑑, 𝑃𝑆𝑡𝑔,𝑜𝑢𝑡𝑅𝑒𝑓1
(3.26)
𝑄𝑐𝑜𝑜𝑙 = 𝐹𝑅𝑒𝑓1 ∙ (ℎ𝑐𝑜𝑜𝑙,𝑖𝑛𝑅𝑒𝑓1
− ℎ𝑐𝑜𝑜𝑙,𝑜𝑢𝑡𝑅𝑒𝑓1
) (3.27)
where hcool = Specific enthalpy of the refrigerant at the inlet and outlet of the intercooler
Qcool
= Heat load in the intercooler
If the refrigerant stream is partially condensed after being cooled in the intercooler, the
liquid formed is separated in a flash unit and then pumped to the compressor discharge
pressure. An isentropic model is considered for the pump, including an isentropic
efficiency, to calculate the power consumption. The vapour fraction of the refrigerant after
the intercooler and the composition of the liquid phase are obtained with Relation 3.28.
The flow rate of the condensed liquid is obtained with Equation 3.29. The specific enthalpy
and entropy of the liquid at the inlet of the pump is obtained with Relation 3.30. The
specific outlet enthalpy and the outlet temperature of the liquid are calculated using
Relation 3.31. The shaft power of the pump is calculated with an energy balance using the
isentropic efficiency (Equation 3.32).
𝑉𝐹𝑐𝑜𝑜𝑙𝑅𝑒𝑓1
, 𝑋𝑐𝑜𝑜𝑙𝐿𝑖𝑞 → 𝑋𝑅𝑒𝑓1, 𝑇𝐶𝑜𝑛𝑑 , 𝑃𝑆𝑡𝑔,𝑜𝑢𝑡
𝑅𝑒𝑓1 (3.28)
Chapter 3 Development and Design of Novel Refrigeration Cycles
80
𝐹𝑃𝑢𝑚𝑝 = [𝐹𝑅𝑒𝑓1 ∙ (1 − 𝑉𝐹𝑐𝑜𝑜𝑙𝑅𝑒𝑓1
)] (3.29)
ℎ𝑖𝑛𝑃𝑢𝑚𝑝, 𝑠𝑖𝑛
𝑃𝑢𝑚𝑝 → 𝑋𝑐𝑜𝑜𝑙𝐿𝑖𝑞 , 𝑇𝐶𝑜𝑛𝑑, 𝑃𝑆𝑡𝑔,𝑜𝑢𝑡
𝑅𝑒𝑓1 (3.30)
ℎ𝑜𝑢𝑡𝑃𝑢𝑚𝑝, 𝑇𝑜𝑢𝑡
𝑃𝑢𝑚𝑝 → 𝑋𝑐𝑜𝑜𝑙𝐿𝑖𝑞 , 𝑃𝐻𝑖𝑔ℎ, 𝑠𝑖𝑛
𝑃𝑢𝑚𝑝 (3.31)
𝑊𝑃𝑢𝑚𝑝 = 𝐹𝑃𝑢𝑚𝑝 ∙ (ℎ𝑜𝑢𝑡
𝑃𝑢𝑚𝑝−ℎ𝑖𝑛
𝑃𝑢𝑚𝑝
𝜂𝐼𝑆) (3.32)
where VFcool = Vapour fraction of the refrigerant stream after the intercooler
XcoolLiq
= Vector for the composition of the liquid condensed after refrigerant intercooling
FPump
= Molar flow rate of the refrigerant stream entering the pump
sinPump
= Specific entropy of the refrigerant stream at the inlet of the pump
hin,outPump
= Specific enthalpy of the refrigerant stream at the inlet and outlet of the pump
ToutPump
= Temperature of the refrigerant stream at the outlet of the pump
WPump
= Shaft power demand of the pump
When two refrigerant streams are mixed at an intermediate pressure level, the temperature
of the resulting stream has to be calculated to determine the inlet conditions of the
refrigerant (temperature and pressure) to the next compression stage. An energy balance is
used to obtain the specific enthalpy of the resulting stream after mixing (Equation 3.33)
and the temperature of the resulting stream is then calculated using Relation 3.34.
ℎ𝑀𝑖𝑥𝑅𝑒𝑓3
= [(𝐹𝑅𝑒𝑓1∙ℎ𝑅𝑒𝑓1)+(𝐹𝑅𝑒𝑓2∙ℎ𝑅𝑒𝑓2)
(𝐹𝑅𝑒𝑓1+𝐹𝑅𝑒𝑓2)] (3.33)
𝑇𝑀𝑖𝑥𝑅𝑒𝑓3
→ 𝑋𝑀𝑖𝑥𝑅𝑒𝑓3
, 𝑃𝑆𝑡𝑔𝑅𝑒𝑓3
, ℎ𝑀𝑖𝑥𝑅𝑒𝑓3
(3.34)
where hMix = Specific enthalpy of the stream resulting from mixing Ref1 and Ref2 in the compressor
TMix = Temperature of the streams resulting from mixing Ref1 and Ref2 in the compressor
The total compression shaft power is calculated as the sum of the power required by each
compressor stage plus the power consumed by the pumps (Equation 3.35):
𝑊𝑇𝑜𝑡𝑎𝑙 = ∑ 𝑊𝑖𝑆𝑡𝑔
+ ∑ 𝑊𝑖𝑃𝑢𝑚𝑝
(3.35)
3.3.2 Modelling of the Two Flash Levels design
The Two Flash Levels refrigeration cycle is displayed in Figure 3.15. The values for the
degrees of freedom are specified, and then the vapour fraction of the refrigerant is obtained
with Relation 3.36. The compositions of the flash unit product streams are obtained with
Relations 3.37 and 3.38, and their corresponding flow rates are calculated with Equations
3.39 and 3.40. The flow rate fed to the second flash unit (F2nd
) is calculated with Equation
Chapter 3 Development and Design of Novel Refrigeration Cycles
81
3.41, given the flow rate fraction of the liquid phase (f 2nd
) that is expanded to the second
pressure level.
𝑉𝐹𝑀𝑅 → 𝑋𝑀𝑅 , 𝑇𝐶𝑜𝑛𝑑, 𝑃𝐻𝑖𝑔ℎ (3.36)
𝑋𝑉𝑎𝑝 → 𝑋𝑀𝑅 , 𝑇𝐶𝑜𝑛𝑑, 𝑃𝐻𝑖𝑔ℎ (3.37)
𝑋𝐿𝑖𝑞 → 𝑋𝑀𝑅 , 𝑇𝐶𝑜𝑛𝑑, 𝑃𝐻𝑖𝑔ℎ (3.38)
𝐹𝑉𝑎𝑝 = 𝑉𝐹𝑀𝑅 ∙ 𝐹𝑀𝑅 (3.39)
𝐹𝐿𝑖𝑞 = (1 − 𝑉𝐹𝑀𝑅) ∙ 𝐹𝑀𝑅 (3.40)
𝐹2𝑛𝑑 = 𝑓2𝑛𝑑 ∙ 𝐹𝐿𝑖𝑞 (3.41)
Figure 3. 15. Two Flash Levels design: degrees of freedom.
The temperature and vapour fraction of the stream that is expanded to the second pressure
level is obtained with Relation 3.42, assuming isenthalpic expansion of the liquid from the
first flash unit. The resulting two-phase refrigerant mixture is separated into vapour and
liquid in the second flash unit. The compositions of the product streams from the second
flash unit are determined with Relations 3.43 and 3.44, whilst the flow rate of each phase is
calculated with Equations 3.45 and 3.46.
𝑇2𝑛𝑑 , 𝑉𝐹2𝑛𝑑 → 𝑋𝐿𝑖𝑞 , 𝑃2𝑛𝑑 , ℎ𝐿𝑖𝑞 (3.42)
𝑋𝑉𝑎𝑝2 → 𝑋𝐿𝑖𝑞 , 𝑇2𝑛𝑑 , 𝑃2𝑛𝑑 (3.43)
𝑋𝐿𝑖𝑞2 → 𝑋𝐿𝑖𝑞 , 𝑇2𝑛𝑑 , 𝑃2𝑛𝑑 (3.44)
Chapter 3 Development and Design of Novel Refrigeration Cycles
82
𝐹𝑉𝑎𝑝2 = 𝑉𝐹2𝑛𝑑 ∙ 𝐹2𝑛𝑑 (3.45)
𝐹𝐿𝑖𝑞2 = (1 − 𝑉𝐹2𝑛𝑑) ∙ 𝐹2𝑛𝑑 (3.46)
where T 2nd
= Temperature of the refrigerant stream entering the second flash unit
VF 2nd
= Vapour fraction of the refrigerant stream entering the second flash unit
X Vap2,Liq2
= Vector for the composition of the vapour and liquid streams from the second flash unit
F Vap2,Liq2
= Molar flow rate of the vapour and liquid streams leaving the second flash unit
The vapour and liquid of the second flash unit are partially mixed (isobarically), and the
resulting composition of Streams 3 and Stream 4 are calculated with Equations 3.47 and
3.48 for each component (xi) in the refrigerant mixture. Equations 3.49 and 3.50 calculate
the flow rate of Stream 3 and Stream 4, respectively.
𝑥𝑖𝑆3 = [
(𝑉𝐹2𝑛𝑑∙𝑓𝑉𝑎𝑝2∙𝑥𝑖𝑉𝑎𝑝2
) + ((1−𝑉𝐹2𝑛𝑑)∙𝑓𝐿𝑖𝑞2∙𝑥𝑖𝐿𝑖𝑞2
)
(𝑉𝐹2𝑛𝑑∙𝑓𝑉𝑎𝑝2) + ((1−𝑉𝐹2𝑛𝑑)∙𝑓𝐿𝑖𝑞2)]; 𝑥𝑖
𝑆3 ∈ 𝑋𝑆3 (3.47)
𝑥𝑖𝑆4 = [
(𝑉𝐹2𝑛𝑑∙(1−𝑓𝑉𝑎𝑝2)∙𝑥𝑖𝑉𝑎𝑝2
) + ((1−𝑉𝐹2𝑛𝑑)∙(1−𝑓𝐿𝑖𝑞2)∙𝑥𝑖𝐿𝑖𝑞2
)
(𝑉𝐹2𝑛𝑑∙(1−𝑓𝑉𝑎𝑝2)) + ((1−𝑉𝐹2𝑛𝑑)∙(1−𝑓𝐿𝑖𝑞2))];
𝑥𝑖𝑆4 ∈ 𝑋𝑆4 (3.48)
𝐹𝑆3 = (𝐹𝑉𝑎𝑝2 ∙ 𝑓𝑉𝑎𝑝2) + (𝐹𝐿𝑖𝑞2 ∙ 𝑓𝐿𝑖𝑞2) (3.49)
𝐹𝑆4 = (𝐹𝑉𝑎𝑝2 ∙ (1 − 𝑓𝑉𝑎𝑝2)) + (𝐹𝐿𝑖𝑞2 ∙ (1 − 𝑓𝐿𝑖𝑞2)) (3.50)
where xiS3,S4
= Mole fraction of the ith component in Stream 3 and Stream 4
xiVap2,Liq2
= Mole fraction of the ith component in the vapour and liquid streams from 2nd flash unit
X S3,S4
= Vector for the composition of Stream 3 and Stream 4
f Vap2
= Flow rate fraction of vapour from the second flash unit, mixed with f Liq2
to create Stream 3
f Liq2
= Flow rate fraction of liquid from the second flash unit, mixed with f Vap2
to create Stream 3
F S3,S4
= Molar flow rate of Stream 3 and Stream 4
S3,S4 = Stream 3, Stream 4
The hot streams (Stream 1 to Stream 4) enter the MSHE and are precooled to their
specified temperature and their specific enthalpy is calculated with Relation 3.51 (Stream 1
and Stream 2) and Relation 3.52 (Stream 3 and Stream 4). After expansion to their
corresponding pressure level, the temperature of each stream is calculated using Relation
3.53 under the assumption of isenthalpic expansion.
ℎ𝐴𝑃 → 𝑋𝑅𝑒𝑓 , 𝑇𝐴𝑃, 𝑃𝐻𝑖𝑔ℎ (3.51)
ℎ𝐴𝑃 → 𝑋𝑅𝑒𝑓 , 𝑇𝐴𝑃, 𝑃2𝑛𝑑 (3.52)
𝑇𝐴𝐸 → 𝑋𝑅𝑒𝑓 , 𝑃𝐴𝐸 , ℎ𝐴𝑃 (3.53)
Chapter 3 Development and Design of Novel Refrigeration Cycles
83
The specific enthalpy of each cold stream at the outlet of the MSHE is calculated with
Relation 3.54. Heat rejected by each hot stream is then calculated with Equation 3.55,
whereas heat absorbed by each cold stream is obtained with Equation 3.56. The overall
refrigerant flow rate is then calculated (Equation 3.57) with an energy balance around the
MSHE.
ℎ𝑜𝑢𝑡 → 𝑋𝑅𝑒𝑓, 𝑇𝑜𝑢𝑡, 𝑃𝐴𝐸 (3.54)
∆ℎℎ𝑜𝑡 = (ℎ𝑖𝑛 − ℎ𝐴𝑃) (3.55)
∆ℎ𝑐𝑜𝑙𝑑 = (ℎ𝑜𝑢𝑡 − ℎ𝐴𝑃) (3.56)
𝐹𝑀𝑅 = [𝑄𝑁𝐺
(∆ℎ𝑐𝑜𝑙𝑑𝑆1 +∆ℎ𝑐𝑜𝑙𝑑
𝑆2 +∆ℎ𝑐𝑜𝑙𝑑𝑆3 +∆ℎ𝑐𝑜𝑙𝑑
𝑆4 ) − (∆ℎℎ𝑜𝑡𝑆1 +∆ℎℎ𝑜𝑡
𝑆2 +∆ℎℎ𝑜𝑡𝑆3 +∆ℎℎ𝑜𝑡
𝑆4 )] (3.57)
Feasible heat transfer is checked as described in Section 3.3.1. Similarly, after the cold
streams have absorbed heat from the hot streams (refrigerants self-cooling and natural gas
condensing), the cold streams are returned to the high pressure level (i.e. the compressor
discharge pressure) and shaft power calculations are performed as described in Section
3.3.1.
3.3.3 Modelling of the Mixing After Precooling design
The Mixing After Precooling refrigeration cycle is shown in Figure 3.16. The values for
the degrees of freedom are first specified. The vapour fraction of the mixed refrigerant is
obtained with Relation 3.58. The refrigerant is flashed and the compositions of the product
streams are obtained with Relations 3.59 and 3.60 for the vapour and liquid phase,
respectively. The flow rate of each stream (vapour and liquid) is calculated using
Equations 3.61 and 3.62.
𝑉𝐹𝑀𝑅 → 𝑋𝑀𝑅 , 𝑇𝐶𝑜𝑛𝑑, 𝑃𝐻𝑖𝑔ℎ (3.58)
𝑋𝑉𝑎𝑝 → 𝑋𝑀𝑅 , 𝑇𝐶𝑜𝑛𝑑, 𝑃𝐻𝑖𝑔ℎ (3.59)
𝑋𝐿𝑖𝑞 → 𝑋𝑀𝑅 , 𝑇𝐶𝑜𝑛𝑑, 𝑃𝐻𝑖𝑔ℎ (3.60)
𝐹𝑉𝑎𝑝 = 𝑉𝐹𝑀𝑅 ∙ 𝐹𝑀𝑅 (3.61)
𝐹𝐿𝑖𝑞 = (1 − 𝑉𝐹𝑀𝑅) ∙ 𝐹𝑀𝑅 (3.62)
Chapter 3 Development and Design of Novel Refrigeration Cycles
84
Figure 3. 16. Mixing After Precooling design: degrees of freedom.
The composition of the hot refrigerants Stream 1 and Stream 2 (XS1
and XS2
) is calculated
with Equations 3.63 and 3.64, respectively, with a mass balance of each component (xi) in
the mixture after partial mixing of vapour and liquid phases from the flash unit (isobaric
mixing is assumed). The flow rate for each hot refrigerant stream is calculated with
Equations 3.65 and 3.66.
𝑥𝑖𝑆1 = [
(𝑉𝐹𝑀𝑅∙𝑓𝑉𝑎𝑝∙𝑥𝑖𝑉𝑎𝑝
) + ((1−𝑉𝐹𝑀𝑅)∙𝑓𝐿𝑖𝑞∙𝑥𝑖𝐿𝑖𝑞
)
(𝑉𝐹𝑀𝑅∙𝑓𝑉𝑎𝑝) + ((1−𝑉𝐹𝑀𝑅)∙𝑓𝐿𝑖𝑞)]; 𝑥𝑖
𝑆1 ∈ 𝑋𝑆1 (3.63)
𝑥𝑖𝑆2 = [
(𝑉𝐹𝑀𝑅∙𝑓𝑉𝑎𝑝∙𝑥𝑖𝑉𝑎𝑝
) + ((1−𝑉𝐹𝑀𝑅)∙𝑓𝐿𝑖𝑞∙𝑥𝑖𝐿𝑖𝑞
)
(𝑉𝐹𝑀𝑅∙𝑓𝑉𝑎𝑝) + ((1−𝑉𝐹𝑀𝑅)∙𝑓𝐿𝑖𝑞)]; 𝑥𝑖
𝑆2 ∈ 𝑋𝑆2 (3.64)
𝐹𝑆1 = (𝐹𝑉𝑎𝑝 ∙ 𝑓𝑉𝑎𝑝) + (𝐹𝐿𝑖𝑞 ∙ 𝑓𝐿𝑖𝑞) (3.65)
𝐹𝑆2 = (𝐹𝑉𝑎𝑝 ∙ (1 − 𝑓𝑉𝑎𝑝)) + (𝐹𝐿𝑖𝑞 ∙ (1 − 𝑓𝐿𝑖𝑞)) (3.66)
The hot refrigerant streams are precooled in the MSHE and then expanded with throttle
valves. The specific enthalpy after precooling is calculated with Relation 3.67. The
temperature of Stream 1 and Stream 2 after expansion is calculated using Relation 3.68,
respectively, assuming isenthalpic expansion. Once expanded, a flow rate fraction ‘α’ from
Stream 1 is mixed with a flow rate fraction ‘β’ from Stream 2 (isobaric mixing is
assumed), in order to create a new cold stream with a different composition (i.e. Stream 5).
In order to partially mix both streams, the refrigerant flow rate fraction α from Stream 1
has to be expanded to the same pressure level of Stream 2. The flow rate fraction α from
Stream 1 is mixed with the flow rate fraction β from Stream 2. The temperature of the
Chapter 3 Development and Design of Novel Refrigeration Cycles
85
refrigerant fraction α, once expanded, is obtained with Relation 3.69. The flow rates of the
three cold streams (Stream 3, Stream 4 and Stream 5) are calculated by Equations 3.70 to
3.72. The composition of the new stream (Stream 5) is calculated similarly as with Stream
1 and Stream 2 (Equation 3.73). The temperature of Stream 5, after flow rate fractions α
and β are mixed, is obtained with Relation 3.75 after obtaining its specific enthalpy value
resulting from the mixing process (Equation 3.74).
ℎ𝐴𝑃 → 𝑋𝑅𝑒𝑓 , 𝑇𝐴𝑃, 𝑃𝐻𝑖𝑔ℎ (3.67)
𝑇𝐴𝐸 → 𝑋𝑅𝑒𝑓 , 𝑃𝐴𝐸 , ℎ𝐴𝑃 (3.68)
𝑇𝐴𝐸𝛼 → 𝑋𝑆1, 𝑃𝐴𝐸
𝑆2, ℎ𝐴𝑃𝑆1 (3.69)
𝐹𝑆3 = 𝐹𝑆1 ∙ (1 − 𝛼) (3.70)
𝐹𝑆4 = 𝐹𝑆2 ∙ (1 − 𝛽) (3.71)
𝐹𝑆5 = (𝐹𝑆1 ∙ 𝛼) + (𝐹𝑆2 ∙ 𝛽) (3.72)
𝑥𝑖𝑆5 = [
(𝐹𝑆1∙ 𝛼 ∙ 𝑥𝑖𝑆1) + (𝐹𝑆2∙ 𝛽 ∙ 𝑥𝑖
𝑆2)
(𝐹𝑆1∙ 𝛼) + (𝐹𝑆2∙ 𝛽)]; 𝑥𝑖
𝑆5 ∈ 𝑋𝑆5 (3.73)
ℎ𝑀𝑖𝑥𝑆5 = [
(𝐹𝑆1∙ 𝛼 ∙ ℎ𝐴𝑃𝑆1 )+(𝐹𝑆2∙ 𝛽 ∙ ℎ𝐴𝑃
𝑆2 )
(𝐹𝑆1∙ 𝛼)+(𝐹𝑆2∙ 𝛽)] (3.74)
𝑇𝑀𝑖𝑥𝑆5 → 𝑋𝑆5, 𝑃𝐴𝐸
𝑆2, ℎ𝑀𝑖𝑥𝑆5 (3.75)
where α = Flow rate fraction from Stream 1, mixed to create Stream 5
β = Flow rate fraction from Stream 2, mixed to create Stream 5
xiS5
= Mole fraction of the ith component in Stream 5
XS5
= Vector for the composition of Stream 5
hMixS5
= Specific enthalpy of Stream 5, resulting from mixing α and β
TMixS5
= Temperature of Stream 5, after mixing α and β
S1,S2,S3,S4,S5 = Stream1, Stream 2, Stream 3, Stream 4, Stream 5
The specific enthalpy of the cold streams at the outlet of the MSHE is then calculated with
Relation 3.76. The heat rejected by each hot stream and the heat absorbed by each cold
stream can thus be calculated by Equations 3.77 and 3.78, respectively. The overall
refrigerant flow rate is calculated from an energy balance around the MSHE (Equation
3.79).
ℎ𝑜𝑢𝑡 → 𝑋𝑅𝑒𝑓, 𝑇𝑜𝑢𝑡, 𝑃𝐴𝐸 (3.76)
∆ℎℎ𝑜𝑡 = (ℎ𝑖𝑛 − ℎ𝐴𝑃) (3.77)
∆ℎ𝑐𝑜𝑙𝑑 = (ℎ𝑜𝑢𝑡 − ℎ𝐴𝑃) (3.78)
Chapter 3 Development and Design of Novel Refrigeration Cycles
86
𝐹𝑀𝑅 = [𝑄𝑁𝐺
(∆ℎ𝑐𝑜𝑙𝑑𝑆3 +∆ℎ𝑐𝑜𝑙𝑑
𝑆4 +∆ℎ𝑐𝑜𝑙𝑑𝑆5 )−(∆ℎℎ𝑜𝑡
𝑆1 +∆ℎℎ𝑜𝑡𝑆2 )
] (3.79)
Feasible heat transfer is then checked at intermediate temperatures in the MSHE as
described in Section 3.3.1. The cold refrigerant streams leaving the MSHE are then
compressed to the compressor discharge pressure. Stream 4 and Stream 5 enter at the same
compression stage since both have the same pressure level. The multistage compression
model, described in Section 3.3.1, is then used to calculate the shaft power demand of the
refrigeration cycle.
3.3.4 Example of mixed refrigerant cycle modelling
In order to evaluate the accuracy of the models of the novel refrigeration cycles (and in the
absence of experimental data), the model of the Two Flash Levels design is simulated in
WORK software and compared to a simulation of the same refrigeration cycle in Aspen
HYSYS v8.2. That is, it is assumed that the results in Aspen HYSYS are an accurate
representation of the real data.
The natural gas stream in WORK software is entered as a T–H profile, whereas in HYSYS,
the full conditions of the natural gas stream are required, i.e. the composition, flow rate,
MSHE inlet and outlet temperatures as well as inlet and outlet pressures. The model in
WORK takes the temperatures of the cold refrigerants at the outlet of the MSHE as input,
and calculates the overall flow rate of the refrigerant with an energy balance. HYSYS, on
the other hand, takes the flow rate of the refrigerant and n–1 outlet temperatures of the cold
streams (n is the total number of cold streams leaving the MSHE) as input and calculates
the temperature of the remaining cold stream at the outlet of the MSHE.
The natural gas stream is first specified (i.e. composition, inlet and outlet temperatures,
etc.) in HYSYS. The MSHE model in HYSYS generates the temperature–enthalpy profiles
of each stream in the MSHE by calculating a user-specified number of intermediate
temperatures (relative to inlet and outlet temperatures of each stream). Thus, the T–H
profile of the natural gas stream is obtained. The natural gas T–H profile is then entered in
WORK software and the parameters for the mixed refrigerant cycle are specified (e.g.
refrigerant composition, streams pressure levels, etc.), allowing calculating the refrigerant
flow rate. The flow rate obtained in WORK is then input to HYSYS to run the simulation.
