Micro-Scale Natural Gas Liquefaction Processes: Design and Optimisation: A Literature Study for the Design and Optimisation of Micro-Scale Natural Gas (NG) Liquefaction Processes has been undertaken for the basis of a Masters dissertation on the subject. The content of this report stipulates the design execution plan in support of the final Master’s dissertation.
The objectives of this execution plan were to:
- Stipulate the basis of design;
- Identify and/or develop processes to be simulated and tested;
- Develop optimisation strategies for proposed processes;
- Discuss financial and economic analysis principles to be applied;
- Introduce basic engineering design strategies; and
- Develop a dissertation schedule.
A detailed research study has been completed on existing work specific to micro-scale (less than 275 tones per day or 5 million GJ per annum) liquefied natural gas (LNG) production facilities. Results were analysed, compared and summarised to serve as initial design