Chapter 3 Development and Design of Novel Refrigeration Cycles
87
For this example, a natural gas stream (1 kmol·s-1
) is to be liquefied at –163°C (110 K)
from 25°C (298 K). The natural gas is assumed to be at a pressure of 45 bar and to leave
the MSHE at 40 bar. The composition of the natural gas stream is taken from Cao et al.
(2006) and is on Table 3.1. The mixed refrigerant is composed of C1 – C4 and nitrogen (see
Table 3.1). The T–H data of the natural gas stream is provided in Table 3.2 and Figure
3.17.
Table 3. 1. Natural gas and mixed refrigerant composition [mole %] for the modelling example.
C1 C2 C3 n-C4 i-C4 N2
Natural gas* 82.0 11.2 4.0 0.9 1.2 0.7
Mixed refrigerant 20.0 40.0 10.0 20.0 0.0 10.0
*Composition obtained from Cao et al. (2006).
Table 3. 2. Natural gas temperature–enthalpy data (from HYSYS, 1 kmol·s-1
) for the modelling example.
Segment Supply Temperature [K] Target Temperature [K] ΔH [kW] CP [kW·K-1
]
1.1 298.1 276.7 1051.2 49.0
1.2 276.7 261.8 754.9 50.8
1.3 261.8 258.4 296.3 85.1
1.4 258.4 246.1 1051.2 86.0
1.5 246.1 233.9 1102.5 90.2
1.6 233.9 222.3 1147.5 98.9
1.7 222.3 213.3 1036.4 115.0
1.8 213.3 206.2 1036.4 145.3
1.9 206.2 201.1 1036.4 202.6
1.10 201.1 197.5 1036.4 290.1
1.11 197.5 195.6 695.5 367.4
1.12 195.6 191.6 340.9 85.5
1.13 191.6 178.1 1036.4 76.6
1.14 178.1 162.8 1036.4 67.8
1.15 162.8 147.5 954.9 62.5
1.16 147.5 130.3 1016.8 59.0
1.17 130.3 110.1 1137.5 56.5
Figure 3. 17. Natural gas temperature–enthalpy data (from HYSYS, 1 kmol·s-1
) for the modelling example.
100
130
160
190
220
250
280
310
0 3 6 9 12 15 18
Tem
per
atu
re [
K]
ΔH [MW]
Chapter 3 Development and Design of Novel Refrigeration Cycles
88
For the simulation it is assumed that the maximum pressure ratio for each compression
stage is 3. The compression isentropic efficiency is considered as 80%. Physical and
thermodynamic properties of the mixed refrigerant are calculated using Peng–Robinson
equation of state. In WORK software, the fluid properties of the mixed refrigerant are
calculated using Peng–Robinson equation by interfacing with Aspen HYSYS.
Table 3.3 compares the simulation in WORK with that in Aspen HYSYS v8.2. The results
obtained with the model of the Two Flash Levels design simulated in WORK are in good
agreement compared to those obtained in HYSYS since there is a difference of only 0.1%
in the shaft power calculated using HYSYS compared to the power demand calculated in
WORK. The main discrepancy in the calculation of the shaft power demand comes from
the pump model. The minimum temperature driving force in the simulation on WORK is
different by only 0.2°C compared to the simulation in HYSYS.
Table 3. 3. Results comparison of the Two Flash Levels design simulation example.
HYSYS v8.2 WORK Difference
Refrigerant flow rate [kmol·s-1
] 2.5a 2.5 -
Outlet pressure of compressor [bar] 45a 45
b -
Pressure second flash unit [bar] 15a 15
b -
Fraction of liquid to the second flash unit 0.6a 0.6
b -
Flow rate fraction of vapour for partial mixing (f Vap
):
- Stream 1
- Stream 2
- Stream 3
- Stream 4
0.75a
0.25a
0.70a
0.30a
0.75b
0.25b
0.70b
0.30b
-
-
-
-
Flow rate fraction of liquid for partial mixing (f Liq
):
- Stream 1
- Stream 2
- Stream 3
- Stream 4
0.17a
0.23a
0.25a
0.75a
0.17b
0.23b
0.25b
0.75b
-
-
-
-
Expansion pressure [bar]
- Stream 1
- Stream 2
- Stream 3
- Stream 4
1.2a
6.0a
11.5a
3.6a
1.2b
6.0b
11.5b
3.6b
-
-
-
-
MSHE outlet temperature [K]:
- Stream 1
- Stream 2
- Stream 3
- Stream 4
296.0a
297.0a
295.0a
284.4
296.0b
297.0b
295.0b
285.2b
-
-
-
0.8°C
Intercooling duty [MW] 35.06 36.07 2.8%
Number of compression stages 5 5 -
ΔTMIN [K] 2.1 1.9 0.2°C
Compression shaft power [MW] 20.00 20.02 0.1% aDenotes input data to HYSYS.
bDenotes input to WORK.
Chapter 3 Development and Design of Novel Refrigeration Cycles
89
3.4 Strategy for the evaluation of the novel designs
To evaluate the novel refrigeration cycles, the same natural gas stream employed in the
CryoMan process is used as the liquefaction duty. The natural gas stream data is provided
only as a T–H profile in a tabular form (see Table 3.4) and was first published by Lee
(2001).
The evaluation of the novel refrigeration cycles is divided in two stages. Firstly, the novel
refrigeration cycles are simulated in Aspen HYSYS v8.2 to evaluate their shaft power
consumption through sensitivity analyses on each degree of freedom (e.g. refrigerant
composition, pressure levels, etc.). Because the natural gas stream is only provided as a T–
H profile, its full conditions (e.g. composition, inlet and outlet pressure, etc.) have to be
determined in order to be simulated in HYSYS for the assessment of the novel
refrigeration cycles on the same liquefaction duty basis as the CryoMan process. An
optimisation approach is employed to find a combination of values for the natural gas
stream conditions that minimises the sum of squared difference of the enthalpies between
the T–H profile of the optimised stream and that in Table 3.4.
In the second stage of the evaluation, only the refrigeration cycles that show shaft power
savings in the sensitivity analyses, compared to the shaft power consumption of the
CryoMan process, are considered for optimisation in WORK software. The degrees of
freedom (e.g. refrigerant composition, pressure levels, etc.) of each novel refrigeration
cycle are optimised and the resulting performance indicators (including shaft power
demand and the ΔTMIN) are compared against those in the CryoMan process as well as
those in the PRICO cycle.
Table 3. 4. Temperature–enthalpy profile of the natural gas stream published by Lee (2001).
Segment Supply
Temperature [°C]
Target
Temperature [°C] ΔH [kW] CP [kW·K
-1]
1.1 25.00 –06.03 –1861.5 60
1.2 –06.03 –34.09 –1964.3 70
1.3 –34.09 –57.65 –1885.0 80
1.4 –57.65 –70.10 –2490.0 200
1.5 –70.10 –74.55 –1780.0 400
1.6 –74.55 –82.26 –3084.0 400
1.7 –82.26 –96.50 –1424.0 100
1.8 –96.50 –115.00 –1850.0 100
1.9 –115.00 –163.00 –3840.0 80
Chapter 3 Development and Design of Novel Refrigeration Cycles
90
3.4.1 Simulation and analysis
The novel refrigeration cycles are built in Aspen HYSYS v8.2 and sensitivity analyses are
implemented around each of their corresponding degrees of freedom (such as the
refrigerant composition, pressure levels, etc.). Each sensitivity analysis involves the
manipulation of only one degree of freedom.
The sensitivity analyses are performed as a preliminary study in order to investigate the
refrigeration cycle designs that show results towards reducing shaft power consumption for
refrigerant compression in the liquefaction of a natural gas stream, compared to the shaft
power required in the CryoMan process. Only designs that show shaft power savings are
considered for optimisation of their operating variables. Chapter 4 provides details of the
results obtained from the sensitivity studies in each of the novel refrigeration cycles.
3.4.2 Optimisation of promising designs
The promising refrigeration cycles are optimised in WORK software in Chapter 5. A
Genetic Algorithm is used as the optimisation method to find a set of values of the
operating variables (e.g. refrigerant composition, evaporating pressures, etc.) that yield the
lowest shaft power consumption of the novel refrigeration cycle. As defined in Section 3.1,
the design of novel refrigeration cycles is constrained to a single MSHE with up to 12
refrigerant streams involved in heat exchange and a single multistage centrifugal
compressor; also, the novel refrigeration are constrained to refrigerant compression with a
maximum pressure ratio (i.e. ratio of outlet to inlet pressure) of 3 in each compression
stage, and up to a maximum of 8 compression stages in the compressor.
Chapter 5 presents a case study for the liquefaction of the natural gas stream (presented in
Table 3.4), in which the operating variables of the novel refrigeration cycles are optimised
to minimise the shaft power consumption of the refrigeration cycle. Results are compared
against the CryoMan process and against the PRICO cycle. A detailed analysis of the
results obtained and an economic comparison (based on operating costs) between the novel
refrigeration cycles and the benchmark processes (i.e. the CryoMan process and the
PRICO cycle) is presented in Chapter 5 as well.
Chapter 3 Development and Design of Novel Refrigeration Cycles
91
3.5 Conclusions
The CryoMan process was analysed and the manipulation of the composition, by partial
mixing of the product streams of the flash unit, was identified as a key degree of freedom.
As a result of the composition manipulation, in the CryoMan process, LP Stream required
less shaft power for compression than LP Stream in the “Pre-flash” design. Moreover, a
trade-off was identified between the flow rate of each refrigerant stream and its pressure
level. The CryoMan process reduced the shaft power consumption, compared to the “Pre-
flash” design, by increasing the flow rate of LP Stream but reducing the pressure
difference between the intermediate pressure level (i.e. the pressure level of HP Stream)
and the compressor discharge pressure.
Conceptual modifications to the structure of the refrigeration cycle were suggested based
on the analysis performed to the CryoMan process and the trade-offs identified between
the operating conditions (e.g. flow rates, pressure levels, compositions, etc.), with the aim
of reducing the overall shaft power demand of the refrigeration cycle. To account for the
low complexity and compactness required in small scale LNG processes, the design of the
novel refrigeration cycles was constrained to use only one multistage compressor in which
the pressure ratio of each stage was not greater than 3, plus only one MSHE with a
maximum of 12 streams involved in heat exchange.
Three novel refrigeration cycles were developed (namely the Bypass design, the Two Flash
Levels design and the Mixing After Precooling design) with the suggested structural
modifications and under the limits defined for complexity. The novel refrigeration cycles
were successfully modelled as it was demonstrated through an example.
In the following Chapter 4, the novel refrigeration cycles are simulated in Aspen HYSYS
v8.2 in order to assess their performance (i.e. shaft power consumption) through sensitivity
analyses towards potential shaft power savings compared to the CryoMan process. Also,
the full conditions the natural gas stream (e.g. composition, inlet and outlet pressure, etc.)
are determined; details of the optimisation approach for the determination of the natural
gas conditions are in Appendix 1. Moreover, Chapter 5 presents a case study for the
liquefaction of the natural gas stream, in which the operating variables of the promising
novel refrigeration cycles are optimised, in order to minimise the total shaft power
consumption for refrigerant compression. Results of the optimised novel refrigeration
Chapter 3 Development and Design of Novel Refrigeration Cycles
92
cycles are compared to those of the CryoMan process and the PRICO cycle (see also
Appendixes 2 and 3). The optimisation is performed in WORK software by means of a
Genetic Algorithm.
Chapter 4 Evaluation of the Novel Refrigeration Cycles
93
Chapter 4 – Evaluation of the Novel Refrigeration Cycles
4.1 Introduction
As mentioned in Section 3.4, the novel refrigeration cycles are evaluated in two stages: a
preliminary assessment using Aspen HYSYS v8.2 (Aspen Technology Inc., 2013) through
sensitivity analyses, and the optimisation of the operating variables in WORK software of
the refrigeration cycles that show shaft power savings compared to the CryoMan process.
This chapter is focused on the first stage of the evaluation.
Aspen HYSYS is used to perform a preliminary study of the novel refrigeration cycles
through sensitivity analyses. An initial simulation of the novel refrigeration cycle is first
performed. Then, sensitivity studies explore the performance of each refrigeration cycle,
using the shaft power consumption and the minimum temperature approach in the multi-
stream heat exchanger (MSHE) as the performance indicators, when each degree of
freedom in their corresponding designs is varied (e.g. the bypass stream flow rate fraction,
the fraction of the streams partially mixed after a second flash separator, etc.). The
compression shaft power demand is compared to that of the reference case – the CryoMan
process – for the same liquefaction duty; therefore, the same natural gas stream employed
with the CryoMan process, provided only as a T–H profile, is used in the preliminary
assessment of the novel refrigeration cycles (see Section 4.2).
The novel refrigeration cycles are modelled in HYSYS as described in Section 3.3. In the
sensitivity analyses, the degrees of freedom of each novel refrigeration cycle are
manipulated, one variable at the time, in order to assess their impact on the shaft power
demand for refrigerant compression. The novel designs that show shaft power savings,
compared to the CryoMan process, are selected for optimisation of their corresponding
operating variables using WORK software. The optimisation of the novel refrigeration
cycles is presented with a case study in Chapter 5.
4.2 Determination of the natural gas conditions
The procedure for obtaining the conditions of the natural gas stream is briefly described in
this section and the detailed procedure is provided in Appendix 1.
The natural gas stream, first published by Lee (2001, Ch. 4), and later employed by Del
Nogal (2006, Ch. 2) and Zheng (2009, Ch. 3), is provided only as a temperature–enthalpy
Chapter 4 Evaluation of the Novel Refrigeration Cycles
94
profile in a tabulated form, as shown in Table 4.1. Remeljej and Hoadley (2006) provided a
composition and mass flow rate for the same natural gas stream in the evaluation of four
different refrigeration cycles, including the PRICO cycle.
Additional information in Lee (2001, Ch. 4) is that the natural gas stream enters the MSHE
at 55 bar and leaves at 50 bar, i.e. there is a pressure drop of 5 bar, although the pressure
drop profile inside the MSHE is not specified. The flow rate of the liquefied natural gas is
not mentioned either.
Table 4. 1. Natural gas stream data presented by Lee (2001), Del Nogal (2006) and Zheng (2009).
Segment Supply
Temperature [°C]
Target
Temperature [°C] ΔH [kW] CP [kW·K
-1]
1.1 25.00 –06.03 –1861.5 60
1.2 –06.03 –34.09 –1964.3 70
1.3 –34.09 –57.65 –1885.0 80
1.4 –57.65 –70.10 –2490.0 200
1.5 –70.10 –74.55 –1780.0 400
1.6 –74.55 –82.26 –3084.0 400
1.7 –82.26 –96.50 –1424.0 100
1.8 –96.50 –115.00 –1850.0 100
1.9 –115.00 –163.00 –3840.0 80
To simulate in HYSYS the natural gas stream that Zheng (2009) employed to optimise the
CryoMan process (see Table 4.1), the full conditions of the stream are required, i.e. the
flow rate, composition, inlet and outlet pressures and temperatures, as well as the pressure
drop profile across the MSHE.
However, the pressure drop profile of the natural gas stream in the MSHE is not provided
in any of the publications previously mentioned. Moreover, the molar flow rate of the
natural gas obtained according to Remeljej and Hoadley (2006) (1.37 kmol·s-1
) is
conflicting when compared to 1 kmol·s-1
used by Zheng (2009) [stated explicitly in Zheng
(2007)] to optimise the CryoMan process. In addition, the T–H profile obtained from the
natural gas stream according to the conditions employed by Remeljej and Hoadley (2006)
does not match that originally presented by Lee (2001), as shown in Figure 4.1. The
conditions of the natural gas stream (composition, flow rate, inlet and outlet pressure, and
pressure drop profile in the MSHE) are thus unclear.
Chapter 4 Evaluation of the Novel Refrigeration Cycles
95
Figure 4. 1. Temperature–enthalpy profile according to data from Remeljej and Hoadley (2006) compared to
that published by Lee (2001).
In order to model the natural gas stream in HYSYS, this work determines the conditions of
the stream (composition, flow rate, inlet and outlet pressures, etc.) through an optimisation
approach that minimises the sum of squared enthalpy differences in the T–H profile
generated with the optimised stream and that provided in Table 4.1, as illustrated in Figure
4.2. The objective function to minimise is that in Equation 4.1. The stream T–H profile is
affected by the composition of the stream, the flow rate and the inlet and outlet pressures.
Thus, the variables to optimise are the natural gas mass flow rate, the molar composition,
and the inlet pressure (represented in Equation 4.1 as the vector x*). The optimisation is
performed with three different pressure drop profiles in the MSHE: i) linear dependence on
the temperature change; ii) linear dependence on the heat rejected by the natural gas stream
(enthalpy change); and iii) zero pressure drop assumed inside the MSHE (i.e. Pin = Pout).
The natural gas is assumed to be composed of methane, ethane, propane, n-butane, i-
butane and nitrogen.
𝑓(𝐱∗) = ∑ (∆𝐻𝑖 − ∆𝐻𝑐𝑎𝑙𝑐)29𝑖=1 𝑖 = 1,2, … ,9 (4.1)
where x* = [x1, x2, x3, x4, x5, m, Pin]
ΔHi = Enthalpy change value of Lee’s natural gas stream at the ith temperature segment
ΔHcalc = Enthalpy change value of the optimised stream at the ith temperature segment
xj = Mole fraction of the jth component in the optimised natural gas stream
m = Normalised value for the mass flow rate of the optimised natural gas stream
Pin = Normalised value for the inlet pressure of the optimised natural gas stream
-200
-150
-100
-50
0
50
0 5 10 15 20 25
Tem
per
atu
re [
°C]
ΔH [MW]
Remeljej and Hoadley (2006)
Lee (2001)
Chapter 4 Evaluation of the Novel Refrigeration Cycles
96
Figure 4. 2. Determination of the natural gas stream conditions through minimisation of the sum of squared
difference of the enthalpy profiles against the data provided by Lee (2001, Ch. 4).
Because the T–H profile of the natural gas depends on the interactions between its
components, and on the overall composition, pressures and flow rate, the optimisation is
performed using fmincon solver in MATLAB for nonlinear problems (The MathWorks
Inc., 2013). Table 4.2 shows the conditions of the natural gas stream that are obtained from
the optimisation. The optimisation in which zero pressure drop is assumed inside the
MSHE resulted in the closest match of the T–H profiles between the optimised stream and
the data originally presented by Lee (2001), as shown in Figure 4.3. According to the flow
rate of the natural gas obtained (24.03 kg·s-1
), the LNG production is 0.75 million t per
annum.
Table 4. 2. Natural gas stream conditions resulted from the optimisation (zero pressure drop in the MSHE).
Composition [mole fraction] Flow rate Inlet pressure
C1 C2 C3 n-C4 N2 i-C4 [kg·s-1
] [bar]
0.9000 0.0940 0.0047 0.0013 0.0000 0.0000 24.03 43.86
Chapter 4 Evaluation of the Novel Refrigeration Cycles
97
Figure 4. 3. Temperature–enthalpy profile of the optimised stream compared to the data provided by Lee
(2001).
The resulting molar flow rate of the optimised natural gas stream is 1.37 kmol·s-1
, which is
higher than that stated by Zheng (2009) (1 kmol·s-1
). Although the value of molar flow rate
is the same as that obtained from Remeljej and Hoadley (2006), the corresponding mass
flow rate is nearly 6% greater in the optimised natural gas stream (i.e. 24.03 kg·s-1
compared to 22.60 kg·s-1
), as a result of the difference in composition,. The composition of
the optimised natural gas stream is compared to that provided by Remeljej and Hoadley
(2006) in Table 4.3. The ethane and propane mole fractions are increased by nearly 0.065
and 0.004, respectively, in the optimised stream compared to that from Remeljej and
Hoadley (2006); the mole fraction of methane, on the other hand, is decreased by nearly
0.070 in the optimised natural gas stream.
Table 4. 3. Composition of optimised natural gas stream compared to that of Remeljej and Hoadley (2006).
Natural gas composition [mole fraction]
C1 C2 C3 n-C4 N2 i-C4
Optimised natural gas stream 0.9000 0.0940 0.0047 0.0013 0.0000 0.0000
Remeljej and Hoadley (2006) 0.9693 0.0294 0.0006 0.0001 0.0006 0.0000
The CryoMan process is simulated in HYSYS using the ‘reconstructed’ natural gas stream
obtained by optimisation (Table 4.2), and the conditions of the CryoMan process reported
in Zheng (2007) – shown in Table 4.4 and Figure 4.4 –, in order to compare the
refrigeration cycle simulation results against those reported by Zheng (2009). The ‘Label’
column in Table 4.4 refers to the letters in the CryoMan process shown in Figure 4.4.
-200
-150
-100
-50
0
50
0 5 10 15 20 25
Tem
per
atu
re [
°C]
ΔH [MW]
OptimisationLee (2001)
Chapter 4 Evaluation of the Novel Refrigeration Cycles
98
Also, in the simulation in HYSYS, the pressure drop of the refrigerant streams in the
MSHE is assumed to be negligible and the T–H profile of each stream is calculated at
equally-spaced intermediate temperatures between their corresponding MSHE inlet and
outlet temperatures. Physical and thermodynamic properties of the refrigerant streams (e.g.
temperatures, enthalpies, etc.) are calculated using the Peng–Robinson equation of state.
The refrigerant compression is performed with 4 compression stages, assuming refrigerant
intercooling to 30°C between compression stages.
The main results of the CryoMan process simulation (performance indicators) are also
presented in Table 4.4. The shaft power consumption of the CryoMan process is expressed
as the total shaft power (MW) needed for the liquefaction of the natural gas stream, and is
also expressed as specific shaft power consumption (kWh·kg-1
of LNG). The specific shaft
power demand is used for comparisons of the novel refrigeration cycles against the
CryoMan process, and can also be used for comparisons against refrigeration cycles for
LNG processes reported in the literature.
According to the simulation in HYSYS, the total shaft power demand is 25.93 MW, which
is within 0.5% of the value reported, 26.05 MW (Zheng, 2009). The shaft power demand in
the CryoMan process, obtained from the simulation in HYSYS, represents a specific shaft
power consumption of 0.2997 kWh·kg-1
of LNG. The minimum temperature approach in
the MSHE is 4.9°C in the simulation, compared to 5.0°C reported by Zheng (2009). The
simulation of the CryoMan process in HYSYS may thus be seen to be in good agreement
with the reported data, and the natural gas stream is used to evaluate the novel refrigeration
cycles.
Total shaft power for compression (MW) is used during the sensitivity analyses and
examples in this chapter; the specific shaft power demand (kWh·kg-1
of LNG) is used to
compare the novel refrigeration cycles at the best conditions found through sensitivity
analyses.
Chapter 4 Evaluation of the Novel Refrigeration Cycles
99
Table 4. 4. The CryoMan process (Zheng, 2009): inputs to HYSYS using the natural gas stream of Table 4.2.
Label in Figure 4.4 Degree of freedom Input Value
A Refrigerant flow rate [kmol·s-1
] 3.2
A Composition [mole fraction]
- Methane 0.2288
- Ethane 0.3703
- Propane 0.1684
- n-Butane 0.1517
- Nitrogen 0.0808
A Discharge pressure [bar] 48.3
B Vapour flow rate fraction 0.883
C Liquid flow rate fraction 0.223
LP Stream HP Stream
D Precooling temperature [°C] –164.6 –79.0
E Expansion pressure [bar] 1.2 9.6
F MSHE outlet temperature [°C] 21.2 24.3
Performance indicators
Number of compression stages 4
ΔTMIN [°C] 4.9
Total shaft power [MW] 25.93
Specific shaft work [kWh·kg-1
LNG] 0.2997
Figure 4. 4. The CryoMan process.
4.3 Sensitivity analyses: Manipulation of the degrees of freedom in the novel
refrigeration cycles
Each novel refrigeration cycle is initially simulated with the same values of the operating
variables as those in the CryoMan process (see Table 4.4 and Figure 4.4). The values of the
new degrees of freedom for the initial simulations are presented in Section 4.4.1 for the
Bypass design, in Section 4.5.1 for the Two Flash Levels design, and in Section 4.6.1 for
the Mixing After Precooling design.
In order to keep consistency with the CryoMan process, a minimum temperature approach
of 5°C in the MSHE is defined for feasible heat transfer between hot and cold streams; the
values for the new degrees of freedom in each novel design are thus selected to achieve a
Chapter 4 Evaluation of the Novel Refrigeration Cycles
100
feasible simulation. Table 4.5 summarises other modelling assumptions and constraints
which are considered in this work in order to maintain consistency with Zheng (2009). The
Peng–Robinson equation of state is used to calculate physical and thermodynamic
properties of the refrigerant mixture (such as enthalpies, entropies, temperatures, phase
equilibrium, etc.).
Table 4. 5. Assumptions in the novel refrigeration cycles to maintain consistency with Zheng (2009).
Compressor MSHE Throttle Valves Condenser Flash Unit Streams
Mixing
Multistage
centrifugal ΔTMIN = 5°C
Isenthalpic
expansion
Refrigerant leaves
at 30°C
Zero pressure
drop Isobaric
Efficiency
ηIS = 80% Zero pressure drop - Zero pressure drop - -
Intercooling
at 30°C - - - - -
As the operating variables of the novel refrigeration cycles are manipulated, the minimum
temperature approach in the MSHE is calculated in HYSYS by comparing a user-specified
number of intermediate temperature values from the hot composite curve against the
corresponding values in the cold composite curve; each temperature comparison between
the hot and cold composite curves is evaluated at the same enthalpy value. Because each
variable is manipulated independently of the remaining variables (i.e. interactions between
the variables are not considered), the constraint of 5°C as minimum temperature difference
between hot and cold streams in the MSHE would reduce significantly the range in which
each variable can be manipulated. Thus, a minimum temperature approach range of
5±0.3°C is considered during the sensitivity analyses.
Sensitivity analyses are performed within HYSYS for the novel refrigeration cycles for
each degree of freedom except for the refrigerant composition, which is manipulated using
MATLAB (see Section 4.3.1). While the selected operating variable is varied, the
remaining degrees of freedom are held constant.
4.3.1 Manipulation of the refrigerant composition
The composition of the refrigerant is complicated to manipulate and to analyse in the
sensitivity analyses. For example, if the mole fraction of a component in the mixture is
decreased, the remaining mole fractions must also be modified to ensure the sum of mole
fractions is unity. Different approaches can be used to manipulate the composition. For
instance, when increasing the mole fraction of one of the components, the reduction of the
remaining proportion can be distributed on an equal share among the other components, or
Chapter 4 Evaluation of the Novel Refrigeration Cycles
101
the mole fraction of the remaining components can be changed while keeping their initial
proportions constant. The resulting compositions for the sensitivity analyses, and their
corresponding impact on the refrigeration cycle performance, depend on the method of
adjusting the composition.
The method selected to vary the composition of the refrigerant during the sensitivity
analyses is to vary the mole fraction of one component while keeping constant the initial
proportions of the remaining components in the refrigerant mixture. This method ensures
that when a component, for example propane, is increased, only its mole faction is
increased and the relative contributions of the remaining components stay constant (see
Figure 4.5). A code was developed in MATLAB (The MathWorks Inc., 2013) to calculate
the refrigerant composition, and linked to Aspen HYSYS to input the refrigerant
composition to the simulation. The initial mole fraction of the components is first input in
the code. Equation 4.2 is employed to calculate the initial proportion of the ith component
in the refrigerant mixture. The reference component in the refrigerant mixture is varied
from 0.05 to 0.95 on a mole fraction basis (xref). As the mole fraction of xref is manipulated,
the mole fraction of the remaining components (xi) is calculated with Equation 4.3
according to their corresponding proportion in the initial composition. The composition of
the optimised CryoMan process (see Table 4.4) is used to calculate the initial proportions
of the components in the refrigerant mixture, which allow calculating the overall
composition of the refrigerant while each component is manipulated.
φ𝑖 =𝑥𝑖
𝑖𝑛𝑖𝑡𝑖𝑎𝑙
1−𝑥𝑟𝑒𝑓𝑖𝑛𝑖𝑡𝑖𝑎𝑙 ; 𝑥𝑖 , 𝑥𝑟𝑒𝑓 ∈ 𝑋𝑀𝑅 (4.2)
𝑥𝑖𝑐𝑎𝑙𝑐 = [1 − 𝑥𝑟𝑒𝑓
𝑣𝑎𝑟𝑖𝑒𝑑] ∙ φ𝑖 (4.3)
where φi = Proportion of the ith component in the initial refrigerant composition
xref = Mole fraction of the reference component
xi = Mole fraction of the ith component
XMR
= Vector for the composition of the overall refrigerant composition
initial = Indicates that the mole fraction of a component is from the initial composition
calc = Indicates the mole fraction of a component, calculated after the reference component is varied
varied = Indicates the mole fraction of the reference component as is varied from 0.05 to 0.95
Chapter 4 Evaluation of the Novel Refrigeration Cycles
102
Figure 4. 5. Example of composition manipulation: propane mole percentage is increased; the mole
percentage proportion between the remaining components remains the same.
4.4 Evaluation of the Bypass design
4.4.1 Bypass design: Initial simulation
The Bypass design, together with the values of the new degrees of freedom is shown in
Figure 4.6. As mentioned in Section 4.3, the values of the remaining degrees of freedom
are those of the optimised CryoMan process, and are shown in Table 4.4. The values of the
new degrees of freedom are chosen in order to achieve a feasible simulation, i.e. a
minimum temperature approach of 5°C between the composite curves in the MSHE. As
shown in Figure 4.6, the flow rate of the refrigerant is increased (3.75 kmol·s-1
), compared
to 3.21 kmol·s-1
in the CryoMan process, because the Bypass Stream reduces by 15% the
flow rate fed to the flash unit. These values are the initial conditions used to evaluate the
Bypass design through sensitivity analyses.
Figure 4. 6. The Bypass design showing the new degrees of freedom and initial values.
23.6%
37.9%
15.0%
15.4%
8.2%
Methane Ethane Propane n-Butane Nitrogen
22.2%
35.7%
20.0%
14.5%
7.7%
Chapter 4 Evaluation of the Novel Refrigeration Cycles
103
The initial simulation resulted in shaft power demand of 26.81 MW (0.3099 kWh·kg-1
of
LNG) for refrigerant compression, which is 3.4% higher than that of the CryoMan process.
The minimum temperature approach between the composite curves in the MSHE is 5.3°C.
4.4.2 Bypass design: Sensitivity studies and discussion
The bypass flow rate fraction is varied and the effect on the shaft power demand
(diamonds) as well as on the minimum driving force in the MSHE (squares) is displayed in
Figure 4.7a. The shaft power demand decreases from 29 MW to 26 MW as the bypass flow
ratio is varied from 0.05 to 0.19.
As the flow rate in the Bypass Stream increases, the flow rate fed to the flash unit
decreases. Thus, increasing the Bypass Stream (which is compressed from 20 bar) implies
reducing the flow rate of Stream 1, which is compressed from 1.2 bar. A greater flow is
compressed from only 20 bar to 48.3 bar as the bypass flow rate fraction increases, and the
shaft power consumption decreases accordingly. However, Stream 1 is the only cold
stream that provides cooling in the temperature range from –170°C to –95°C (see Figure
4.7b), and as the flow rate fraction of the Bypass Stream is increased further (above 0.17),
the flow rate of Stream 1 (1.73 kmol·s-1
) fails to provide cooling to the hot streams in the
MSHE while maintaining a minimum temperature difference of 5°C. Thus, increasing the
bypass flow rate fraction reduces the flow rate of Stream 1 to bring shaft power savings but
decreasing the flow rate of Stream 1 may also result in infeasible heat transfer inside the
MSHE (i.e. minimum temperature approach between hot and cold streams less than 5°C).
Figure 4. 7. Effect of increasing the Bypass Stream flow rate fraction: a) shaft power demand and minimum
driving force in the MSHE; b) the composite curves in the MSHE.
Increasing the pressure level of the Bypass Stream also reduces the shaft power
consumption of the refrigeration cycle (see diamonds in Figure 4.8a). As the pressure level
of the Bypass Stream is increased from 15 bar to 25 bar, the shaft power demand decreases
Chapter 4 Evaluation of the Novel Refrigeration Cycles
104
from 28.1 MW to 26.3 MW. This decreasing trend in the shaft power demand is expected
because, as illustrated in the pressure–enthalpy diagram of a refrigeration cycle of Figure
4.8b, the shaft power required for refrigerant compression decreases as the expansion
pressure (refrigerant pressure level) increases, for a fixed compressor discharge pressure
considering the same flow rate and composition of the refrigerant and the same
compression efficiency. The minimum temperature difference between the composite
curves in the MSHE is unaffected by the variation of the pressure level of the Bypass
Stream (squares in Figure 4.8a). This null effect on the minimum driving force indicates
that the minimum temperature approach of the composite curves in the MSHE is not within
the temperature range in which the Bypass Stream provides cooling (i.e. between –95°C
and 7°C according to Figure 4.6). According to the sensitivity analyses, higher pressure
levels of the Bypass Stream (compared to the initial pressure level, 20 bar) would help to
reduce the shaft power demand in the refrigeration cycle.
Figure 4. 8. Effect of increasing the pressure level of the Bypass Stream: a) shaft power demand and
minimum driving force in the MSHE; b) P–H diagram.
The effect of the compressor discharge pressure on the shaft power for refrigerant
compression (diamonds) and minimum driving force in the MSHE (squares) is shown in
Figure 4.9a. The shaft power required for refrigerant compression hardly changes (increase
from 26.76 MW to 26.81 MW) as the discharge pressure increases from 43 bar to 49 bar.
As the discharge pressure is increased further, from 49 bar to 52 bar in Figure 4.9a
(diamonds), the shaft power consumption decreases to 26.79 MW. Increasing the
compressor discharge pressure significantly reduces the minimum driving force in the
MSHE (squares in Figure 4.9a). As the compressor discharge pressure increases, more
liquid refrigerant is obtained after the condenser; this effect of pressure on the vapour
fraction of the refrigerant at the condenser temperature is illustrated with an isotherm in a
P–H diagram in Figure 4.9b. Increasing the pressure reduces the vapour fraction of the
Chapter 4 Evaluation of the Novel Refrigeration Cycles
105
refrigerant, i.e. the liquid fraction of the refrigerant increases. The flow rate of Stream 1 is
consequently reduced. Beyond 50 bar, the flow rate of Stream 1 (1.71 kmol·s-1
) is not able
to provide cooling to the hot streams in the MSHE while meeting the minimum
temperature difference constraint (see squares in Figure 4.9a). The sensitivity analyses thus
suggest that the compressor outlet pressure should decrease, compared to the base case
(48.3 bar), to avoid infeasible heat transfer in the MSHE.
Figure 4. 9. Effect of increasing the compressor discharge pressure: a) power demand and minimum driving
force; b) the vapour fraction of the refrigerant.
The flow rates of Stream 1 and Stream 2 are also controlled by the flow rate fractions of
the vapour (f Vap
) and liquid (f Liq
) that are partially mixed after the flash unit (labelled ‘B’
and ‘C’ in Figure 4.6, respectively). Both variables indicate the fraction of their
corresponding flow rates that is mixed to create Stream 1. Figure 4.10 shows how the flow
rates of Stream 1 and Stream 2 (diamonds and squares in Figure 4.10b, respectively)
change as f Liq
is increased from 0.15 to 0.35, which leads to increasing the shaft power
demanded by the refrigeration cycle (diamonds in Figure 4.10a). As the liquid flow rate
fraction increases from 0.15 to 0.35, the shaft power demand increases from 26 MW to
28.2 MW. Increasing f Liq
increases the flow rate of Stream 1 (which is at the lowest
pressure level), leading to increase in the shaft power demand. As the value of f Liq
is
decreased below 0.19, the flow rate of Stream 1 (1.68 kmol·s-1
) cannot provide cooling to
the hot streams in the MSHE with a minimum temperature difference of at least 5°C (see
squares in Figure 4.10a).
Chapter 4 Evaluation of the Novel Refrigeration Cycles
106
Figure 4. 10. Effect of increasing the flow rate fraction of liquid from the flash unit (f
Liq): a) shaft power
demand and minimum driving force in the MSHE; b) flow rates of Stream 1 and Stream 2.
Additionally, as f Liq
increases, the composition of Stream 1 yields a higher specific heat of
vaporisation (see Figure 4.11a). That is, the heat required to fully vaporise 1 kmol of the
refrigerant increases as the amount of liquid that is fed to Stream 1 increases. The heat of
vaporisation is a function of the composition of the refrigerant, at a fixed pressure, and
increases as the composition of the stream becomes richer in ‘heavy’ components (i.e.
propane and n-butane), which have greater heat of vaporisations compared to ‘light’
components (nitrogen, methane and ethane), as exemplified in Figure 4.11b.
Figure 4.11b shows the heat of vaporisation of the streams involved in a flash separation in
order to have three streams (inlet refrigerant, vapour phase and liquid phase) with different
compositions. According to Figure 4.11b, less refrigerant flow would be required, to
provide the same cooling duty, with the liquid phase stream since its heat of vaporisation is
greater than that of the inlet refrigerant and of the vapour phase stream. Thus, increasing
the heat of vaporisation of Stream 1 may help to reduce the overall refrigerant flow rate.
Reducing the overall refrigerant flow rate would reduce the shaft power consumption.
Consequently, a trade-off exists between decreasing the value of f Liq
to reduce of flow rate
of Stream 1, which would bring shaft power savings, and increasing the value of f Liq
to
modify the composition of Stream 1, increasing its heat of vaporisation and reducing the
overall refrigerant flow rate to bring shaft power savings. The sensitivity analyses suggest
that decreasing f Liq
would rather take advantage of reducing the flow rate of Stream 1,
leading to shaft power savings in the refrigeration cycle (see diamonds in Figure 4.10a).
Chapter 4 Evaluation of the Novel Refrigeration Cycles
107
Figure 4. 11. a) Increasing the value of f
Liq increases the heat of vaporisation of Stream 1; b) heat of
vaporisation of the refrigerant is increased as heavy components in the composition are increased.
The effect of the overall composition of the refrigerant on the power demand is shown for
each component in Figure 4.12 as the line with diamonds in each plot. The descending
trends of the power demand observed as the proportion of propane (Figure 4.12c) and n-
butane (Figure 4.12d) increases agrees with the discussion in Section 3.2.1; in this section
it was seen that the shaft power for compression decreases as the mole fraction of the
heavier components (i.e. propane and n-butane, compared to ethane, methane and nitrogen)
increases. As the mole fraction of methane (Figure 4.12a), ethane (Figure 4.12b) and
nitrogen (Figure 4.12e) increases, the shaft power for compression increases accordingly,
although the shaft power demand is less sensitive to changes in the mole fraction of ethane.
For example, the power demand increases by 2 MW as the mole fraction of methane
increases by 0.05, whereas the shaft power increases by only 0.1 MW as the mole fraction
of ethane increases by 0.05. However, the mole fraction of each component can be varied
only within a limited range, in order to meet the constraint of minimum temperature
approach inside the MSHE.
The minimum temperature approach between the composite curves inside the MSHE is
also represented in each plot of Figure 4.12 (squares). As can be seen in Figure 4.12, the
range of compositions for which heat transfer is feasible (i.e. minimum temperature
approach inside the MSHE of 5°C) is highly constrained. For example, in Figure 4.12a
feasible heat transfer is potentially achieved only with the compositions obtained when the
mole fraction of methane in the refrigerant mixture is greater than 0.20 but less than 0.30.
Table 4.6 shows the mole fraction ranges of each component in which the composition
could result in feasible heat transfer according to the sensitivity analyses.
The composition of the overall refrigerant stream determines the flow rate and composition
of the flash unit product streams (at fixed compressor discharge pressure and condenser
Chapter 4 Evaluation of the Novel Refrigeration Cycles
108
temperature). Because the mixing flow rate fractions of vapour and liquid, and the flow
rate fed to the flash unit are held constant, the overall composition determines the actual
molar flow rates and compositions of Stream 1 and Stream 2. Increasing the flow rate of
Stream 1 reduces the flow rate of Stream 2, and vice versa. Thus, as the composition of the
overall refrigerant stream is varied, infeasible heat transfer comes from reducing either
Stream 1 or Stream 2.
Figure 4. 12. Effect of composition on power demand and minimum driving force in the Bypass design:
a) methane; b) ethane; c) propane; d) n-butane; e) nitrogen.
Table 4. 6. Composition ranges for feasible heat transfer in the Bypass design.
Lower bound [mole fraction] Upper bound [mole fraction]
Methane 0.20 0.30
Ethane 0.30 0.40
Propane 0.10 0.20
n-Butane 0.05 0.20
Nitrogen 0.05 0.15
As the mole fraction of the light components (methane, ethane and nitrogen) increases, two
effects take place: i) the vapour fraction of the refrigerant mixture at the condenser
Chapter 4 Evaluation of the Novel Refrigeration Cycles
109
temperature increases (illustrated in Figure 4.13a), and ii) the heat of vaporisation of the
refrigerant decreases (previously illustrated with Figure 4.11b). As the mole fraction of the
light components increases, the molar flow rate of vapour leaving the flash unit increases,
which leads to increasing the flow rate of Stream 1 and reducing the molar flow rate of
Stream 2 (since the flow rate fed to the flash unit and the vapour and liquid flow rate
fractions partially mixed are constant). Increasing the molar flow rate of Stream 1 leads to
increasing compression shaft power. Additionally, for example, when methane represents
over 30% mole of the refrigerant mixture, the molar flow rate of Stream 2 is not able to
provide cooling to the hot streams in the MSHE with a minimum temperature difference of
5°C (according to Table 4.6).
As the mole fraction of methane increases, the specific heat of vaporisation of the overall
refrigerant stream decreases. Thus, a refrigerant with a low specific heat of vaporisation
would require a greater molar flow rate to provide the same cooling duty compared to a
refrigerant with a high heat of vaporisation, as previously discussed and illustrated with
Figure 4.11b. However, a refrigerant with light components can provide cooling at colder
temperatures than a refrigerant with heavy components, at the same pressure level, because
light components evaporate at lower temperatures (see Figure 4.13b).
On the other hand, as mole fraction of the heavy components (propane, n-butane) increase
in the refrigerant mixture, the opposite effects occur compared to the increase of light
components mole fraction: for instance, when n-butane represents over 20% mole of the
refrigerant mixture (according to Table 4.6), the liquid condensed and separated in the
flash unit increases; the flow rate of Stream 1 decreases to bring shaft power savings but is
not able to provide cooling in the MSHE with a minimum driving force of 5°C. This
infeasible heat transfer occurs despite the increase of the specific heat of vaporisation in
the refrigerant stream.
Thus, a complex trade-off exists in the selection of the overall refrigerant composition:
heavy components can provide larger cooling duties per unit mole than light components,
which would reduce the overall refrigerant flow rate and bring shaft power savings;
however, light components provide cooling at colder temperatures (compared to heavy
components) at the same pressure level, but at the expense of increasing the shaft power
demand since light compositions require more shaft power for compression than heavy
compositions (see Section 3.2.1).
Chapter 4 Evaluation of the Novel Refrigeration Cycles
110
Figure 4. 13. As the overall refrigerant composition becomes lighter: a) its vapour fraction increases; and
b) its evaporating temperature decreases.
4.4.3 Bypass design: Manipulation of its degrees of freedom
According to the trends and trade-offs observed in the sensitivity analyses previously
discussed, the degrees of freedom in the Bypass design are then manipulated in order to
reduce the shaft power consumption for refrigerant compression. The initial values of the
degrees of freedom are those of the initial simulation (see Section 4.4.1). As mentioned in
Section 4.3, a minimum temperature approach of 5±0.3°C is considered for feasible heat
transfer during the manipulation of the degrees of freedom.
For example, Figure 4.14a shows the manipulation of the Bypass Stream flow rate fraction,
which is increased from 0.15, in the initial simulation, to 0.17, and the shaft power demand
is reduced from 26.8 MW to 26.4 MW while the minimum temperature difference in the
MSHE is reduced from 5.3°C to 4.8°C. Increasing the Bypass Stream flow rate fraction
takes advantage of reducing the flow rate of Stream 1 from 1.77 kmol·s-1
to 1.73 kmol·s-1
.
In Figure 4.14b, the fraction of the liquid stream leaving the flash unit and mixed to create
Stream 1 (f Liq
) is manipulated after the value of the Bypass Stream flow rate fraction is
changed. The value of f Liq
is decreased from 0.22 to 0.20, which further reduces the flow
rate of Stream 1 (to 1.69 kmol·s-1
) and the shaft power demand (to 26.1 MW). The
minimum temperature approach in the MSHE remains as 4.8°C.
Chapter 4 Evaluation of the Novel Refrigeration Cycles
111
Figure 4. 14. Manipulation of the operating variables in the Bypass design: a) Bypass Stream flow rate
fraction is varied; b) the liquid mixing fraction (f Liq
) is varied.
Each of the operating variables of the Bypass design is manipulated in order to reduce the
shaft power demand, as illustrated with Figure 4.14, and the values that yield the lowest
shaft power demand are shown in Table 4.7. The ‘Label’ column in Table 4.7 refers to the
letters in the Bypass design shown in Figure 4.6. The shaft power demand is 25.68 MW
(0.2968 kWh·kg-1
of LNG), i.e. a reduction of 1% compared to the CryoMan process. The
minimum temperature approach between the composite curves in the MSHE is 5.1°C.
Table 4. 7. Operating variables of the Bypass design after the sensitivity analyses.
Label in Figure 4.6 Process variable Value
A Refrigerant flow rate [kmol·s-1
] 3.65
A Composition [mole fraction]:
- Methane 0.2569
- Ethane 0.3552
- Propane 0.1650
- n-Butane 0.1547
- Nitrogen 0.0682
A Discharge pressure [bar] 41
A Bypass fraction 0.13
B Vapour flow rate fraction 0.80
C Liquid flow rate fraction 0.08
Stream 1 Stream 2 Bypass
D Precooling temperature [°C] –163 –68 –73
E Expansion pressure [bar] 1.2 7.5 20
F MSHE outlet temperature [°C] 15 25 24
Performance indicators
Number of compression stages 4
ΔTMIN [°C] 5.1
Total shaft power [MW] 25.68
Specific shaft power [kWh·kg-1
LNG] 0.2968
Shaft power savings (%) 1.0
Even though the mole fraction of methane (light component) in the overall refrigerant
stream is increased, the results in Table 4.7 suggest that the Bypass design takes advantage
of reducing the flow rate of Stream 1 by creating the Bypass Stream to bring shaft power
savings. The overall refrigerant flow rate is reduced (3.75 kmol·s-1
in the initial simulation
compared to 3.65 kmol·s-1
in Table 4.7); the flow rate fraction of the liquid stream leaving
Chapter 4 Evaluation of the Novel Refrigeration Cycles
112
the flash unit (f Liq
), mixed to create Stream 1, is decreased from 0.22 (in the initial
simulation) to 0.08, while the bypass flow rate fraction is also decreased from 0.15 (initial
simulation) to 0.13. The Bypass design appears to offer advantages compared to the
CryoMan process.
Further power savings are likely to be achieved in the Bypass design if its operating
variables are optimised. Therefore, the Bypass design is considered for optimisation using
WORK software, as presented in Chapter 5.
4.5 Evaluation of the Two Flash Levels design
4.5.1 Two Flash Levels design: Initial simulation
The Two Flash Levels design and the initial values of the new degrees of freedom are
shown in Figure 4.15. These initial values are chosen in order to achieve a feasible
simulation (i.e. the minimum temperature difference between hot and cold streams inside
the MSHE is 5°C). The values of the remaining degrees of freedom are those of the
CryoMan process (see Section 4.3).
Note that the order in which the refrigerant streams are fed to the compressor, according to
Figure 4.15, does not represent the actual arrangement of the refrigerant streams in the
compressor. The order of the refrigerant streams shown in Figure 4.15 is only for clarity in
labelling the streams and values of the operating variables of the new degrees of freedom
in the Two Flash Levels design.
Figure 4. 15. The Two Flash Levels design showing the new degrees of freedom and initial values.
Chapter 4 Evaluation of the Novel Refrigeration Cycles
113
According to the initial simulation of the Two Flash Levels design, the resulting shaft
power demand is 26.78 MW (0.3096 kWh·kg-1
of LNG), which is 3.3% greater than that of
the CryoMan process; the minimum driving force for heat transfer in the MSHE is 5.2°C.
Similar to the Bypass design, the refrigerant flow rate increases compared to the CryoMan
process (by nearly 1.3%, to 3.25 kmol·s-1
) since 30% of the liquid produced from the first
flash unit is fed to the second flash separator (initial value for the simulation) and
consequently the minimum temperature approach in the MSHE would be less than 5°C for
the initial flow rate.
Note that, in the CryoMan process (Figure 4.4), the LP Stream (at 1.2 bar) is compressed in
two compression stages to the pressure level of the HP Stream (9.6 bar) (the maximum
pressure ratio is 3), which leads to one intermediate pressure level between the pressure
levels of the LP Stream and the HP Stream (see Figure 4.16a). In the initial conditions of
the Two Flash Levels design, the pressure level of the new two streams (2.4 bar for Stream
3 and 4.8 bar for Stream 4) are accommodated between the pressure levels of Stream 1 and
Stream 2 (1.2 bar and 9.6 bar, respectively), as illustrated in Figure 4.16b. Consequently,
the compression of Stream 1 (at 1.2 bar) to the pressure level of Stream 2 (9.6 bar) in the
Two Flash Levels design is performed in three compression stages. Therefore, the overall
compression of the refrigerant takes place with 5 compression stages, compared to 4
compression stages in the CryoMan process. Since after each compression stage the
refrigerant is cooled down to 30°C with an intercooler (as illustrated in Figure 4.16) that
helps reducing the shaft power demand, the Two Flash Levels design would take
advantage of the additional compression stage and associated intercooling stage. However,
this additional compression stage would also increase the capital costs of the Two Flash
Levels design.
Chapter 4 Evaluation of the Novel Refrigeration Cycles
114
Figure 4. 16. Streams arrangement in the multistage compressor: a) CryoMan process;
b) Two Flash Levels design.
4.5.2 Two Flash Levels design: Sensitivity studies and discussion
Similar to the Bypass design, the variables of the Two Flash Levels design are evaluated in
terms of their effect on shaft power demand and minimum temperature approach between
the composite curves in the MSHE. Trends of the operating variables are presented and
discussed, where the aim is to reduce the shaft power demand in the novel refrigeration
cycle.
The effect of the flow rate fraction that is fed to the second flash unit (f 2nd
), on the shaft
power demand (diamonds) of the refrigeration cycle and minimum temperature approach
in the MSHE (squares), is shown in Figure 4.17a. The shaft power demand increases from
26 MW to 27.9 MW as the flow rate fraction fed to the second flash unit increases from
0.15 to 0.50 (Figure 4.17a). Since the flow rate fraction of liquid from the first flash unit (f
Liq), mixed to create Stream 1, is held constant, increasing f
2nd reduces the flow rate of
Stream 2 to create the new streams after the second flash unit (Stream 3 and Stream 4) (see
Figure 4.17b). The pressure levels of Stream 3 and Stream 4 are 2.4 bar and 4.8 bar,
respectively, whereas the pressure level of Stream 2 is 9.6 bar. Thus, as fraction f 2nd
is
increased, the refrigerant flow rate at lower pressure levels than that of Stream 2 (9.6 bar)
is increased accordingly. As previously illustrated in Figure 4.8b, the shaft power demand
is increased as the difference between the refrigerant pressure level and the compressor
Chapter 4 Evaluation of the Novel Refrigeration Cycles
115
discharge pressure is increased. Thus, reducing f 2nd
would decrease the shaft power for
refrigerant compression. However, as f 2nd
is decreased below 0.20, the cold streams in the
MSHE fail to provide cooling to the hot streams with a minimum temperature difference of
5°C in the MSHE (Figure 4.17a), which indicates that the minimum temperature difference
between the composite curves in the MSHE is in the temperature range in which Stream 3
and Stream 4 provide cooling (i.e. from –63°C to 7°C). So, reducing f 2nd
(below 0.30)
would help reducing the total shaft power demand because the flow rate at a high pressure
level (Stream 1 at 9.6 bar) is increased compared to that at pressure levels of Stream 3 and
Stream 4 (2.4 bar and 4.8 bar, respectively), but values below 0.20 may lead to infeasible
heat transfer within the MSHE.
Figure 4. 17. Effect of increasing the flow rate fraction of liquid fed to the second flash unit (f
2nd): a) shaft
power demand and minimum driving force in the MSHE; b) flow rate of Stream 2.
The pressure in the second flash unit (P2nd
in Figure 4.15) is varied from 15 bar to 35 bar;
the effect on the power demand (diamonds) and minimum temperature approach in the
MSHE (squares) is presented in Figure 4.18a. The variation of the shaft power demand is
negligible (diamonds in Figure 4.18a), decreasing by 0.05 MW over the whole pressure
range. The minimum temperature difference between hot and cold streams in the MSHE
(squares in Figure 4.18a) decreases as the pressure in the second flash unit is increased.
The pressure in the second flash unit determines the vapour-liquid equilibrium of the
refrigerant; the temperature of the refrigerant before the throttle valve of the second flash
unit is assumed to be that of the condenser, and the pressure and composition of the
refrigerant before the throttle valve are held constant. The pressure in the second flash unit
thus determines the flow rate of vapour and liquid leaving the second flash unit and their
corresponding composition.
Because Stream 3 and Stream 4 are created by partially mixing the vapour and liquid from
the second flash unit, the pressure of the second flash unit would then affect the flow rate
Chapter 4 Evaluation of the Novel Refrigeration Cycles
116
of Stream 3 and Stream 4 and their corresponding composition. For example, Figure 4.18b
shows the effect of the pressure in the second flash unit on the flow rate (diamonds) and
heat of vaporisation, ΔHvap (squares, related to the composition), of Stream 3. As the
pressure increases from 15 bar to 35 bar, the heat of vaporisation of Stream 3 is increased
by 1.6% and, according to the discussion of Figure 4.11b in Section 4.4.2, would help to
reduce the overall flow rate fed to the second flash unit to bring shaft power savings. On
the other hand, as P2nd
is increased, the vapour fraction of the refrigerant stream fed to the
second flash is reduced, decreasing the flow rate of Stream 3 by 8%. Consequently, the
minimum temperature approach in the MSHE is decreased from 5.3°C to 5.0°C (squares in
Figure 4.18a). Thus, as P2nd
increases, there is a trade-off between increasing the heat of
vaporisation of Stream 3 or Stream 4 to reduce the refrigerant flow rate in the second flash
unit, and reducing the flow rate of Stream 3 or Stream 4 leading to infeasible heat transfer.
Figure 4. 18. Effect of increasing the pressure of the second flash unit (P
2nd): a) power demand and minimum
driving force in the MSHE; b) heat of vaporisation and flow rate of Stream 3.
Figure 4.19a shows the effect of the precooling temperature of Stream 3 on the power
demand (diamonds) and minimum temperature approach in the MSHE (squares). As the
precooling temperature of Stream 3 is decreased, the flow rate of the refrigerant would be
expected to increase because the cooling duty in the MSHE would increase (and so would
the shaft power demand) in order to maintain a constant minimum temperature difference
between hot and cold streams in the MSHE. However, the flow rate of the refrigerant is
held constant during the sensitivity analyses and therefore, the minimum temperature
difference in the MSHE would vary and indicate the temperature range in which Stream 3
can be precooled with the current refrigerant flow rate (3.25 kmol·s-1
) with feasible heat
transfer, and the shaft power demand would remain constant.
In Figure 4.19a (squares), the minimum temperature approach in the MSHE remains
constant at 5.2°C as the precooling temperature of Stream 3 increases from –93° to –63°C;
Chapter 4 Evaluation of the Novel Refrigeration Cycles
117
the minimum temperature approach then decreases when the precooling temperature of
Stream 3 is further increased. As illustrated in Figure 4.19b, as the precooling temperature
of the mixed refrigerant stream decreases, at fixed pressure, the temperature at which the
stream provides cooling, once expanded at the throttle valve outlet pressure, decreases
accordingly. Thus, the evaporating temperature of Stream 3 increases as its precooling
temperature increases. At precooling temperatures above –55°C, Stream 3 evaporates at
temperatures above –75°C, resulting in infeasible heat transfer at around –80°C, (see
Figure 4.19c). To avoid infeasible heat transfer inside the MSHE, Stream 3 then would
have to be precooled to temperatures below –55°C to help provide cooling at colder
temperatures. The shaft power demand (diamonds in Figure 4.19a), as previously
discussed, remain constant.
Figure 4. 19. Effect of increasing the precooling temperature of Stream 3: a) power demand and minimum
driving force in the MSHE; b) evaporating temperature; c) infeasible heat transfer in the MSHE.
The overall refrigerant composition in the Two Flash Levels design is explored according
to Section 4.3.1. Figure 4.20 displays the effect of the refrigerant composition on the shaft
power demand of the refrigeration cycle as well as on the minimum temperature approach
Chapter 4 Evaluation of the Novel Refrigeration Cycles
118
in the MSHE. Similar to the Bypass design, Figure 4.20 shows that the shaft power
demand increases when the mole fraction of the light components (i.e. methane, ethane and
nitrogen) increases, and decreases as the mole fraction of the heavy components (i.e.
propane and n-butane) increases. According to the trends of the minimum temperature
difference inside the MSHE in Figure 4.20, heat transfer is feasible potentially only within
the composition ranges shown in Table 4.8.
Figure 4. 20. Effect of composition on power demand and minimum driving force in the Two Flash Levels
design: a) methane; b) ethane; c) propane; d) n-butane; e) nitrogen.
Table 4. 8. Composition ranges for feasible heat transfer in the Two Flash Levels design.
Lower bound [mole fraction] Upper bound [mole fraction]
Methane 0.20 0.30
Ethane 0.30 0.40
Propane 0.10 0.20
n-Butane 0.10 0.20
Nitrogen 0.05 0.10
The operating variables of the Two Flash Levels design are manipulated, one variable at
the time, as exemplified in Section 4.4.2 for the Bypass design, according to the sensitivity
Chapter 4 Evaluation of the Novel Refrigeration Cycles
119
analyses and trends observed. As mentioned in Section 4.3, the minimum temperature
approach for heat transfer between hot and cold streams inside the MSHE is considered
between 4.7°C and 5.3°C. After the operating variables are manipulated in the Two Flash
Levels design, the values of the variables that result in the lowest shaft power demand for
refrigerant compression are shown in Table 4.9. The ‘Label’ column in Table 4.9 refers to
the letters in the Two Flash Levels design shown in Figure 4.15. The shaft power obtained
is 25.93 MW (0.2997 kWh·kg-1
of LNG), which is the same as that in the CryoMan
process. The minimum temperature approach in the MSHE is 4.8°C.
Table 4. 9. Operating variables of the Two Flash Levels design after the sensitivity analyses.
Label in Figure 4.15 Process variable Value
A Refrigerant flow rate [kmol·s-1
] 3.45
A Composition [mole fraction]:
- Methane 0.2332
- Ethane 0.4000
- Propane 0.0503
- n-Butane 0.2397
- Nitrogen 0.0768
A Discharge pressure [bar] 41
H Pressure 2nd flash unit [bar] 30
G Liquid fraction to 2nd flash unit 0.224
Stream 1 Stream 2 Stream 3 Stream 4
D Precooling temperature [°C] –163 –78 –43 –53
E Expansion pressure [bar] 1.2 7.5 5 18.5
F MSHE outlet temperature [°C] 17 25 25 20
B, I Vapour flow rate fraction 0.74 0.26 0.10 0.90
C, J Liquid flow rate fraction 0.25 0.526 0.20 0.80
Performance indicator
Number of compression stages 5
ΔTMIN [°C] 4.8
Total shaft power [MW] 25.93
Specific shaft power [kWh·kg-1
LNG] 0.2997
Shaft power savings (%) 0.0
The Two Flash Levels design thus, according to the results shown in Table 4.9, takes
advantage of decreasing the flow rate fraction fed to the second flash unit and increasing
the pressure of the second flash unit to manipulate the heat of vaporisation of Stream 3 and
Stream 4. Moreover, the pressure level of Stream 4 is increased from 4.8 bar to 18.5 bar, to
help reduce the refrigerant flow rate that is compressed at low pressure levels (i.e. 1.2 bar
of Stream 1 and 5 bar of Stream 3). Also, note that despite the increase in the pressure level
of Stream 4 from 4.8 bar to 18.5 bar, the overall refrigerant in the Two Flash Levels design
is still compressed in 5 compression stages. As discussed in Section 4.5.1, the additional
compression stage, compared to the CryoMan process, helps to reduce the shaft power
demand because of the associated intercooling stage, although at the expense of increasing
the capital costs.
Chapter 4 Evaluation of the Novel Refrigeration Cycles
120
Thus, according to Table 4.9, the Two Flash Levels design is very likely to achieve shaft
power savings, compared to the CryoMan process, once its operating variables are
optimised. Therefore, the Two Flash Levels design is considered for optimisation in
Chapter 5.
4.6 Evaluation of the Mixing After Precooling design
4.6.1 Mixing After Precooling design: Initial simulation
The Mixing After Precooling design and the values of the new degrees of freedom are
shown in Figure 4.21. The values of the remaining degrees of freedom are those of the
CryoMan process (see Section 4.3). Similar to the initial simulation of the previous novel
refrigeration cycles, the initial values of the new degrees of freedom in the Mixing After
Precooling design are chosen in order to achieve a feasible simulation.
Note that, according to Figure 4.21, Stream 1 is expanded to an intermediate pressure
(shown in Figure 4.21 as Stream 3) whereas Stream 2 is expanded to the lowest pressure
level (shown in Figure 4.21 as Stream 4). Thus, Stream 1 in the Mixing After Precooling
design is simulated with the conditions (pressure level, precooling temperature,
composition, etc.) of the HP Stream in the CryoMan process, and Stream 2 is simulated
with the conditions of the LP Stream in the CryoMan process. That is, Stream 1 and
Stream 2 in the Mixing After Precooling design are in the opposite pressure order
compared to Stream 1 and Stream 2 in the Bypass design and in the Two Flash Levels
design.
Stream 5 is created from the mixing of the flow rate fraction α (from Stream 1) and the
flow rate fraction β (from Stream 2). Because the mixing of the flow rate fractions α and β
is assumed to be isobaric, Stream 5 is constrained to have the same pressure level as
Stream 4, as illustrated in Figure 4.21.
Chapter 4 Evaluation of the Novel Refrigeration Cycles
121
Figure 4. 21. The Mixing After Precooling design showing the new degrees of freedom and initial values.
For the initial simulation, the overall refrigerant flow rate is increased to 3.65 kmol·s-1
(+13.7%), compared to 3.21 kmol·s-1
in the CryoMan process; the shaft power demand is
increased accordingly, to 30.48 MW (0.3523 kWh·kg-1
of LNG), i.e. 17.6% greater than
that of the CryoMan process. The minimum temperature approach in the MSHE is 5.0°C.
4.6.2 Mixing After Precooling design: Sensitivity studies and discussion
The effect of the flow rate fraction α (from Stream 1) on the shaft power demand and
minimum temperature approach between hot and cold streams in the MSHE, is shown in
Figure 4.22a. The heat of vaporisation of Stream 5 increases as the flow rate fraction α
increases (Figure 4.22b) which indicates that the mole fraction of heavy components (i.e.
propane and n-butane) in Stream 5 increases, as previously illustrated with Figure 4.11b.
As discussed in Section 4.4.2, as the heat of vaporisation increases, the flow rate needed
for the same cooling duty decreases. Increasing the heat of vaporisation of Stream 5 helps
to reduce the overall refrigerant flow rate to decrease the shaft power demand for
compression. On the other hand, however, increasing the heat of vaporisation of Stream 5
(by increasing flow rate fraction α) implies also increasing its flow rate. Consequently, the
shaft power demand of the refrigeration cycle (diamonds in Figure 4.22a) increases
accordingly from 29.1 MW to 31 MW because the refrigerant flow rate in the lowest
pressure level increases as the flow rate of Stream 5 increases.
Furthermore, as seen in Figure 4.22a, increasing flow rate fraction α beyond 0.15 also
leads to infeasible heat transfer. As the mole fraction of the heavy components in Stream 5
increases, its evaporating temperature increases accordingly (see Figure 4.23a), at a fixed
pressure. Consequently, when the value of flow rate fraction α is increased above 0.15,
Chapter 4 Evaluation of the Novel Refrigeration Cycles
122
Stream 5 evaporates from –129°C (compared to –132°C in the initial simulation) and the
flow rate of Stream 4 fails to provide cooling in the temperature range of –171°C to –
132°C to the hot streams in the MSHE with a minimum temperature difference of 5°C (see
Figure 4.23b). Thus, a trade-off exists, as the flow rate fraction α increases, between
increasing the heat of vaporisation of Stream 5 to reduce the overall flow rate, and
increasing the flow rate in the lowest pressure level. According to the sensitivity study in
Figure 4.22a, values below 0.15 for the flow rate fraction α would help reducing the shaft
power demand, compared to the initial values of the simulation of the Mixing After
Precooling design, and would also avoid infeasible heat transfer in the MSHE.
Figure 4. 22. Effect of increasing flow rate fraction α: a) shaft power demand and minimum driving force in
the MSHE; b) heat of vaporisation of Stream 5.
Figure 4. 23. a) Evaporating temperature of Stream 5 increases as flow rate fraction α increases; b) infeasible
heat transfer as a result of the increased evaporating temperature of Stream 5.
According to the Mixing After Precooling design shown in Figure 4.21, once hot Stream 2
is precooled in the MSHE and expanded, it is split to create cold Stream 4, and cold Stream
5 with the flow rate fraction β. As the flow rate fraction β from Stream 2 is increased, the
shaft power demand (Figure 4.24a) remains constant while the minimum temperature
difference between hot and cold streams in the MSHE continuously decreases. Because
Stream 5 and Stream 4 are at the same pressure level (the lowest pressure level, 1.2 bar),
Chapter 4 Evaluation of the Novel Refrigeration Cycles
123
both are fed to the compressor at the first compression stage. The flow rate fed to the
compressor at the first compression stage would remain constant, regardless of the
refrigerant flow rate in Stream 4 and Stream 5, because the flow rate fraction α is held
constant. Further, the composition of the stream at the inlet of the compressor would also
remain unaffected because the flow rate fraction α is held constant, even though the
composition of Stream 4 and Stream 5 might be different. Thus, as the flow rate fraction β
from Stream 2 increases, the flow rate and composition resulting from mixing Stream 4
and Stream 5 at the inlet of the compressor, after providing cooling in the MSHE, remains
the same and therefore leads to a constant shaft power demand for refrigerant compression.
However, as the flow rate fraction β is increased above 0.13, the hot and cold streams in
the MSHE are not able to transfer heat with a minimum temperature difference of 5°C (see
squares in Figure 4.24a). As the flow rate fraction β increases above 0.13, the flow rate of
Stream 4 decreases and is no longer able to provide cooling with a minimum temperature
difference of 5°C, in the temperature range between –171°C and –137°C as shown in
Figure 4.24b. Values of the flow rate fraction β lower than 0.13 would then avoid
infeasible heat transfer in the MSHE.
Figure 4. 24. Effect of increasing flow rate fraction β on: a) shaft power demand and minimum driving force
in the MSHE; b) the composite curves in the MSHE.
Regarding the effect of the overall composition of the refrigerant, Figure 4.25 shows the
trends of the shaft power demand and of the minimum temperature approach in the MSHE
as each component is manipulated, as described in Section 4.3.1. The shaft power demand
increases as the mole fraction of methane or nitrogen increases, and decreases as the mole
fraction of propane or n-butane increases. According to the minimum temperature
approach trends (Figure 4.25), the mole fraction ranges of each component in which heat
transfer is potentially feasible are those listed in Table 4.10.
Chapter 4 Evaluation of the Novel Refrigeration Cycles
124
Figure 4. 25. Effect of composition on power demand and minimum driving force in the Mixing After
Precooling design: a) methane; b) ethane; c) propane; d) n-butane; e) nitrogen.
Table 4. 10. Composition ranges for feasible heat transfer in the Mixing After Precooling design. Lower bound [mole fraction] Upper bound [mole fraction]
Methane 0.20 0.30
Ethane 0.20 0.40
Propane 0.10 0.20
n-Butane 0.05 0.20
Nitrogen 0.05 0.15
The operating variables of the Mixing After Precooling design are manipulated according
to the trends and trade-offs observed in the sensitivity analyses. As mentioned in Section
4.3, during the manipulation of the operating variables of the Mixing After Precooling
design, the range for feasible heat transfer is from 4.7°C to 5.3°C. The values of the
operating variables that yield the lowest shaft power are shown in Table 4.11. The ‘Label’
column in Table 4.11 refers to the letters in the Mixing After Precooling design shown in
Figure 4.21. The shaft power is 26.58 MW (0.3073 kWh·kg-1
of LNG, 2.5% greater than
Chapter 4 Evaluation of the Novel Refrigeration Cycles
125
that of the CryoMan process) with a minimum temperature approach of 4.8°C for heat
transfer in the MSHE.
Table 4. 11. Operating variables of the Mixing After Precooling design after the sensitivity analyses.
Label in Figure 4.21 Process variable Value
A Refrigerant flow rate [kmol·s-1
] 3.27
A Composition [mole fraction]:
- Methane 0.2373
- Ethane 0.3500
- Propane 0.1754
- n-Butane 0.1548
- Nitrogen 0.0825
A Discharge pressure [bar] 47.5
G Fraction α of Stream 1 0.02
H Fraction β of Stream 2 0.01
Stream 1 Stream 2 Stream 3 Stream 4 Stream 5
D Precooling temperature [°C] –93 –166 - - -
E Expansion pressure [bar] - - 10.8 1.2 1.2
F MSHE outlet temperature [°C] - - 25 25 5
B Vapour flow rate fraction 0.16 0.84 - - -
C Liquid flow rate fraction 0.67 0.33 - - -
Performance indicator
Number of compression stages 4
ΔTMIN [°C] 4.8
Total shaft power [MW] 26.58
Specific shaft power [kWh·kg-1
LNG] 0.3073
Shaft power savings (%) –2.5
The values of the new degrees of freedom in Table 4.11 are in agreement with the trends
observed during the sensitivity analyses. Reducing both flow rate fractions α and β help to
reduce the total shaft power demand of the refrigeration cycle from 30.48 MW (in the
initial simulation) to 26.58 MW. However, Table 4.11 also suggests that the trade-offs of
the operating variables in the Mixing After Precooling design are moving towards
minimising the flow rate of Stream 5, rather than increasing the heat of vaporisation of
Stream 5 to reduce the overall refrigerant flow rate. Even though Stream 5 might benefit
from the increased heat of vaporisation (by increasing both flow rate fractions α and β), the
total shaft power would increase because the compressor would receive an increased flow
rate of refrigerant at the lowest pressure level (1.2 bar) since Stream 5 is constrained to
have the same pressure level as Stream 4.
Consequently, according to Table 4.11, the structural modification in the refrigeration
cycle shows no effect towards bringing shaft power savings to the refrigeration cycle since
the values of the flow rate fractions α and β are close to zero (0.02 and 0.01, respectively).
Therefore, the Mixing After Precooling Design is not further considered for optimisation.
Chapter 4 Evaluation of the Novel Refrigeration Cycles
126
4.7 Selection of the novel designs for optimisation
According to the sensitivity studies, the Bypass design achieved shaft power savings of
1%, compared to the CryoMan process, through the manipulation of the variables. The
Bypass design benefits from reducing the flow rate of the stream in the lowest pressure
level (Stream 1 at 1.2 bar) by creating the Bypass Stream. The total shaft power demand
achieved with the Bypass design is 25.68 MW (0.2968 kWh·kg-1
of LNG).
The sensitivity analyses also show that the Two Flash Levels design benefits from an
additional compression stage (and associated intercooling stage), compared to the
CryoMan process. Manipulating the heat of vaporisation of Stream 3 and Stream 4 allows
a reduction in the refrigerant the flow rate fed to the second flash level compared to the
initial simulation. Additionally, because the pressure level of Stream 4 is increased from
4.8 bar to 18.5 bar, the refrigerant flow rate at the lowest pressure levels of the compressor
is also reduced (Stream 1 at 1.2 bar and Stream 3 at 5 bar). The Two Flash Levels design
achieved the same shaft power demand as that in the CryoMan process, i.e. 25.93 MW of
total shaft power consumption (0.2997 kWh·kg-1
of LNG).
Based on the sensitivity studies, the minimum total shaft power achieved with the Mixing
After Precooling design is 26.58 MW (0.3073 kWh·kg-1
of LNG), which is 2.5% higher
than that in the CryoMan process. The sensitivity analyses on the Mixing After Precooling
design suggest that the structural modifications proposed in the novel refrigeration cycle
(two streams partially mixed, after precooled in the MSHE, to create a new stream with
intermediate composition) would not have a significant effect on the performance of the
refrigeration cycle in order to bring shaft power savings, compared to the CryoMan
process.
Thus, the Bypass design already achieved shaft power savings of 1%, compared to the
CryoMan process, during the sensitivity analyses and is therefore considered for
optimisation of its operating variables. The Two Flash Levels design showed the same
shaft power demand as that of the CryoMan process during the sensitivity analyses.
However, in the sensitivity analyses, the operating variables are manipulated one at the
time. That is, the interactions between the operating variables are not considered. Thus, it
was concluded that the Two Flash Levels design is very likely to show shaft power
savings, compared to the CryoMan process, if its operating variables are simultaneously
optimised. Therefore, the Two Flash Levels design is considered for optimisation. The
Chapter 4 Evaluation of the Novel Refrigeration Cycles
127
Mixing After Precooling design, on the other hand, showed no shaft power savings
compared to the CryoMan process. Moreover, the sensitivity analyses suggest that the
structural modification in the Mixing After Precooling design has no significant effect in
the performance of the refrigeration cycle, compared to the CryoMan process. The Mixing
After Precooling design is not further considered for optimisation.
4.8 Conclusions
The natural gas stream provided by Lee (2001, Ch. 4) only as a T–H profile was simulated
in Aspen HYSYS v8.2. The natural gas stream was reconstructed using an optimisation
approach that consisted in minimising the sum of the squared difference of the enthalpy
profiles between the data provided and the optimised stream, at the intermediate
temperatures indicated in Table 4.1. This regression of the natural gas stream data allowed
a robust evaluation of the novel refrigeration cycles, comparing their performance (shaft
power demand) against that of the CryoMan process. The novel refrigeration cycles were
successfully modelled and simulated in Aspen HYSYS v8.2.
The sensitivity analyses showed that the Bypass design and the Two Flash Levels design
are refrigeration cycles that can bring shaft power savings, compared to the CryoMan
process, as they benefit from the structural modifications implemented. The Bypass design
already showed shaft power savings of 1% compared to the CryoMan process, whilst the
Two Flash Levels design yielded the same shaft power consumption as that of the
CryoMan process (0.2997 kWh·kg-1
of LNG). On the other hand, the sensitivity analyses
on the Mixing After Precooling design suggested that the structural modification would not
impact significantly in the performance of the refrigeration cycle and, therefore, would not
bring shaft power savings compared to the CryoMan process.
Consequently, the Bypass design and the Two Flash Levels design are taken into account
for optimisation of their corresponding operating variables, whereas the Mixing After
Precooling design is not further considered for optimisation.
In the following Chapter 5, the operating variables of the Bypass design and the Two Flash
Levels design are optimised in a case study for the liquefaction of the natural gas stream.
Chapter 5 Case Study: Optimisation of Novel Refrigeration Cycles
128
Chapter 5 – Case Study: Optimisation of Novel Refrigeration
Cycles
5.1 Introduction
According to the assessment performed on the novel refrigeration cycles in Chapter 4, the
Bypass design and the Two Flash Levels design are selected to optimise their
corresponding operating variables. The Mixing After Precooling design, on the other hand,
is discarded as the sensitivity analyses suggested that the structural modification proposed
would not bring shaft power savings to the refrigeration cycle, compared to the CryoMan
process.
In the case of the Bypass design, the sensitivity analyses revealed that the Bypass Stream
allowed reducing the flow rate of Stream 1 (1.62 kmol·s-1
) which is at the lowest pressure
level (1.2 bar), compared to LP Stream in the CryoMan process (1.79 kmol·s-1
at 1.2 bar),
to bring shaft power savings in spite the increase of methane in the composition of the
overall refrigerant stream. The specific shaft power is reduced to 0.2968 kWh·kg-1
LNG,
which represents shaft power savings of 1% compared to the CryoMan process.
Regarding the Two Flash Levels design, the sensitivity studies suggested that the trade-off
between refrigerant flow rate and heat of vaporisation (related to the refrigerant
composition) of Stream 3 and Stream 4 is being exploited, to reduce the refrigerant flow
rate fed to the second flash unit, in order to bring shaft power savings. Additionally, the
Two Flash Levels design would also benefit from an additional compression stage and its
corresponding intercooling stage, although the extra compression stage would increase the
capital costs relative to the CryoMan process. Therefore, the operating cost savings that
can be achieved with the Two Flash Levels design would have to justify the increase in
capital costs. The specific shaft power demand of the Two Flash Levels design, achieved
only with sensitivity studies, is 0.2997 kWh·kg-1
LNG.
The corresponding operating variables of these two novel refrigeration cycles are
optimised in a case study – presented in this chapter – using WORK software. Compared
to HYSYS, regarding the optimisation capabilities for refrigeration cycles, WORK
software is able to optimise the composition of the mixed refrigerant. Further, WORK is
also capable to use stochastic optimisation techniques (e.g. Genetic Algorithm). Even
Chapter 5 Case Study: Optimisation of Novel Refrigeration Cycles
129
though MATLAB can be used to perform stochastic optimisations and input the values of
the operating variables to the simulation in HYSYS (including the composition of the
mixed refrigerant) at each iteration of the optimisation, the flowsheet in HYSYS would
have to be well robust to include, for example, scenarios in which the pressure level of the
refrigerant streams are changed and thus, the refrigerant streams need to be fed to the
compressor at different compression stages.
The novel refrigeration cycles are optimised in order to find the combination of the values
of their corresponding operating variables that yield the lowest power consumption for the
liquefaction of a given natural gas stream. The optimisation is performed in WORK
software by means of a Genetic Algorithm method. The resulting shaft power demand of
the optimised designs is compared to that of the CryoMan process and the PRICO cycle;
the power savings achieved are also illustrated on an economic basis by comparing the
annual operating costs and savings obtained relative to the CryoMan process.
5.2 Optimisation criteria
The operating variables of the novel refrigeration cycles are optimised to minimise the
total shaft power demand for refrigerant compression. Thus, the objective function is
formulated as in Equation 5.1, and the mathematical formulation for the optimisation,
according to the modelling described in Chapter 3, is as follows:
Minimise: 𝑊𝑇𝑜𝑡𝑎𝑙 = ∑ 𝑊𝑖𝑆𝑡𝑔(Φ)𝑛
𝑖=1 𝑖 = 1,2, … , 𝑛 (5.1)
Subject to: Equations 3.1 to 3.35 (for the Bypass design)
Equations 3.17 to 3.57 (for the Two Flash Levels design)
Δ𝑇𝑀𝐼𝑁 ≥ 5℃ (5.2)
𝑃𝑅𝐴𝑇 ≤ 3 (5.3)
∑ 𝑥𝑗 = 1𝑚𝑗=1 𝑥𝑗 ∈ 𝑋𝑀𝑅 𝑗 = 1,2, … , 𝑚 (5.4)
𝑙𝑏 ≤ Φ ≤ 𝑢𝑏 (5.5)
where Φ = [𝑋𝑀𝑅 , 𝑃𝐻𝑖𝑔ℎ, 𝑓𝑉𝑎𝑝, 𝑓𝐿𝑖𝑞 , 𝑇𝐴𝑃𝑘 , 𝑃𝐴𝐸
𝑘 , 𝑇𝑜𝑢𝑡𝑘 , 𝛼]
𝑘 = 𝑆𝑡𝑟𝑒𝑎𝑚 1, 𝑆𝑡𝑟𝑒𝑎𝑚 2, 𝐵𝑦𝑝𝑎𝑠𝑠 (5.6a)
Φ = [𝑋𝑀𝑅 , 𝑃𝐻𝑖𝑔ℎ, 𝑃2𝑛𝑑 , 𝑓𝑉𝑎𝑝, 𝑓𝐿𝑖𝑞 , 𝑓𝑉𝑎𝑝2, 𝑓𝐿𝑖𝑞2, 𝑓2𝑛𝑑 , 𝑇𝐴𝑃𝑘 , 𝑃𝐴𝐸
𝑘 , 𝑇𝑜𝑢𝑡𝑘 ]
𝑘 = 𝑆𝑡𝑟𝑒𝑎𝑚 1, … , 𝑆𝑡𝑟𝑒𝑎𝑚 4 (5.6b)
Chapter 5 Case Study: Optimisation of Novel Refrigeration Cycles
130
Thus, Φ is a vector that includes the values of all the operating variables to be optimised in
each design, including compressor discharge pressure, refrigerant streams pressure level,
precooling temperatures and MSHE outlet temperatures, vapour and liquid flow rate
fractions, and refrigerant composition. Note that Φ includes the operating variables of the
Bypass design (bypass flow rate fraction; bypass stream pressure level, precooling
temperature and MSHE outlet temperature) only when the Bypass design is optimised
(Equation 5.6a), and includes the operating variables of the Two Flash Levels design (flow
rate fraction fed to the second flash unit; pressure in the second flash level; pressure levels,
precooling temperatures and MSHE outlet temperatures of the streams resulting from the
second flash level) only when the Two Flash Levels design is optimised (Equation 5.6b).
Equations 3.1 to 3.35 represent the mass and energy balances in the Bypass design and
Equations 3.17 to 3.57 are those for the Two Flash Levels design (detailed in Section 3.3.1
and 3.3.2, respectively); Equation 5.2 ensures that the minimum temperature difference
across the length of the composite curves in the MSHE is at least 5°C for feasible heat
transfer between hot and cold streams. With Equation 5.3, the pressure ratio in any
compression stage is limited to a maximum of 3. Equation 5.4 states that the sum of the
mole fractions of the components in the refrigerant mixture must be unity. During the
optimisation, each value in vector Φ is varied within a range that is limited by lower and
upper bounds; the lower bound for each variable is included in vector lb, and the upper
bound for each variable is included in vector ub (Equation 5.5).
The optimisation parameters for the optimisation of both the Bypass design and the Two
Flash Levels design are displayed in Table 5.1. The population size is 250 whereas the
maximum number of generations is 400. That is, the optimisation is performed from 250
different starting points in order to help avoiding local optima. Crossover is usually the
parameter with highest probability for producing new members in the population,
compared to mutation (Bäck et al., 1997, Ch. B1.1). Crossover probability is often around
0.90, whereas mutation rate usually has a probability of 0.01 (Poli et al., 2008, Ch. 2.4).
Alternatively, the values of both the crossover probability and mutation rate can be finely
adjusted – e.g. by performing sensitivity analyses – as they might have an impact on the
optimum solution found or in the computational time needed for the optimisation. The
values for crossover probability and mutation rate in this work – shown in Table 5.1 – are
selected according to the recommended values previously discussed.
Chapter 5 Case Study: Optimisation of Novel Refrigeration Cycles
131
Table 5. 1. Genetic Algorithm parameters for the optimisation of the novel refrigeration cycles.
Parameter Population Size Maximum Generations Crossover Probability Mutation Rate
Value 250 400 0.85 0.01
5.3 Problem statement
The operating variables of the novel refrigeration cycles are optimised in order to fully
liquefy a natural gas stream. The case study is that first published by Lee (2001), and later
used by Zheng (2009) to optimise the PRICO cycle and the CryoMan process.
The case study optimises the operating variables of the refrigeration cycle for liquefying a
given natural gas stream, where the objective is to minimise the shaft power consumption
of the refrigeration cycle. The natural gas stream is provided as a temperature–enthalpy
profile (shown in Table 5.2). WORK software is employed to perform the optimisation
and, as mentioned in Section 4.2 and also described in Appendix 1, the natural gas stream
is input directly as the T–H profile. According to the case study, the natural gas stream
enters the multi-stream heat exchanger (MSHE) at 25°C and 55 bar and leaves the MSHE
at –163°C and 50 bar as liquefied natural gas. The operating variables that are optimised
are the refrigerant composition, the compressor discharge pressure, the pressure level of
the refrigerant streams, the flow rate fraction of vapour and liquid streams to be split after
each flash separation unit, the precooling temperature of the refrigerant streams in the
MSHE and the MSHE outlet temperature of the refrigerant streams. The bypass flow rate
fraction is a variable only in the Bypass design, and the flow rate fraction of liquid fed to
the second flash unit and the pressure of the second flash unit are variables only in the Two
Flash Levels design.
In order to maintain consistency with the optimisation of the CryoMan process (Zheng,
2009), the following assumptions are made: the refrigerant mixture comprises methane,
ethane, propane, n-butane and nitrogen. The isentropic efficiency of the refrigerant
compression is assumed to be 80%. The maximum pressure ratio of each compression
stage is 3. The minimum temperature approach between the composite curves in the
MSHE is 5°C. During the optimisation, physical and thermodynamic properties of the
refrigerant mixture (e.g. temperatures, enthalpies) are calculated using Peng–Robinson
equation of state by interfacing with Aspen HYSYS v8.2 (Aspen Technology Inc., 2013).
Chapter 5 Case Study: Optimisation of Novel Refrigeration Cycles
132
Table 5. 2. Temperature–enthalpy profile of the natural gas stream to be liquefied (Lee, 2001).
Segment Supply
Temperature [°C]
Target
Temperature [°C] ΔH [kW] CP [kW·K
-1]
1.1 25.00 –06.03 –1861.5 60
1.2 –06.03 –34.09 –1964.3 70
1.3 –34.09 –57.65 –1885.0 80
1.4 –57.65 –70.10 –2490.0 200
1.5 –70.10 –74.55 –1780.0 400
1.6 –74.55 –82.26 –3084.0 400
1.7 –82.26 –96.50 –1424.0 100
1.8 –96.50 –115.00 –1850.0 100
1.9 –115.00 –163.00 –3840.0 80
5.4 Bypass design: Optimisation and discussion
Figure 5.1 shows the objective function of the optimisation (i.e. total shaft power demand)
plotted as a function of the generation (iteration) number, as the optimisation of the Bypass
design progresses. According to Figure 5.1, the optimisation reaches a steady objective
function value (i.e. total shaft power demand) of 25.22 MW. Further, the optimisation is
performed three times to gain confidence in the optimum solution found (on average, the
total shaft power demand varies only within a range of ±0.4%, see Appendix 2). The
optimisation takes 4,490 minutes – on average – using an Intel Core i5-4570 processor
with 3.20 GHz and 8.00 GB of RAM memory, resulting from the relatively large
population size and number of iterations of the optimisation.
Figure 5. 1. Objective function progression in the optimisation of the operating variables of the Bypass
design.
The resulting values of the operating variables of the optimised Bypass design are
displayed in Table 5.3 (the ‘Label’ column refers to the letters in Figure 5.2). The total
shaft power demand obtained is 25.22 MW (0.2915 kWh·kg-1
LNG). This shaft power
Chapter 5 Case Study: Optimisation of Novel Refrigeration Cycles
133
value represents shaft power savings of 3.2% when compared to 26.05 MW of shaft power
demand in the CryoMan process (0.3011 kWh·kg-1
LNG) and 10.8% in power savings
when compared to 28.27 MW of shaft power consumption in the PRICO cycle (0.3268
kWh·kg-1
LNG) as reported by Zheng (2009), who employed the same models,
assumptions, equation of state and optimisation approach.
Table 5. 3. Optimised operating variables of the Bypass design.
Label in Figure 5.2 Process variable Value A Refrigerant flow rate [kmol·s
-1] 3.43
A Composition [mole fraction]: - Methane 0.2250 - Ethane 0.3743 - Propane 0.1770 - n-Butane 0.1509 - Nitrogen 0.0728 A Discharge pressure [bar] 39.4 A Bypass fraction 0.0868 B Vapour flow rate fraction 0.8527 C Liquid flow rate fraction 0.1106 Stream 1 Stream 2 Bypass D Precooling temperature [°C] –166.3 –97.2 –99.8 E Expansion pressure [bar] 1.21 7.48 20.2 F MSHE outlet temperature [°C] 22.1 25.0 25.0 Performance indicators Value
Number of compression stages 4 ΔT
MIN [°C] 5.0
Total shaft power [MW] 25.22
Specific shaft power [kWh·kg-1
LNG] 0.2915 Shaft power savings (%)* 3.2
*compared to the CryoMan process (Zheng, 2009).
Figure 5. 2. Bypass design.
Chapter 5 Case Study: Optimisation of Novel Refrigeration Cycles
134
According to the optimised values of the operating variables displayed in Table 5.3, the
bypass flow rate fraction is decreased from 0.13 (see Table 4.7 in Section 4.4.3 for results
of the sensitivity analyses) to 0.08, even though the sensitivity analyses suggested that
increasing the flow rate of Bypass Stream would bring shaft power savings as a result of
decreasing the flow rate of Stream 1 (which is at the lowest pressure level). However, the
overall refrigerant flow rate is decreased from 3.65 kmol·s-1
(after the sensitivity analyses)
to 3.43 kmol·s-1
, which also leads to decreasing the flow rate of Stream 1. The compressor
discharge pressure is decreased from 41 bar to 39.4 bar, which is in agreement with the
trend shown in the sensitivity analyses (presented in Section 4.4.2) to avoid infeasible heat
transfer that would otherwise occur as a consequence of decreasing the vapour fraction of
the overall refrigerant stream (i.e. decreasing the vapour flow rate obtained in the flash
unit).
The flow rate fraction of the liquid from the flash unit (f Liq
) that is partially mixed to create
Stream 1 is increased from 0.08 (after the sensitivity analyses; see Table 4.7) to 0.11 after
the optimisation. As discussed in Section 4.4.2, increasing the value of f Liq
increases the
total shaft power demand of the refrigeration cycle (because the flow rate of Stream 1
would be increased, which is at the lowest pressure level). However, increasing the value
of f Liq
also increases the heat of vaporisation of Stream 1, which can help to reduce the
overall refrigerant flow rate to bring shaft power savings. On the other hand, decreasing the
flow rate of Stream 1 might also lead to infeasible heat transfer between hot and cold
streams in the MSHE. The optimised Bypass design thus exploits the trade-off between
increasing the heat of vaporisation of Stream 1 to reduce the overall refrigerant flow rate
and bring shaft power savings, and increasing the power consumption as a result of
increasing the flow rate of Stream 1 (at the lowest pressure level, 1.21 bar).
The Bypass Stream is 8.7% of the overall refrigerant stream. Introducing the Bypass
Stream allows to reduce the flow rate of the stream at the lowest pressure level in the
Bypass design (Stream 1) to bring shaft power savings. Figure 5.3 compares the flow rate
and heat load of Stream 1 in the optimised Bypass design (Figure 5.3a) against those of LP
Stream in the CryoMan process (Figure 5.3b) using the composite curves in the MSHE.
The flow rate of LP Stream in the CryoMan process is 1.8 kmol·s-1
compared to 1.7
kmol·s-1
of Stream 1 in the Bypass design. The heat duty of LP Stream in the CryoMan
process is 39.7 MW, whereas Stream 1 in the Bypass design has a heat load of 37.6 MW.
Note that the composite curves shown in Figure 5.3 represent an ideal scenario in terms of
Chapter 5 Case Study: Optimisation of Novel Refrigeration Cycles
135
heat transfer inside the MSHE; the actual composite curves might be affected if, for
example, the actual arrangement of the refrigerant streams inside the MSHE is considered.
However, determination of the actual arrangement of the refrigerant streams inside the
MSHE is beyond the scope of this work.
Reducing the heat load of Stream 1 in the Bypass design, compared to LP Stream in the
CryoMan process, resulted in shortening the length (x axis) of its T–H profile (Figure
5.3a), which would produce infeasible heat transfer at –83°C because the minimum
temperature approach between hot and cold composite curves would be less than 5°C.
Such a minimum driving force violation is avoided since both the Bypass Stream and
Stream 2 provide cooling to the hot streams from temperatures near –100°C, compared to
HP Stream in the CryoMan process which evaporates at –80°C. Thus, the Bypass Stream
allowed a reduction in the flow rate of Stream 1 by providing cooling at the necessary
temperature (–100°C) to avoid a minimum temperature approach violation, but at a higher
pressure level (20.2 bar) than Stream 1 (1.21 bar) to bring shaft power savings.
Figure 5. 3. Composite curves in the MSHE: a) the Bypass design; b) the CryoMan process.
Regarding the composition of the overall refrigerant stream, the optimised mole fraction
values of the components (in Table 5.3) are in agreement with the ranges obtained with the
sensitivity analyses (Table 4.5) in which heat transfer is feasible.
Furthermore, the mole fraction of methane in the overall composition of the refrigerant is
reduced to 0.2250, compared to 0.2288 in the CryoMan process; the mole fraction of
nitrogen is reduced to 0.0728 (compared to 0.0808 in the CryoMan process). Decreasing
the mole fraction of the light components in the refrigerant mixture helps reducing the total
Chapter 5 Case Study: Optimisation of Novel Refrigeration Cycles
136
shaft power demand of the refrigeration cycle, as previously discussed and illustrated in
Section 4.4.2.
The Bypass design achieves 3.2% savings in shaft power demand, compared to the
CryoMan process, by creating a new refrigerant stream that bypasses the flash unit (Bypass
Stream). In the optimised Bypass design, the compression of the overall refrigerant stream
takes place in 4 compression stages, i.e. with the same number of compression stages as
that in the CryoMan process. Thus, compared to the CryoMan process, the increase in
complexity of the Bypass design is only reflected with two additional refrigerant streams in
the MSHE (i.e. hot and cold Bypass Stream).
5.5 Two Flash Levels design: Optimisation and discussion
Figure 5.4 shows the minimisation of the objective function (i.e. total shaft power demand)
plotted as the optimisation of the Two Flash Levels design progresses, i.e. increasing the
number of generations (iterations). The optimisation reaches a steady objective function
value of 25.40 MW. Similar to the Bypass design, the optimisation is performed three
times in order to gain confidence in the optimum solution obtained (on average, the
objective function varies only within a range of ±0.4%, see Appendix 2). The optimisation
takes around 6,592 minutes – on average – with an Intel Core i5-4570 processor with 3.20
GHz and 8.00 GB of RAM memory.
Figure 5. 4. Objective function progression in the optimisation of the operating variables of the Two Flash
Levels design.
The optimised values of the operating variables in the Two Flash Levels design are shown
in Table 5.4 (where the ‘Label’ column refers to the letters in Figure 5.5). The total shaft
Chapter 5 Case Study: Optimisation of Novel Refrigeration Cycles
137
power consumption is 22.40 MW (0.2936 kWh·kg-1
LNG), which is 2.5% in shaft power
savings compared to 26.05 MW of shaft power demand in the CryoMan process (0.3011
kWh·kg-1
LNG), and 10.2% savings in shaft power compared to 28.27 MW of shaft power
demand in the PRICO cycle (0.3268 kWh·kg-1
LNG) according to Zheng (2009).
Table 5. 4. Optimised operating variables of the Two Flash Levels design.
Label in Figure 5.5 Process variable Value A Refrigerant flow rate [kmol·s
-1] 3.42
A Composition [mole fraction]: - Methane 0.2372 - Ethane 0.4013 - Propane 0.0587 - n-Butane 0.2285 - Nitrogen 0.0743 A Discharge pressure [bar] 41.1 H Pressure 2nd flash unit [bar] 26.7 G Liquid fraction to 2nd flash unit 0.2082 Stream 1 Stream 2 Stream 3 Stream 4 D Precooling temperature [°C] –162.3 –110.2 –96.5 –56.3 E Expansion pressure [bar] 1.2 7.8 4.7 11.2 F MSHE outlet temperature [°C] 19.6 24.0 18.9 25.0
B, I Vapour flow rate fraction 0.7506 0.2494 0.2975 0.7025 C, J Liquid flow rate fraction 0.2193 0.5725 0.2559 0.7441
Performance indicators Value
Number of compression stages 6 ΔT
MIN [°C] 5.0
Total shaft power [MW] 25.40
Specific shaft power [kWh·kg-1
LNG] 0.2936 Shaft power savings (%)* 2.5
*compared to the CryoMan process (Zheng, 2009).
Figure 5. 5. Two Flash Levels design.
Chapter 5 Case Study: Optimisation of Novel Refrigeration Cycles
138
The compression of the overall refrigerant stream is achieved with six compression stages.
As discussed in Section 4.5.1, additional compression stages, compared to the CryoMan
process, could help with shaft power savings in the refrigeration cycle because of their
corresponding intercooling stages. In Section 4.5.1, the compression of the overall
refrigerant in the initial simulation of the Two Flash Levels design is achieved with five
compression stages. After the optimisation, because of the design constraint of a maximum
pressure ratio of 3 in each compression stage, the resulting pressure level of the refrigerant
streams (shown in Table 5.4) lead to a compression of the overall refrigerant stream that is
achieved with six compression stages. Nevertheless, because of the two extra compression
stages and associated intercooling stages relative to the CryoMan process, the capital costs
of the Two Flash Levels design would increase accordingly (see Section 5.6 and Appendix
3 for further discussion).
According to the values in Table 5.4, the flow rate fraction of liquid from the first flash
unit that is fed to the second flash unit (f 2nd
) is decreased from 0.2240 (after the sensitivity
analyses, see Table 4.9) to 0.2082. The optimised value of f 2nd
is in agreement with the
discussion in Section 4.5.2, in which reducing the value of f
2nd decreases the shaft power
demand (as a consequence of reducing the flow rate of the streams leaving the second flash
unit). On the other hand, values of f 2nd
below 0.20 would lead to infeasible heat transfer in
the temperature range in which Stream 3 and Stream 4 (resulting from the second flash
unit) provide cooling in the MSHE, i.e. from –96°C to 25°C for the optimised conditions
(see Figure 5.6).
Figure 5. 6. Composite curves in the MSHE of the optimised Two Flash Levels design.
Chapter 5 Case Study: Optimisation of Novel Refrigeration Cycles
139
The pressure in the second flash unit (P2nd
), according to the optimised values in Table 5.4,
is decreased to 26.7 bar (from 30 bar after the sensitivity analyses, see Table 4.9).
According to the sensitivity analyses in Section 4.5.2 and Figure 4.17, more vapour is
produced in the second flash unit as its pressure level is decreased.
Also, the optimised values of the vapour and liquid flow rate fractions in Table 5.4,
indicates that Stream 4 has above 70% of the flow rate fed to the second flash unit. The
pressure level of Stream 4 is 11.2 bar, which is the highest pressure level amongst the
refrigerant streams. Thus, over 70% of the flow rate fed to the second flash unit is
compressed from a relatively high pressure level. According to a previous discussion in
Section 4.4.2 (illustrated with Figure 4.7b for the Bypass design), the shaft power demand
for compression decreases as the pressure level of the refrigerant streams increases, for a
fixed compressor discharge pressure.
Regarding the optimised composition of the overall refrigerant stream in the Two Flash
Levels design, the mole fraction of n-butane is increased from 0.1517, in the CryoMan
process, to 0.2285. Increasing the mole fraction of n-butane in the composition of the
overall refrigerant stream reduces the total shaft power consumption for refrigerant
compression (see discussion in Section 4.4.2). However, the increase in the mole fraction
of n-butane is at the expense of decreasing the mole fraction of propane to 0.0587
(compared to 0.1684 in the CryoMan process), which increases the shaft power demand.
The mole fraction of ethane is increased to 0.4013 from 0.3703 in the CryoMan process,
which also increases the shaft power demand; although the overall shaft power
consumption is less sensitive to changes in the mole fraction of ethane compared to
changes in the mole fraction of the remaining components in the refrigerant mixture (see
Section 4.4.2).
Also, note that the mole fraction values of ethane, propane and n-butane are outside of the
ranges provided in Table 4.7 (see Section 4.5.2) in which heat transfer is feasible according
to the sensitivity analyses. For example, the optimised mole fraction value for propane is
0.0587 (see Table 5.4), whereas the range in which the mole fraction of propane would
lead to feasible heat transfer in the MSHE, according to the sensitivity analyses in Section
4.5.2, is between 0.10 and 0.20. After the optimisation (where the operating variables are
manipulated simultaneously), the complex interactions and trade-offs between the
operating variables in the refrigeration cycle (e.g. pressure levels, stream flow rates,
Chapter 5 Case Study: Optimisation of Novel Refrigeration Cycles
140
compositions from flash separations) allowed the mole fraction of the components in the
overall refrigerant stream to be outside the ranges obtained through sensitivity analyses
(provided in Table 4.7, Section 4.5.2), in which only the composition is manipulated whilst
the remaining degrees of freedom (e.g. refrigerant flow rate, stream pressure levels) are
held constant.
The optimised Two Flash Levels design thus exploits the complex interactions and trade-
offs between refrigerant flow rate, composition and pressure levels of the new streams in
the refrigeration cycle (Stream 3 and Stream 4). The design constraint of maximum
pressure ratio in a compression stage is also exploited, leading to a compression of the
overall refrigerant stream achieved with six compression stages. On the other hand, as
mentioned before, the two extra compression stages would significantly increase the
capital costs of the Two Flash Levels design.
Overall, the Two Flash Levels design achieves shaft power savings of 2.5% relative to the
CryoMan process. The structural modification proposed in the Two Flash Levels design
consists in throttling and flashing a portion of the liquid from the first flash unit, to a lower
pressure level. Two new refrigerant streams (Stream 3 and Stream 4) result from partially
mixing the product streams of the second flash unit. The compression of the overall
refrigerant stream is achieved with six compression stages. Thus, compared to the
CryoMan process, the increase in the complexity of the Two Flash Levels design is
reflected with four additional refrigerant streams in the MSHE (two hot and two cold
streams), and also with two extra compression stages.
5.6 Operating costs comparison between the novel refrigeration cycles and the
benchmark processes
The refrigerant compression in the refrigeration cycles is the most energy-consuming stage
in the liquefaction process, and hence, dominates operating costs of LNG plants (Mokhatab
et al., 2014b, Ch. 3.2). In this section, the total shaft power achieved with the novel
refrigeration cycles is compared against those in the benchmark processes (i.e. the PRICO
cycle and the CryoMan process) on an annual operating cost basis, in order to illustrate
operating cost savings as a result of the shaft power savings.
According to the results shown in Table 5.3 and Table 5.4, shaft power savings of 3.2%
and 2.5% are achieved with the Bypass design and with the Two Flash Levels design,
respectively, compared to the CryoMan process. The unit shaft power energy cost is
Chapter 5 Case Study: Optimisation of Novel Refrigeration Cycles
141
assumed to be £0.0955·kWh-1
for extra-large scale industrial consumers (U.K. Department
of Energy & Climate Change, 2015). Table 5.5 presents the annual operating costs of the
novel refrigeration cycles as well as those of the benchmark processes, and compares the
operating cost savings relative to the CryoMan process.
According to Table 5.5, the shaft power demand achieved with the Bypass design (0.2915
kWh·kg-1
LNG) would lead to operating cost savings of nearly £0.70 million per annum
compared to the CryoMan process and over £2.5 million per annum compared to the
PRICO cycle. The optimised Two Flash Levels design yielded a specific shaft power
consumption of 0.2936 kWh·kg-1
LNG, which represents operating cost savings of £0.54
million per annum relative to the CryoMan process and £2.4 million per annum when
compared to the PRICO cycle. Thus, significant operating cost savings are achieved with
the novel refrigeration cycles, according to the shaft power savings obtained, compared to
the benchmark processes.
Table 5. 5. Operating cost savings comparison (relative to the CryoMan process) between the novel
refrigeration cycles and benchmark processes.
PRICO cycle CryoMan process Bypass design Two Flash Levels design
Total shaft power
[MW] 28.27 26.05 25.22 25.40
Specific shaft power
[kWh·kg-1
LNG] 0.3268 0.3011 0.2915 0.2936
Shaft power savings
[%] – 8.5 - + 3.2 + 2.5
Number of
compression stages 4 4 4 6
Operating costs
[£ million per annum] 23.65 21.79 21.10 21.25
Operating cost savings
[£ million per annum] – 1.86 - + 0.69 + 0.54
However, the optimised Two Flash Levels design is impacted by the design constraint of
maximum compression ratio in a single compression stage (Equation 5.3): the overall
refrigerant stream is compressed to the compressor discharge pressure with six
compression stages. Thus, the capital costs of the Two Flash Levels design would
significantly increase as a consequence of the two extra compression stages compared to
the CryoMan process.
Chapter 5 Case Study: Optimisation of Novel Refrigeration Cycles
142
In order to compare the Two Flash Levels design and the CryoMan process on the same
basis, the CryoMan process is optimised for the case that the compression of the overall
refrigerant stream is achieved with six compression stages. The optimisation of the
CryoMan process is presented in Appendix 3. The total shaft power demand of the
CryoMan process optimised with six compression stages is 24.94 MW (0.2883 kWh·kg-1
of LNG), which is 1.8% lower than that of the optimised Two Flash Levels design.
That is, the comparison on a basis of equal number of compression stages suggests that the
Two Flash Levels design mainly benefits from the extra compression stages, rather than
from the structural modification proposed. Thus, the shaft power savings and the
corresponding operating cost savings achieved by the Two Flash Levels design – shown in
Table 5.5 – would not justify the increase in capital costs.
5.7 Conclusions
The operating variables of the novel refrigeration cycles (the Bypass design and the Two
Flash Levels design) were successfully optimised to evaluate their performance in the LNG
production case study (0.75 million t per annum) that was employed by Zheng (2009) to
optimise the CryoMan process and the PRICO cycle. The optimisation was performed
using WORK software.
Genetic Algorithm was employed as the optimisation method in order to help avoiding
local optima. The objective function to minimise was the total shaft power demand for
refrigerant compression. The Genetic Algorithm method demonstrated to be a robust
optimisation method, allowing explore thoroughly the combination of values of the
operating variables of the novel refrigeration cycles to minimise the objective function.
The optimisation was performed three times on each novel refrigeration cycle. The similar
solutions resulting from their corresponding optimisations, gives confidence in the optimal
results obtained.
The structural modification in the Bypass design allowed a reduction of the flow rate of the
refrigerant stream at the lowest pressure level (Stream 1 at 1.21 bar) to bring shaft power
savings. The optimised Bypass design achieved a shaft power consumption of 25.22 MW
(0.2915 kWh·kg-1
LNG), i.e. shaft power savings of 3.2% compared to the CryoMan
process and 10.8% compared to the PRICO cycle. The shaft power savings achieved with
the Bypass design would be equivalent to savings of £0.69 million per annum in operating
Chapter 5 Case Study: Optimisation of Novel Refrigeration Cycles
143
costs, compared to the CryoMan process and £2.55 million per annum compared to the
PRICO cycle. Furthermore, the compression of the overall refrigerant stream in the Bypass
design was achieved with four compression stages, i.e. the same number of compression
stages as that in the CryoMan process. Thus, the shaft power savings (and corresponding
operating cost savings) offered by the Bypass design, compared to the CryoMan process,
are at the expense of only minor increase in the structure complexity of the refrigeration
cycle design, that is an additional refrigerant stream that bypasses the flash unit.
The optimised Two Flash Levels design yielded a total shaft power demand of 25.40 MW
(0.2936 kWh·kg-1
LNG), which is 2.5% in shaft power savings compared to the CryoMan
process and 10.2% compared to the PRICO cycle. The shaft power savings are equivalent
to operating cost savings of £0.54 million per annum, compared to the CryoMan process.
However, the compression of the overall refrigerant stream, in the optimised Two Flash
Levels design, resulted with six compression stages, i.e. two additional compression stages
compared the CryoMan process. It was concluded – in Appendix 3 – that the Two Flash
Levels design was mainly benefited from the intercooling associated with the additional
compression stages, rather than from the structural modification proposed (i.e. two new
refrigerant streams from a second flash unit). Thus, the operating cost savings achieved
would not justify the increase in complexity and associated capital costs.
As previously discussed in Section 3.1, an important trade-off exists in the design of
refrigeration cycles for small scale LNG processes; on the one hand, the refrigeration cycle
should be of low complexity to help keeping the capital costs low. On the other hand, the
energy-efficiency of the refrigeration cycle should be high to minimise the shaft power
consumption in the refrigeration cycle, which dominates operating costs in LNG processes.
The novel refrigeration cycles in this work, developed by modifying the configuration of
the CryoMan process, showed that the Bypass design exploited this trade-off as it was
benefited from the structural modification proposed (a stream bypasses the flash unit).
Significant shaft power savings and, therefore, significant operating cost savings were
achieved, compared to the CryoMan process, with minor increase in design complexity. In
the case of the Two Flash Levels design, even though shaft power savings were achieved,
the operating cost savings achieved does not justify the increase in capital costs associated
with two additional compression stages compared to the CryoMan process.
Chapter 6 Conclusions and Future Work
144
Chapter 6 – Conclusions and Future Work
6.1 Conclusions
In the current scenario on global energy demand and supply, natural gas is a major source
of energy. Trends project that the production and market of natural gas will continue
growing in the next two decades. Thus, small natural gas reserves are likely to become
increasingly attractive for commercial exploitation as liquefied natural gas (LNG) at small
scale production, i.e. up to 1 million t per annum (Mokhatab et al., 2014b, Ch. 3.3).
LNG processes are both capital- and energy-intensive. The design of refrigeration cycles
for LNG process at small scale is challenging: high energy-efficiency of the refrigeration
cycle is important to minimise the shaft power demand for refrigerant compression (which
dominates operating costs). However, energy savings are usually at the expense of
increasing the complexity of the design and associated capital costs. In small scale LNG
processes, complexity of the refrigerant cycle design should be kept low to help keep
capital costs relatively low.
The literature review – presented in Chapter 2 – revealed that a limited range of mixed
refrigerant cycles have been studied for the production of LNG at small scale. The
CryoMan process (Zheng, 2009; Kim and Zheng, 2011), developed by structurally
modifying the simplest commercial mixed refrigerant cycle (the PRICO cycle), showed
significant shaft power savings (nearly 8%) compared to the PRICO cycle. The structural
modifications consisted on a flash unit after the partial condenser to separate the overall
refrigerant stream into vapour and liquid phases (that have different composition); these
vapour and liquid phases are partially mixed to create the two actual refrigerant streams.
The flow rate and composition of the refrigerant streams can thus be manipulated. The
shaft power savings in the CryoMan process resulted from the exploitation of the trade-offs
between refrigerant flow rate, refrigerant composition and pressure level of the refrigerant
streams.
Further structural modifications in the CryoMan process were explored in this work, in
order to explore and exploit the trade-off in the design of refrigeration cycles for LNG at
small scale: complexity (which can be related to capital costs) against energy-efficiency (in
particular shaft power demand for refrigerant compression, to account for operating costs).
However, the complexity of a refrigeration cycle has not been clearly defined in the open
Chapter 6 Conclusions and Future Work
145
research literature. Thus, in this work, design constraints in the refrigerant compressor and
in the multi-stream heat exchanger (MSHE) were defined to impose a limit of how
‘complex’ the novel refrigeration cycles can be.
In this thesis, three novel mixed refrigerant cycles were proposed – in Chapter 3 –, namely
the Bypass design, the Two Flash Levels design, and the Mixing After Precooling design.
The novel refrigeration cycles were developed by exploring structural modifications
applied to the base case refrigeration cycle (i.e. the CryoMan process). The modification
proposed for the Bypass design consisted of a stream that bypasses the flash unit, with an
independent pressure level. In the Two Flash Levels design, a portion of the liquid stream
obtained from the first flash unit is expanded to a second pressure level and flashed; two
new refrigerant streams are created by partially mixing the resulting vapour and liquid
phases from the second flash unit. In the Mixing After Precooling design, a new refrigerant
stream is created by partially mixing the two hot refrigerant streams once precooled in the
MSHE and expanded; the new refrigerant stream is constrained to have the same pressure
level as that of the refrigerant stream at the lowest pressure level.
The structural modifications aimed to exploit trade-offs between the operating variables of
the refrigerant streams (refrigerant flow rate, composition, pressure level, etc.) to bring
further shaft power savings to the refrigeration cycles, with relatively minor increase in the
complexity of the design to help keep capital costs low.
The three novel refrigeration cycles were first assessed and screened – in Chapter 4 –
through sensitivity analyses, in commercial simulator software (i.e. Aspen HYSYS), for
the liquefaction of the same natural gas stream as in the CryoMan process. The promising
designs were further considered for optimisation of their corresponding operating
variables. Because the natural gas stream is provided as a temperature–enthalpy profile
only, its conditions (flow rate, composition, etc.) were determined – in Appendix 1 – with
an optimisation approach that minimises the sum of squared enthalpy differences between
the published data and the optimised stream.
The preliminary assessment showed that the Bypass design could achieve shaft power
savings of 1% compared to the CryoMan process. The new stream in the Bypass design
allowed a reduction of the flow rate of the refrigerant stream at the lowest pressure level,
by providing cooling at the necessary temperature (–100°C) but at a higher pressure (20.2
bar). Additionally, the new stream impacted the composition of the overall refrigerant
Chapter 6 Conclusions and Future Work
146
stream by reducing the mole fraction of the light components in the mixture, resulting in
shaft power savings. The Two Flash Levels design demonstrated potential for shaft power
savings; this novel design was benefited from reducing the flow rate of the new refrigerant
streams and increasing their corresponding pressure level, and also from the intercooling
associated with an additional compression stage. The extra compression stage would,
however, increase the capital costs of the refrigeration cycle. In the Mixing After
Precooling design, on the other hand, the structural modification showed no effect towards
bringing shaft power savings to the refrigeration cycle. The studies suggested that, as the
flow rate of the new stream increases, its composition would help to reduce the overall
refrigerant flow rate. However, because the new stream is constrained to have the lowest
pressure level, increasing its flow rate resulted in increasing the shaft power demand.
Only the Bypass design and the Two Flash Levels design were optimised – in Chapter 5 –:
the benefits obtained from their corresponding structural modifications were illustrated
with an industrially-relevant case study that has been studied in several research
publications. The case study consists in the liquefaction of a given natural gas stream while
minimising the shaft power consumption in the refrigeration cycle. An optimisation using
Genetic Algorithm was performed to the novel refrigeration cycles. The optimisation
considered the simultaneous manipulation of the degrees of freedom of each novel
refrigeration cycle (e.g. refrigerant flow rate and composition, streams pressure level) and
explored the complex interactions between their corresponding operating variables, thus
avoiding local optima.
The optimised Bypass design yielded 3.2% savings in shaft power compared to the
CryoMan process. That is, power savings equivalent to operating cost savings of £0.69
million per annum were possible. Furthermore, the compression of the overall refrigerant
stream was achieved with only four compression stages, i.e. the same number of
compression stages as that in the CryoMan process. Therefore, the shaft power savings
(and corresponding operating cost savings) achieved by the Bypass design come at the
expense of only minor increase in complexity, i.e. an additional refrigerant stream that
bypasses the flash unit, compared to the CryoMan process.
In the case of the optimised Two Flash Levels design, 2.5% savings in shaft power
consumption were achieved, which are equivalent to operating cost savings of £0.54
million per annum. However, the compression of the overall refrigerant stream was
Chapter 6 Conclusions and Future Work
147
achieved with two additional compression stages compared to the CryoMan process. It was
concluded – in Appendix 3 – that the novel refrigeration cycle was mainly benefited from
the intercooling associated with the extra compression stages, rather than the structural
modification proposed. The additional compression stages would increase the capital costs
of this novel design; the operating cost savings achieved in the case study would not justify
the increase in complexity and associated capital costs.
The Genetic Algorithm optimisation method was shown to be a robust optimisation
approach to thoroughly explore the complex interactions between the operating variables
in the novel refrigeration cycles. Moreover, the optimisation was performed three times for
each novel refrigeration cycle; the similar results obtained allowed gaining confidence in
the optimal solutions found.
The structural modifications proposed in this work allowed the novel designs to bring shaft
power savings with relatively minor increase in the complexity of the refrigeration cycles.
Shaft power savings achieved with the novel refrigeration cycles yielded up to 3.2% in the
case of the Bypass design (equivalent to operating cost savings of £0.69 million per
annum) compared to the CryoMan process. Compared to the PRICO cycle, the shaft power
savings achieved by the novel refrigeration cycles yielded up to 10.8% (equivalent to
operating cost savings of £2.55 million per annum).
6.2 Future work
The constraints for complexity defined in this work may be different from those considered
in industrial practice, and thus other criteria for the complexity could be established in the
development of refrigeration cycles for LNG processes at small scale. For example, the
constraint of the number of streams inside the MSHE could be replaced with a constraint
for a maximum heat transfer area. Alternatively, the criteria could be combined to include
both constraints in the MSHE.
Also, the evaluation of structural modifications in the liquefaction process could be
strengthened by including an assessment of the capital cost of the novel refrigeration
cycles. For example, even though the design constraints were defined to include only one
compressor, its cost would be expected to increase not only as the shaft power demand
increases, but also as the number of compression stages increases (i.e. as the compressor
becomes more complex). The same would be expected for the cost of the MSHE as the
Chapter 6 Conclusions and Future Work
148
total heat transfer area increases, or as the number of streams in a single MSHE increases
and the equipment design becomes more complicated. The objective function of the
optimisation could thus be the minimisation of the total annualised costs in order to
account for both shaft power and complexity of the refrigeration cycle design.
The scope of structural modifications for the development of novel refrigeration cycles
should be extended to cascade cycles and, therefore, should be applied to larger scale LNG
processes (e.g. LNG production above 2.5 million t per annum). Thus, novel cascade
refrigerant cycles could be developed, in which the modifications applied would improve
further the high energy-efficiencies achieved with commercial cascade cycles, and bring
significant operating cost savings.
Furthermore, a robust integrated design methodology, e.g. that includes power systems
design and driver selection to run the refrigerant compressors, such as that proposed by Del
Nogal (2006, Ch. 4 and 5), could be developed to include the novel refrigeration cycles.
The benefits of structural modifications in the refrigeration cycles, presented in this thesis,
could be further explored and extended by exploiting the interactions between the
refrigeration cycles and the power systems (e.g. steam turbines, gas turbines, electric
motors) in the production of LNG.
Regarding the optimisation, the values of the Genetic Algorithm parameters (e.g. crossover
probability, mutation rate) could be further studied in order to investigate potential benefits
on the optimisation performance, both on the final solution obtained (objective function)
and on the computational time needed. Finally, although the stochastic optimisation
approach employed in this work (i.e. Genetic Algorithm) gives confidence that the
solutions found are closer to a global optimum compared to those from deterministic
methods, the results obtained are not guaranteed to be optimal. Thus, a different
optimisation approach could be studied. For instance, a combination of stochastic and
deterministic optimisation methods could be implemented to explore thoroughly the
solution space and also to guarantee optimality of the final results obtained.
Appendix 1 Determination of the Natural Gas Stream Conditions
149
Appendix 1 – Determination of the Natural Gas Stream
Conditions
The temperature–enthalpy profile of a natural gas stream first published by Lee (2001, Ch.
4), presented in Table A1.1 and Figure A1.1, has been employed in the open research
literature by Lee (2001, Ch. 4), Del Nogal (2006, Ch. 2), Remeljej and Hoadley (2006),
and Zheng (2009, Ch. 3), to assess the shaft power consumption of a refrigeration cycle to
liquefy the natural gas stream.
Lee (2001), also in (Lee et al., 2002), presented the natural gas data (T–H profile) in a case
study in which the composition of the mixed refrigerant in the PRICO cycle is optimised in
order to minimise the shaft power demand for refrigerant compression. Similarly, Del
Nogal (2006), also in (Del Nogal et al., 2008), employed the natural gas T–H data in a case
study to optimise the operating variables of the PRICO cycle to minimise the total
compression shaft power demand, but considering simultaneous manipulation of the
degrees of freedom (refrigerant composition, pressure levels, refrigerant flow rate).
Remeljej and Hoadley (2006) used the natural gas stream to simulate and compare the
performance (using the shaft power demand as the performance indicator) of four different
refrigeration cycles, including the PRICO cycle, in the liquefaction of the natural gas
stream for small scale LNG processes. Zheng (2009) employed the natural gas data to
optimise the operating variables of five refrigeration cycle designs, including the PRICO
cycle and the CryoMan process, to minimise the total shaft power demand in order to
explore structural modifications to the PRICO refrigeration cycle.
Since Zheng (2009) employed the previously mentioned natural gas stream to evaluate the
performance of the CryoMan process, this natural gas stream is also used to assess the
novel refrigeration cycles developed in this work. The evaluation of the novel refrigeration
cycles is performed in two stages: a preliminary assessment using Aspen HYSYS v8.2
(Aspen Technology Inc., 2013) (see Chapter 4), and a case study in which the operating
variables of the novel refrigeration cycles are optimised to minimise the total shaft power
demand and the resulting specific shaft power consumption is compared to the benchmark
processes, i.e. the PRICO cycle and the CryoMan process (see Chapter 5).
The natural gas T–H profile data is used by Zheng (2009) to optimise the CryoMan process
in WORK software; no details of the composition of the natural gas stream, flow rate or
Appendix 1 Determination of the Natural Gas Stream Conditions
150
pressure drop profile are needed. On the other hand, to simulate the natural gas stream in
HYSYS, the full conditions of the stream are needed, i.e. flow rate, composition,
temperatures at the inlet and outlet of the multi-stream heat exchanger (MSHE), inlet and
outlet pressures as well as the pressure drop profile.
Table A1. 1. Temperature–enthalpy profile of the natural gas stream (Lee, 2001; Del Nogal, 2006; Zheng,
2009).
Segment Supply Temperature
[°C]
Target Temperature
[°C] ΔH [kW] CP [kW·K
-1]
1.1 25.00 –06.03 –1861.5 60
1.2 –06.03 –34.09 –1964.3 70
1.3 –34.09 –57.65 –1885.0 80
1.4 –57.65 –70.10 –2490.0 200
1.5 –70.10 –74.55 –1780.0 400
1.6 –74.55 –82.26 –3084.0 400
1.7 –82.26 –96.50 –1424.0 100
1.8 –96.50 –115.00 –1850.0 100
1.9 –115.00 –163.00 –3840.0 80
Figure A1. 1. Temperature–enthalpy profile of the natural gas stream (Lee, 2001).
In Lee (2001), the data available for the natural gas stream is the T–H profile, the inlet
pressure (55 bar) and outlet pressure (50 bar). In Del Nogal (2006), only the T–H profile
data of the natural gas stream are shown, but details of its inlet and outlet pressures are
omitted, and the flow rate is not stated. Moreover, even though Remeljej and Hoadley
(2006) provide a composition and mass flow rate (22.60 kg·s-1
) of the natural gas stream,
in addition to the inlet and outlet pressure, the resulting molar flow rate (1.37 kmol·s-1
) is
different to that stated by Zheng (2007) (1 kmol·s-1
). Additionally, the T–H profile
-200
-150
-100
-50
0
50
0 5 10 15 20 25
Tem
per
atu
re [
°C]
ΔH [MW]
Appendix 1 Determination of the Natural Gas Stream Conditions
151
obtained according to the natural gas stream conditions used by Remeljej and Hoadley
(2006) does not match that originally published by Lee (2001), as shown in Table A1.2 and
Figure A1.2. In Zheng (2009), the inlet and outlet pressures of the natural gas are omitted.
Furthermore, the pressure drop profile of the natural gas in the MSHE is not reported in
any of the previously mentioned publications. Consequently, the conditions of the natural
gas stream (composition, flow rate, inlet and outlet pressure, and pressure drop profile in
the MSHE) are unclear.
Table A1. 2. Natural gas stream profile according to Remeljej and Hoadley (2006) compared to Lee (2001).
Segment Supply
Temperature [°C]
Target
Temperature [°C]
Cumulative ΔH [kW] (Lee, 2001)
Cumulative ΔH [kW]
(Remeljej and Hoadley, 2006) Difference
[%]
1.1 25.00 –06.03 20178.8 18972.9 6.0
1.2 –06.03 –34.09 18317.3 17119.6 6.5
1.3 –34.09 –57.65 16353 15313.1 6.4
1.4 –57.65 –70.10 14468 13467.6 6.9
1.5 –70.10 –74.55 11978 11891.1 0.7
1.6 –74.55 –82.26 10198 10263.7 –0.6
1.7 –82.26 –96.50 7114 7692.4 –8.1
1.8 –96.50 –115.00 5690 5722.0 –0.6
1.9 –115.00 –163.00 3840 3871.3 –0.8
Figure A1. 2. Temperature–enthalpy profile comparison between the data provided by Lee (2001) and that
according Remeljej and Hoadley (2006).
Thus, an optimisation is undertaken in order to find a stream with the same T–H profile as
that provided in Table A1.1. The objective function of the optimisation is to minimise the
sum of squared difference of the enthalpies compared to the data in Table A1.1, at the
fixed intermediate target temperatures shown in Table A1.1. Squares of differences of
enthalpy values account for both positive and negative differences against the published
-200
-150
-100
-50
0
50
0 5 10 15 20 25
Tem
per
atu
re [
°C]
ΔH [MW]
Remeljej and Hoadley (2006)Lee (2001)
Appendix 1 Determination of the Natural Gas Stream Conditions
152
data. The degrees of freedom optimised, i.e. the variables to be varied, are the molar
fraction of (n – 1) components, the overall mass flow rate as well as the inlet pressure of
the natural gas stream.
The overall pressure drop of the natural gas streams is assumed as 5 bar (Lee, 2001).
Because the pressure drop profile across the MSHE is not reported, the optimisation is
performed with three different pressure drop profile assumptions: i) linear dependence on
the overall temperature change of the stream (Equation A1.1); ii) linear dependence on the
heat that the natural gas stream rejects during the liquefaction process (Equation A1.2); iii)
zero pressure drop is also assumed, i.e. the natural gas stream is liquefied at constant
pressure and, therefore, its inlet and outlet pressures are the same (Equation A1.3).
𝑃𝑖 = 𝑃𝑖𝑛 − [𝑇𝑖𝑛−𝑇𝑖
𝑇𝑖𝑛−𝑇𝑜𝑢𝑡∙ ∆𝑃𝑇𝑜𝑡𝑎𝑙]; 𝑖 = 1, 2, … , 9 (A1.1)
𝑃𝑖 = 𝑃𝑖𝑛 − [∆𝐻𝑖𝑛−∆𝐻𝑖
∆𝐻𝑖𝑛−∆𝐻𝑜𝑢𝑡∙ ∆𝑃𝑇𝑜𝑡𝑎𝑙]; 𝑖 = 1, 2, … , 9 (A1.2)
∆𝑃𝑇𝑜𝑡𝑎𝑙 = 0 (A1.3)
where ΔPTotal = Overall pressure drop of the natural gas stream across the MSHE
Pin = Pressure of the natural gas stream at the inlet of the MSHE
Pi = Pressure of the natural gas stream at the ith temperature segment
Tin = Temperature of the natural gas stream at the inlet of the MSHE
Tout = Temperature of the natural gas stream at the outlet of the MSHE
Ti = Temperature of the natural gas stream at the ith segment
ΔHin = Cumulative enthalpy change value of the natural gas stream at the inlet of the MSHE
ΔHout = Cumulative enthalpy change value of the natural gas stream at the outlet of the MSHE
ΔHi = Cumulative enthalpy change value of the natural gas stream at the ith temperature segment
The optimisation is performed for each pressure drop profile assumption. It is assumed that
the natural gas stream is composed of methane, ethane, propane, n-butane, i-butane and
nitrogen, because those are the components commonly found in natural gas to be liquefied
(Mokhatab et al., 2014a, Ch. 1.3). The mathematical formulation of the objective function
is as in Equation A1.4:
Appendix 1 Determination of the Natural Gas Stream Conditions
153
Minimise: 𝑓(𝐱∗) = ∑ (∆𝐻𝑖 − ∆𝐻𝑐𝑎𝑙𝑐)29𝑖=1 𝑖 = 1, 2, … , 9 (A1.4)
Subject to: ∑ 𝑥𝑗6𝑗=1 = 1 (A1.5)
0 ≤ 𝑥𝑗 ≤ 1; 𝑗 = 1, 2, … , 5 (A1.6)
0.67 ≤ 𝑚 ≤ 1 (A1.7)
0.73 ≤ 𝑃𝑖𝑛 ≤ 1 (A1.8)
where x* = [x1, x2, x3, x4, x5, m, Pin]
ΔHi = Enthalpy change value of the natural gas stream in Table A1.1 at the ith temperature segment
ΔHcalc = Enthalpy change value of the optimised natural gas stream at the ith temperature segment
xj = Mole fraction of the jth component in the optimised natural gas stream
m = Normalised value of the mass flow rate of the optimised natural gas stream
Pin = Normalised value of the inlet pressure of the optimised natural gas stream
Equation A1.5 states that the sum of the components in the natural gas stream must yield
unity. During the optimisation, the mole fraction of five components in the natural gas
stream (x1 to x5) is varied from 0 to 1 (Equation A1.6).
Based on the mass flow rate reported by Remeljej and Hoadley (2006) (22.60 kg·s-1
), the
mass flow rate of the natural gas stream is allowed to vary between 20 kg·s-1
and 30 kg·s-1
.
The inlet pressure of the natural gas stream is varied from 55 bar to 40 bar during the
optimisation for each pressure drop profile assumption. The mass flow rate (m) and inlet
pressure (Pin) ranges are normalised by dividing by their corresponding upper bound (i.e.
30 kg·s-1
and 55 bar, respectively) as stated by Equation A1.7 and Equation A1.8. The
optimum combination of composition, flow rate, inlet pressure and pressure drop profile
would present minimum difference of enthalpy change with the published data.
The temperature–enthalpy profile of the natural gas depends on the interactions between
the components in the mixture, and on the overall composition, pressures and flow rate.
The optimisation is thus performed using fmincon solver within MATLAB for nonlinear
problems (The MathWorks Inc., 2013). The solution obtained (values of vector x*) in
nonlinear problems is likely to depend on the initial conditions (starting point) of the
optimisation (Edgar et al., 2001, Ch. 10.1). Thus, a set of different starting points is
proposed to explore thoroughly the possible solutions achieved by optimisation, as shown
in Table A1.3.
Appendix 1 Determination of the Natural Gas Stream Conditions
154
Since the main component of the natural gas is methane, the compositions selected as the
starting points are composed of methane ranging from 88% to 98%. Ethane is varied from
2% to 10%. For each starting point it is assumed that the portions of methane and ethane in
the mixture do not exceed 98% of the total composition of the natural gas stream (except in
Starting Point 1 where the summed fractions yield 99.50%). For each value of the methane
mole fraction, there is a number of starting points as the ethane fraction is varied in order
to yield up to 98% of the overall composition (see Table A1.3). The proportions between
the components in the composition of the natural gas stream are thus also modified. The
remainder of the stream is composed of propane, n-butane, nitrogen and i-butane in
proportions of 50%, 16%, 24% and 10%, respectively.
Table A1. 3. Starting points for the natural gas stream optimisation.
Composition - Mole fraction
Starting
point C1 C2 C3 n-C4 N2 i-C4
Normalised
flow rate
Normalised
inlet pressure
1 0.98 0.0150 0.0025 0.0008 0.0012 0.0005
0.78 0.83
2 0.89 0.92
3 0.96 0.0200 0.0100 0.0032 0.0048 0.0020
0.78 0.83
4 0.89 0.92
5
0.94
0.0200 0.0200 0.0064 0.0096 0.0040 0.78 0.83
6 0.89 0.92
7 0.0400 0.0100 0.0032 0.0048 0.0020
0.78 0.83
8 0.89 0.92
9
0.92
0.0200 0.0300 0.0096 0.0144 0.0060 0.78 0.83
10 0.89 0.92
11 0.0400 0.0200 0.0064 0.0096 0.0040
0.78 0.83
12 0.89 0.92
13 0.0600 0.0100 0.0032 0.0048 0.0020
0.78 0.83
14 0.89 0.92
15
0.90
0.0200 0.0400 0.0128 0.0192 0.0080 0.78 0.83
16 0.89 0.92
17 0.0400 0.0300 0.0096 0.0144 0.0060
0.78 0.83
18 0.89 0.92
19 0.0600 0.0200 0.0064 0.0096 0.0040
0.78 0.83
20 0.89 0.92
21 0.0800 0.0100 0.0032 0.0048 0.0020
0.78 0.83
22 0.89 0.92
23
0.88
0.0200 0.0500 0.0160 0.0240 0.0100 0.78 0.83
24 0.89 0.92
25 0.0400 0.0400 0.0128 0.0192 0.0080
0.78 0.83
26 0.89 0.92
27 0.0600 0.0300 0.0096 0.0144 0.0060
0.78 0.83
28 0.89 0.92
29 0.0800 0.0200 0.0064 0.0096 0.0040
0.78 0.83
30 0.89 0.92
31 0.1000 0.0100 0.0032 0.0048 0.0020
0.78 0.83
32 0.89 0.92
Appendix 1 Determination of the Natural Gas Stream Conditions
155
For each initial composition of the natural gas stream, there are two values of the
normalised mass flow rate and two values of the normalised inlet pressure (see Table
A1.3). The two normalised values selected for these variables, are equally-spaced values
between their corresponding upper and lower bounds.
In the optimisations, when the pressure drop profile of the natural gas stream is assumed
linearly dependent on the temperature change (Equation A1.1) and also when the pressure
drop profile is assumed linearly dependent on the heat rejected (Equation A1.2), the inlet
pressure of the natural gas stream was initially varied between 50 bar and 55 bar (i.e.
between normalised values 0.90 and 1.0) in order to represent a similar scenario to that
stated by Lee (2001). However, as shown in Figure A1.3a, the resulting natural gas stream
was not fully covering the condensing zone (indicated by the change of slope) of the
original T–H profile. The condensing enthalpy change (ΔHCond) of the natural gas stream
increases as its pressure level decreases, as illustrated with two isobars in a temperature–
enthalpy diagram in Figure A1.3b. Thus, during the optimisations, the lower bound of the
natural gas inlet pressure was decreased to 40 bar (i.e. normalised value of 0.73) to allow
the condensing enthalpy change in the constructed natural gas stream to increase.
Figure A1. 3. a) Constructed profile (squares, Starting Point 16) failing to fully cover the condensing zone;
b) condensing enthalpy change increases as pressure of the stream decreases.
After the optimisation, the selected combination of stream composition, inlet pressure,
overall mass flow rate and pressure drop profile, is that with the lowest sum of squared
errors (SSE) between the enthalpy values of the optimised T–H profile and the published
data. Figure A1.4 shows the SSE values plotted for each of the starting points after the
optimisation and for each of the three pressure drop profiles assumed for the natural gas
stream across the MSHE. SSE values when the pressure drop profile is assumed as linearly
Appendix 1 Determination of the Natural Gas Stream Conditions
156
dependent on the temperature change (ΔP/ΔT = k) are displayed in squares; SSE values
when the pressure drop profile is assumed as linearly dependent on the heat rejected
(ΔP/ΔH = k) are shown in triangles; SSE values when zero pressure drop in the MSHE is
assumed (ΔP = 0) are displayed in diamonds.
The approach assuming constant pressure during the liquefaction of the natural gas stream
(i.e. a zero pressure drop across the MSHE) resulted in lower values of SSE compared to
those achieved when pressure drop was modelled based either on temperature change or
heat rejection dependence. Multiple starting points (2-6, 10-12, 15, 18-20, 22, 25, 28 and
32) yielded the lowest SSE value (68×103), and also resulted in the same x* vector of
values for stream composition, normalised flow rate and normalised inlet pressure. The
optimisation is considered successful since a wide range of starting points with different
initial values ended up with the same solution.
Figure A1. 4. Sum of squared errors (SSE) of enthalpy values for each starting point and for each pressure
drop profile assumption.
The values of the vector x* that achieved the closest match to the published T–H profile
are those presented in Table A1.4. The values of normalised flow rate and normalised inlet
pressure represent a mass flow rate of 24.03 kg·s-1
and an inlet pressure of 43.86 bar,
respectively, of the natural gas stream. The mass flow rate obtained (24.03 kg·s-1
) is
equivalent to LNG production of 0.75 million t per annum, which is within the production
limits of liquefied natural gas at small scale, i.e. up to 1 million t per annum (Mokhatab et
al., 2014b, Ch. 3.3).
65×10³
70×10³
75×10³
80×10³
85×10³
90×10³
0 5 10 15 20 25 30 35
SS
E
Starting Point Number
ΔP = 0
ΔP/ΔT = k
ΔP/ΔH = k
Appendix 1 Determination of the Natural Gas Stream Conditions
157
Table A1. 4. Optimised values of the vector x* for the natural gas stream (zero pressure drop in the MSHE).
C1 C2 C3 n-C4 N2 i-C4 Normalised
flow rate
Normalised
inlet pressure
0.9000 0.0940 0.0047 0.0013 0.0000 0.0000 0.8009 0.7975
The molar flow rate of the optimised natural gas is 1.37 kmol·s-1
, which is greater than that
stated by Zheng (2009) (1 kmol·s-1
). The mass flow rate of the optimised natural gas
stream (24.03 kg·s-1
) is higher than the mass flow rate employed by Remeljej and Hoadley
(2006) (22.60 kg·s-1
) by around 6%. Regarding the composition, Table A1.5 shows the
composition of the optimised natural gas stream and that provided by Remeljej and
Hoadley (2006). The molar fraction of methane is decreased by nearly 0.070, in the
optimised natural gas stream, whereas the mole fractions of ethane and propane are
increased by 0.065 and 0.004, respectively. In the optimised natural gas stream, the low
fraction of n-butane (0.0013) and i-butane (0.0000) is expected since the natural gas stream
undergoes a series of separation processes including heavy hydrocarbons (C4+) removal
prior to liquefaction; the low mole fraction of nitrogen (0.0000) is also expected, as
nitrogen is usually in low concentrations (mole fraction < 0.0100) in natural gas to be
liquefied (Kidnay and Parrish, 2006b, Ch. 13.2).
Table A1. 5. Composition of the optimised natural gas stream and that of Remeljej and Hoadley (2006).
Natural gas composition [mole fraction]
C1 C2 C3 n-C4 N2 i-C4
Optimised natural gas stream 0.9000 0.0940 0.0047 0.0013 0.0000 0.0000
Remeljej and Hoadley (2006) 0.9693 0.0294 0.0006 0.0001 0.0006 0.0000
Table A1.6 compares the T–H profile of the natural gas stream obtained from the
optimisation against that originally published by Lee (2001) and later used by Zheng
(2009) to optimise the CryoMan process. The largest deviation in the T–H profile of the
optimised natural gas stream occurs at –82°C (the difference is 3% compared to the
published data). Figure A1.5 displays the T–H profile obtained with the optimisation
(squares) as compared to the published data (line with diamonds). Figure A1.5
demonstrates that a good agreement between the T–H curves is achieved.
Appendix 1 Determination of the Natural Gas Stream Conditions
158
Table A1. 6. Temperature–enthalpy data comparison between the optimised stream and the published data.
Segment Supply
Temperature [°C]
Target
Temperature [°C]
Cumulative ΔH [kW] (Lee, 2001)
Cumulative ΔH [kW]
(optimisation) Difference
[%]
1.1 25.00 –06.03 20178.8 20147.6 0.2
1.2 –06.03 –34.09 18317.3 18255.9 0.3
1.3 –34.09 –57.65 16353 16427.2 –0.5
1.4 –57.65 –70.10 14468 14448.0 0.1
1.5 –70.10 –74.55 11978 11911.7 0.6
1.6 –74.55 –82.26 10198 10225.2 –0.3
1.7 –82.26 –96.50 7114 7330.0 –3.0
1.8 –96.50 –115.00 5690 5616.2 1.3
1.9 –115.00 –163.00 3840 3846.2 –0.2
Figure A1. 5. Temperature–enthalpy profiles comparison between data published by Lee (2001) and the
stream constructed by optimisation.
The values of the composition, flow rate and inlet pressure obtained from the optimisation,
are employed in HYSYS to simulate the natural gas stream used by Zheng (2009) to
optimise the CryoMan process. This back-calculation allowed the natural gas stream to be
defined and so evaluate the novel refrigeration cycles with the same production rate (0.75
million t per annum), and compare their performance (especially specific shaft power
demand) against that of the CryoMan process (see Chapter 4).
-200
-150
-100
-50
0
50
0 5 10 15 20 25
Tem
per
atu
re [
°C]
ΔH [MW]
Optimisation
Lee (2001)
Appendix 2 Bypass design and Two Flash Levels design: Optimisation Results
159
Appendix 2 – Bypass design and Two Flash Levels design:
Optimisation Results
A2.1 Optimisation results of the Bypass design
Figure A2. 1. Bypass refrigeration cycle.
Figure A2. 2. Optimisation progression of the Bypass design (run 1).
Appendix 2 Bypass design and Two Flash Levels design: Optimisation Results
160
Table A2. 1. Optimised operating variables of the Bypass design (run 1).
Label in Figure A2.1 Process variable Value A Refrigerant flow rate [kmol·s
-1] 3.43
A Composition [mole fraction]: - Methane 0.2250 - Ethane 0.3743 - Propane 0.1770 - n-Butane 0.1509 - Nitrogen 0.0728 A Discharge pressure [bar] 39.4 A Bypass fraction 0.0868 B Vapour flow rate fraction 0.8527 C Liquid flow rate fraction 0.1106 Stream 1 Stream 2 Bypass D Precooling temperature [°C] –166.3 –97.2 –99.8 E Expansion pressure [bar] 1.21 7.48 20.20 F MSHE outlet temperature [°C] 22.1 25.0 25.0 Performance indicators Value
Number of compression stages 4 ΔT
MIN [°C] 5.0
Total shaft power [MW] 25.22
Figure A2. 3. Optimisation progression of the Bypass design (run 2).
Appendix 2 Bypass design and Two Flash Levels design: Optimisation Results
161
Table A2. 2. Optimised operating variables of the Bypass design (run 2).
Label in Figure A2.1 Process variable Value A Refrigerant flow rate [kmol·s
-1] 3.56
A Composition [mole fraction]: - Methane 0.2239
- Ethane 0.3797
- Propane 0.1897
- n-Butane 0.1295
- Nitrogen 0.0772
A Discharge pressure [bar] 42.8
A Bypass fraction 0.1142
B Vapour flow rate fraction 0.8835
C Liquid flow rate fraction 0.0821
Stream 1 Stream 2 Bypass D Precooling temperature [°C] –167.5 –66.5 –72.9
E Expansion pressure [bar] 1.32 8.12 22.04
F MSHE outlet temperature [°C] 23.9 21.5 24.7
Performance indicators Value
Number of compression stages 4 ΔT
MIN [°C] 5.0
Total shaft power [MW] 25.37
Figure A2. 4. Optimisation progression of the Bypass design (run 3).
Appendix 2 Bypass design and Two Flash Levels design: Optimisation Results
162
Table A2. 3. Optimised operating variables of the Bypass design (run 3).
Label in Figure A2.1 Process variable Value A Refrigerant flow rate [kmol·s
-1] 3.66
A Composition [mole fraction]: - Methane 0.2495
- Ethane 0.3697
- Propane 0.1565
- n-Butane 0.1502
- Nitrogen 0.0741
A Discharge pressure [bar] 42.6
A Bypass fraction 0.0718
B Vapour flow rate fraction 0.7205
C Liquid flow rate fraction 0.1459
Stream 1 Stream 2 Bypass D Precooling temperature [°C] –163.6 –96.6 –90.8
E Expansion pressure [bar] 1.23 9.52 22.94
F MSHE outlet temperature [°C] 24.7 23.5 24.9
Performance indicators Value
Number of compression stages 4 ΔT
MIN [°C] 5.0
Total shaft power [MW] 25.38
A2.2 Optimisation results of the Two Flash Levels design
Figure A2. 5. Two Flash Levels refrigeration cycle.
Appendix 2 Bypass design and Two Flash Levels design: Optimisation Results
163
Figure A2. 6. Optimisation progression of the Two Flash Levels design (run 1).
Table A2. 4. Optimised operating variables of the Two Flash Levels design (run 1).
Label in Figure A2.5 Process variable Value A Refrigerant flow rate [kmol·s
-1] 3.42
A Composition [mole fraction]: - Methane 0.2372 - Ethane 0.4013 - Propane 0.0587 - n-Butane 0.2285 - Nitrogen 0.0743 A Discharge pressure [bar] 41.1 H Pressure 2nd flash unit [bar] 26.7 G Liquid fraction to 2nd flash unit 0.2082 Stream 1 Stream 2 Stream 3 Stream 4 D Precooling temperature [°C] –162.3 –110.2 –96.5 –56.3 E Expansion pressure [bar] 1.2 7.8 4.7 11.2 F MSHE outlet temperature [°C] 19.6 24.0 18.9 25.0
B, I Vapour flow rate fraction 0.7506 0.2494 0.2975 0.7025 C, J Liquid flow rate fraction 0.2193 0.5725 0.2559 0.7441
Performance indicators Value
Number of compression stages 6 ΔT
MIN [°C] 5.0
Total shaft power [MW] 25.40
Appendix 2 Bypass design and Two Flash Levels design: Optimisation Results
164
Figure A2. 7. Optimisation progression of the Two Flash Levels design (run 2).
Table A2. 5. Optimised operating variables of the Two Flash Levels design (run 2).
Label in Figure A2.5 Process variable Value A Refrigerant flow rate [kmol·s
-1] 3.38
A Composition [mole fraction]:
- Methane 0.2569
- Ethane 0.3536
- Propane 0.1271
- n-Butane 0.1792
- Nitrogen 0.0832
A Discharge pressure [bar] 42.75
H Pressure 2nd flash unit [bar] 33.4
G Liquid fraction to 2nd flash unit 0.1508
Stream 1 Stream 2 Stream 3 Stream 4 D Precooling temperature [°C] –162.8 –106.3 –74.7 –54.2
E Expansion pressure [bar] 1.3 8.6 4.9 13.3
F MSHE outlet temperature [°C] 24.9 25.0 25.0 24.9
B, I Vapour flow rate fraction 0.7256 0.2744 0.4717 0.5283
C, J Liquid flow rate fraction 0.2743 0.5749 0.3599 0.6401
Performance indicators Value
Number of compression stages 6 ΔT
MIN [°C] 5.0
Total shaft power [MW] 25.59
Appendix 2 Bypass design and Two Flash Levels design: Optimisation Results
165
Figure A2. 8. Optimisation progression of the Two Flash Levels design (run 3).
Table A2. 6. Optimised operating variables of the Two Flash Levels design (run 3).
Label in Figure A2.5 Process variable Value A Refrigerant flow rate [kmol·s
-1] 3.41
A Composition [mole fraction]:
- Methane 0.2327
- Ethane 0.4000
- Propane 0.0537
- n-Butane 0.2377
- Nitrogen 0.0759
A Discharge pressure [bar] 40.22
H Pressure 2nd flash unit [bar] 33.67
G Liquid fraction to 2nd flash unit 0.2301
Stream 1 Stream 2 Stream 3 Stream 4 D Precooling temperature [°C] –162.2 –109.6 –44.4 –47.3
E Expansion pressure [bar] 1.2 7.1 9.6 12.5
F MSHE outlet temperature [°C] 23.1 24.9 24.8 24.9
B, I Vapour flow rate fraction 0.7542 0.2458 0.0141 0.9859
C, J Liquid flow rate fraction 0.2394 0.5305 0.0515 0.9485
Performance indicators Value
Number of compression stages 6 ΔT
MIN [°C] 5.0
Total shaft power [MW] 25.46
Appendix 3 Optimisation of the CryoMan Process with Six Compression Stages
166
Appendix 3 – Optimisation of the CryoMan Process with Six
Compression Stages
A3.1 Introduction
The optimised Two Flash Levels design – presented in the case study in Chapter 5 –
showed that operating cost savings (from its corresponding shaft power savings) of £0.54
million per annum are achieved, compared to the CryoMan process (Zheng, 2009). In the
Two Flash Levels design, the compression of the overall refrigerant stream is performed
with six compression stages, i.e. two additional compression stages compared to the
CryoMan process. There is an intercooling stage after each compression stage (according
to the model described in Section 3.3.1), that helps reducing the shaft power demand by
decreasing the temperature and volumetric flow rate of the refrigerant stream at the inlet of
the next compression stage. The Two Flash Levels thus takes advantage of the intercooling
associated with the two additional compression stages.
However, the two extra compression stages (compared to four in the CryoMan process)
would be expected to significantly increase the capital investment of the Two Flash Levels
design because the compressors are the most expensive equipment in the refrigeration
cycle (Mokhatab et al., 2014b, Ch. 3.2).
The Two Flash Levels design is thus compared – in this appendix – to the CryoMan
process optimised to have also six compression stages with their corresponding
intercooling stages. This comparison is in order to evaluate whether the operating cost
savings achieved by the Two Flash Levels design (see Section 5.6) can offset the increase
in capital investment resulting from additional compression stages.
The rationale for this study is as follows: the optimised Two Flash Levels design, which
has six compression stages, is compared to the CryoMan process optimised with six
compression stages. This study would allow comparing both designs in the same basis of
complexity of the compressor. Thus, any shaft power savings yielded by the Two Flash
Levels design, relative to the CryoMan process with six compression stages, would now
represent operating cost savings that can be achieved as a result of the structural
modification proposed (two new streams from a second flash unit), regardless of the
intercooling stages. That is, the operating cost savings achieved by the Two Flash Levels
Appendix 3 Optimisation of the CryoMan Process with Six Compression Stages
167
design (see Section 5.6) would justify the increase in capital investment associated with the
additional compression stages.
A3.2 Problem formulation
First, to model the CryoMan process with additional compression stages, either the Bypass
design model (Section 3.3.1) or the Two Flash Levels design model (Section 3.3.2) can be
used. If the Bypass design model is used, then Equation 3.1 (calculation of the flow rate in
the Bypass Stream) is forced to be zero. If the Two Flash Levels design model is selected,
Equation 3.41 (calculation of the flow rate fed to the second flash unit) is forced to be zero.
Similar to the Bypass design and the Two Flash Levels design, the objective function of
the optimisation is the minimisation of the total shaft power demand for refrigerant
compression. The model of the Bypass configuration is used to simulate the CryoMan
process. Thus, the objective function (Equation A3.1) and the mathematical formulation
for the optimisation of the CryoMan process are as follows:
Minimise: 𝑊𝑇𝑜𝑡𝑎𝑙 = ∑ 𝑊𝑖𝑆𝑡𝑔(Φ)𝑛
𝑖=1 𝑖 = 1,2, … , 𝑛 (A3.1)
Subject to: Equations 3.1 to 3.35 (for the Bypass design)
α = 0 (Bypass flow rate fraction) (A3.2)
Δ𝑇𝑀𝐼𝑁 ≥ 5℃ (A3.3)
𝑃𝑅𝐴𝑇 ≤ 2 (A3.4)
∑ 𝑥𝑗 = 1𝑚𝑗=1 𝑥𝑗 ∈ 𝑋𝑀𝑅 𝑗 = 1,2, … , 𝑚 (A3.5)
𝑙𝑏 ≤ Φ ≤ 𝑢𝑏 (A3.6)
where
Φ = [𝑋𝑀𝑅 , 𝑃𝐻𝑖𝑔ℎ, 𝑓𝑉𝑎𝑝, 𝑓𝐿𝑖𝑞, 𝑇𝐴𝑃𝑘 , 𝑃𝐴𝐸
𝑘 , 𝑇𝑜𝑢𝑡𝑘 ]
𝑘 = 𝐿𝑃 𝑆𝑡𝑟𝑒𝑎𝑚, 𝐻𝑃 𝑆𝑡𝑟𝑒𝑎𝑚 (A3.7)
Φ is a vector that includes the values of the operating variables to be optimised, i.e.
refrigerant composition, refrigerant streams pressure level, precooling temperatures and
outlet temperatures from the multi-stream heat exchanger (MSHE), vapour and liquid flow
rate fractions, and compressor discharge pressure (Equation A3.7).
Appendix 3 Optimisation of the CryoMan Process with Six Compression Stages
168
Note that, this time, the maximum pressure ratio for each compression stage (Equation
A3.4) is 2, in order to achieve the compression of the overall refrigerant stream with six
compression stages.
Equation A3.2 ensures that the flow rate of the Bypass Stream is zero and, therefore, the
adapted CryoMan process is modelled. Equation A3.3 guarantees feasible heat transfer
inside the MSHE, i.e. heat transfer between hot and cold streams with a minimum
temperature difference of 5°C. Equation A3.5 states that the sum of mole fractions of the
components, in the refrigerant mixture, is unity. Each operating variable (e.g. refrigerant
composition, streams pressure level) is varied within a lower and upper bounds (Equation
A3.6).
The CryoMan process is optimised using WORK software. The Genetic Algorithm method
is used for the optimisation. The values of the optimisation parameters (shown in Table
A3.1), which are employed in the optimisation of both the Bypass design and the Two
Flash Levels design (see Section 5.2), are also employed in the optimisation of the
CryoMan process.
Table A3. 1. Genetic Algorithm parameters for optimisation of the CryoMan process.
Parameter Population Size Maximum Generations Crossover Probability Mutation Rate
Value 250 400 0.85 0.01
A3.3 Problem statement
The optimisation is aimed at minimising the total shaft power demand needed for
refrigerant compression in the liquefaction of a given natural gas stream. The case study is
that first published by Lee (2001), and later employed by Zheng (2009) to optimise the
CryoMan process (in which refrigerant compression is achieved with four compression
stages). The natural gas stream enters the MSHE at 25°C and leaves as LNG at –163°C
(the LNG production is assumed as 0.75 million t per annum). The data of the natural gas
stream is provided in Table A3.2 as a temperature–enthalpy profile.
In order to keep consistency with the optimisation performed by Zheng (2009), the
following assumptions are adopted: the refrigerant stream mixture comprises methane,
ethane, propane, n-butane and nitrogen. The compressor isentropic efficiency is assumed
as 80%. Zero pressure drop of the refrigerant streams inside of the MSHE is assumed. The
minimum temperature approach inside the MSHE for feasible heat transfer is 5°C. Physical
Appendix 3 Optimisation of the CryoMan Process with Six Compression Stages
169
and thermodynamic properties of the refrigerant streams (e.g. temperatures, enthalpies) are
calculated using Peng–Robinson equation of state by interfacing with Aspen HYSYS v8.2
(Aspen Technology Inc., 2013).
Table A3. 2. Natural gas stream data (temperature–enthalpy profile).
Segment Supply
Temperature [°C]
Target
Temperature [°C] ΔH [kW] CP [kW·K
-1]
1.1 25.00 –06.03 –1861.5 60
1.2 –06.03 –34.09 –1964.3 70
1.3 –34.09 –57.65 –1885.0 80
1.4 –57.65 –70.10 –2490.0 200
1.5 –70.10 –74.55 –1780.0 400
1.6 –74.55 –82.26 –3084.0 400
1.7 –82.26 –96.50 –1424.0 100
1.8 –96.50 –115.00 –1850.0 100
1.9 –115.00 –163.00 –3840.0 80
A3.4 Optimisation of the CryoMan process
Figure A3.1 shows the minimisation of the objective function (i.e. total shaft power
demand) as the optimisation progresses. According to Figure A3.1, the objective function
reaches a steady value of 24.94 MW of total shaft power demand.
Figure A3. 1. Objective function progression in the optimisation of the operating variables of the CryoMan
process.
The optimised values of the operating variables in the CryoMan process are shown in
Table A3.3 (the ‘Label’ column in Table A3.3 refers to the letters in Figure A3.2). The
total shaft power demand is 24.94 MW, i.e. 0.2883 kWh·kg-1
of LNG.
Appendix 3 Optimisation of the CryoMan Process with Six Compression Stages
170
Compared to the CryoMan process in which refrigerant compression is achieved with four
compression stages (26.05 MW), the two additional compression stages allowed a
reduction in the total shaft power demand of 4.3%.
Table A3. 3. Operating variables of the optimised CryoMan process with six and four compression stages.
Six compression stages Four compression stages
Label in Figure A3.2 Process variable Value Value
A Refrigerant flow rate [kmol·s-1
] 3.15 3.21
A Composition [mole fraction]:
- Methane 0.2362 0.2288
- Ethane 0.3739 0.3703
- Propane 0.1640 0.1684
- n-Butane 0.1407 0.1517
- Nitrogen 0.0852 0.0808
A Discharge pressure [bar] 47.7 48.3
B Vapour flow rate fraction 0.8640 0.8830
C Liquid flow rate fraction 0.2020 0.2230
LP Stream HP Stream LP Stream HP Stream
D Precooling temperature [°C] –162.6 –88.7 –164.6 –79.0
E Expansion pressure [bar] 1.3 8.8 1.2 9.6
F MSHE outlet temperature [°C] 22.4 24.9 21.2 24.3
Performance indicators Value Value
Number of compression stages 6 4
ΔTMIN
[°C] 5.0 5.0
Total shaft power [MW] 24.94 26.05
Specific shaft power [kWh·kg-1
LNG] 0.2883 0.3011
Figure A3. 2. The CryoMan refrigeration cycle.
In the CryoMan process with six compression stages, the pressure level of LP Stream is
increased from 1.2 bar to 1.3 bar (which helps reduce the shaft power demand, see Section
4.4.2) at the expense of increasing the mole fraction of the light components (mole fraction
Appendix 3 Optimisation of the CryoMan Process with Six Compression Stages
171
of methane is increased from 0.2288 to 0.2362; mole fraction of nitrogen is increased from
0.0808 to 0.0852) which increases the shaft power demand (as discussed in Section 4.4.2).
However, because of the design constraint of maximum pressure ratio (Equation A3.4), the
pressure level of the refrigerant streams lead to compression of the overall refrigerant
stream with six compression stages and thus, take advantage of the intercooling associated
with the two additional compression stages.
The Two Flash Levels design can now be compared to the CryoMan process on the basis
of equal number of compression stages and equal number of intercooling stages. The total
shaft power demand achieved by the CryoMan process, when compression of the overall
refrigerant stream is performed with six compression stages, is 24.94 MW; the total shaft
power demand of the Two Flash Levels design (also with six compression stages, see
Section 5.5) is 25.40 MW. That is, the power demand in the Two Flash Levels design is
1.8% higher than that of the CryoMan process. Therefore, shaft power savings are no
longer achieved with the Two Flash Levels design (relative to the CryoMan process with
six compression stages).
It can be then concluded that the shaft power demand achieved by the optimised Two Flash
Levels design – 25.40 MW – in the case study presented in Chapter 5, is significantly
affected by the intercooling associated with the two additional compression stages,
compared to the CryoMan process with four compression stages.
The comparison suggests that the shaft power savings and corresponding operating cost
savings achieved by the Two Flash Levels design, compared to the CryoMan process with
four compression stages, are achieved mainly as a result of cooling the refrigerant stream
after each compression stage. The structural modification proposed, i.e. two new
refrigerant streams created from a second flash unit, does not show a significant effect
towards bringing shaft power demand savings. Thus, the operating cost savings achieved
by the Two Flash Levels design – shown in Chapter 5 – would not justify the increase in
capital costs that the two additional compression stages represent.
References
172
References
Air Products and Chemicals Inc. (2013). LNG technology brochure [Online]. Available:
http://www.airproducts.com/~/media/downloads/brochure/L/en-lng-brochure-and-data-
sheets.pdf [Accessed 23 April 2016].
Aspen Technology Inc. (2013). Aspen HYSYS v8.2. Bedford, USA.
Bäck, T., Fogel, D. B. & Michalewicz, Z. (1997). Chapter B1 - Evolutionary Algorithms
and Their Standard Instances. Handbook of Evolutionary Computation. Bristol, UK: IOP
Publishing Ltd and Oxford University Press.
Borgnakke, C. & Sonntag, R. (2009). Chapter 11 - Power and Refrigeration Systems with
Phase Change. Fundamentals of Thermodynamics. NJ, USA: John Wiley & Sons, Ltd.
BP Global (2015). BP Statistical Review of World Energy.
Cao, W.-S., Lu, X.-S., Lin, W.-S. & Gu, A.-Z. (2006). Parameter comparison of two small-
scale natural gas liquefaction processes in skid-mounted packages. Applied Thermal
Engineering, 26(8–9), 898-904.
Castillo, L. & Dorao, C. A. (2010). Influence of the plot area in an economical analysis for
selecting small scale LNG technologies for remote gas production. Journal of Natural Gas
Science and Engineering, 2(6), 302-309.
Del Nogal, F., Kim, J.-K., Perry, S. & Smith, R. (2008). Optimal design of mixed
refrigerant cycles. Industrial & Engineering Chemistry Research, 47(22), 8724-8740.
Del Nogal, F. L. (2006). Optimal Design and Integration of Refrigeration and Power
Systems. Ph.D. Thesis, The University of Manchester, UK.
Dincer, I. & Kanoglu, M. (2010). Chapter 03 - Refrigeration System Components.
Refrigeration Systems and Applications. Chichester, UK: John Wiley & Sons, Ltd.
Edgar, T. F., Himmelblau, D. M. & Lasdon, L. S. (2001). Chapter 10 - Global
Optimization for Problems with Continuous and Discrete Variables. Optimization of
Chemical Processes. New York, USA: The McGraw-Hill Companies, Inc.
Elliot, J. R. & Lira, C. T. (1999). Chapter 06 - Engineering Equations of State for PVT
Properties. Introductory Chemical Engineering Thermodynamics. Upper Saddle River, NJ,
USA: Prentice Hall, Inc.
ESDU International plc (2006). Selection and costing of heat exchangers, plate-fin type.
ESDU 97006. London, UK.
Gaumer, L. & Newton, C. (1973). Combined cascade and multicomponent refrigeration
system and method. Air Products and Chemicals, Inc. US3763658.
Grootjans, H. F., Nagelvoort, R. K. & Vink, K. J. (2002). Liquefying a stream enriched in
methane. Shell Oil Company. US6370910.
References
173
Hanlon, P. C. (2001). Chapter 03 - Compressor Performance - Dynamic. Compressor
Handbook. USA: McGraw-Hill.
He, T. & Ju, Y. (2014a). Design and optimization of a novel mixed refrigerant cycle
integrated with NGL recovery process for small-scale LNG plant. Industrial &
Engineering Chemistry Research, 53(13), 5545-5553.
He, T. B. & Ju, Y. L. (2014b). Performance improvement of nitrogen expansion
liquefaction process for small-scale LNG plant. Cryogenics, 61(0), 111-119.
Hesselgreaves, J. E. (2001). Chapter 02 - Industrial Compact Exchangers. Compact Heat
Exchangers. Oxford, UK: Pergamon.
Hughes, J. H. (1971). Gas liquefaction by refrigeration with parallel expansion of the
refrigerant. Phillips Petroleum Company. US3581510.
Hwang, J.-H., Roh, M.-I. & Lee, K.-Y. (2013). Determination of the optimal operating
conditions of the dual mixed refrigerant cycle for the LNG FPSO topside liquefaction
process. Computers & Chemical Engineering, 49(0), 25-36.
IEA (2012). World Energy Outlook.
Kidnay, A. J. & Parrish, W. R. (2006a). Chapter 01 - Overview of Natural Gas Industry.
Fundamentals of Natural Gas Processing. Boca Raton, FL, USA: CRC Taylor & Francis.
Kidnay, A. J. & Parrish, W. R. (2006b). Chapter 13 - Liquefied Natural Gas. Fundamentals
of Natural Gas Processing. Boca Raton, FL, USA: CRC Taylor & Francis.
Kim, J.-K. & Zheng, X. (2011). Refrigeration process. The University of Manchester.
PCT/GB2011/050617.
Kumar, S., Kwon, H.-T., Choi, K.-H., Hyun Cho, J., Lim, W. & Moon, I. (2011). Current
status and future projections of LNG demand and supplies: A global prospective. Energy
Policy, 39(7), 4097-4104.
Kyle, B. G. (1999a). Chapter 03 - The Behavior of Fluids. Chemical and Process
Thermodynamics. Upper Saddle River, NJ, USA: Prentice Hall, Inc.
Kyle, B. G. (1999b). Chapter 15 - Physicomechanical Processes. Chemical and Process
Thermodynamics. Upper Saddle River, NJ, USA: Prentice Hall, Inc.
Lee, G.-C. (2001). Optimal Design and Analysis of Refrigeration Systems for Low
Temperature Processes. Ph.D. Thesis, UMIST, UK.
Lee, G. C., Smith, R. & Zhu, X. X. (2002). Optimal synthesis of mixed-refrigerant systems
for low-temperature processes. Industrial & Engineering Chemistry Research, 41(20),
5016-5028.
Li, Q. Y. & Ju, Y. L. (2010). Design and analysis of liquefaction process for offshore
associated gas resources. Applied Thermal Engineering, 30(16), 2518-2525.
References
174
Linnhoff, B. & Dhole, V. R. (1992). Shaftwork targets for low-temperature process design.
Chemical Engineering Science, 47(8), 2081-2091.
Ludwig, E. E. (2001). Chapter 12 - Compression Equipment (Including Fans). Applied
Process Design for Chemical & Petrochemical Plants. Gulf Professional Publishing.
Mokhatab, S., Mak, J. Y., Valappil, J. V. & Wood, D. A. (2014a). Chapter 01 - LNG
Fundamentals. Handbook of Liquefied Natural Gas. Kidlington, Oxford, UK: Gulf
Professional Elsevier Inc.
Mokhatab, S., Mak, J. Y., Valappil, J. V. & Wood, D. A. (2014b). Chapter 03 - Natural
Gas Liquefaction. Handbook of Liquefied Natural Gas. Kidlington, Oxford, UK: Gulf
Professional Elsevier Inc.
Mokhatab, S., Mak, J. Y., Valappil, J. V. & Wood, D. A. (2014c). Chapter 05 - Natural
Gas Liquefaction Cycle Enhancements and Optimization. Handbook of Liquefied Natural
Gas. Kidlington, Oxford, UK: Gulf Professional Elsevier Inc.
Montanez-Morantes, M. (2015). Operational Optimisation of Low-Temperature Energy
Systems. Ph.D. Thesis, The University of Manchester, UK.
Morosuk, T., Tesch, S., Hiemann, A., Tsatsaronis, G. & Bin Omar, N. (2015). Evaluation
of the PRICO liquefaction process using exergy-based methods. Journal of Natural Gas
Science and Engineering, (0).
Mortazavi, A., Somers, C., Hwang, Y., Radermacher, R., Rodgers, P. & Al-Hashimi, S.
(2012). Performance enhancement of propane pre-cooled mixed refrigerant LNG plant.
Applied Energy, 93(0), 125-131.
Peng, D.-Y. & Robinson, D. B. (1976). A new two-constant equation of state. Industrial &
Engineering Chemistry Fundamentals, 15(1), 59-64.
Poli, R., Langdon, W. B. & McPhee, N. F. (2008). Chapter 2 - Representation,
Initialisation and Operators in Tree-based GP. A field guide to Genetic Programming.
Published via http://lulu.com and freely available at http://www.gp-field-guide.org.uk.
Querol, E., González-Regueral, B. & Pérez-Benedito, J. L. (2013). Chapter 02 - Exergy
Concept and Determination. Practical Approach to Exergy and Thermoeconomic Analyses
of Industrial Processes. London, Heidelberg, New York, Dordrecht: Springer-Verlag
London.
Radermacher, R. (1989). Thermodynamic and heat transfer implications of working fluid
mixtures in Rankine cycles. International Journal of Heat and Fluid Flow, 10(2), 90-102.
Reap, E. (2015). The risk of hydraulic fracturing on public health in the UK and the UK’s
fracking legislation. Environmental Sciences Europe, 27(1), 1-7.
Remeljej, C. W. & Hoadley, A. F. A. (2006). An exergy analysis of small-scale liquefied
natural gas (LNG) liquefaction processes. Energy, 31(12), 2005-2019.
Shelton, M. R. & Grossmann, I. E. (1985). A shortcut procedure for refrigeration systems.
Computers & Chemical Engineering, 9(6), 615-619.
References
175
Shelton, M. R. & Grossmann, I. E. (1986). Optimal synthesis of integrated refrigeration
systems—I: Mixed-integer programming model. Computers & Chemical Engineering,
10(5), 445-459.
Shirazi, M. H. & Mowla, D. (2010). Energy optimization for liquefaction process of
natural gas in peak shaving plant. Energy, 35(7), 2878-2885.
Smith, R. (2005a). Chapter 01 - The Nature of Chemical Process Design and Integration.
Chemical Process Design and Integration. Chichester, UK: John Wiley & Sons, Ltd.
Smith, R. (2005b). Chapter 24 - Cooling and Refrigeration Systems. Chemical Process
Design and Integration. Chichester, UK: John Wiley & Sons, Ltd.
Swenson, L. K. (1977). Single mixed refrigerant, closed loop process for liquefying
natural gas. J. F. Pritchard and Company. US4033735.
The MathWorks Inc. (2013). MATLAB Version 2013a. Natick, USA.
U.K. Department of Energy & Climate Change. (2015). International industrial energy
prices [Online]. Available: https://www.gov.uk/government/statistical-data-
sets/international-industrial-energy-prices [Accessed 04 February 2016].
Vaidyaraman, S. & Maranas, C. D. (2002). Synthesis of mixed refrigerant cascade cycles.
Chemical Engineering Communications, 189(8), 1057-1078.
Whitmarsh, L., Nash, N., Upham, P., Lloyd, A., Verdon, J. P. & Kendall, J. M. (2015). UK
public perceptions of shale gas hydraulic fracturing: The role of audience, message and
contextual factors on risk perceptions and policy support. Applied Energy, 160, 419-430.
Yin, Q. S., Li, H. Y., Fan, Q. H. & Jia, L. X. (2008). Economic analysis of mixed-
refrigerant cycle and nitrogen expander cycle in small scale natural gas liquefier. AIP
Conference Proceedings, 985(1), 1159-1165.
Zheng, X. (2007). Design of Refrigeration and Power Systems (Presentation). Process
Integration Research Consortium Annual Meeting. Centre for Process Integration, The
University of Manchester, UK.
Zheng, X. (2009). Design and Integration of Refrigeration and Power Systems. Ph.D.
Thesis, The University of Manchester, UK.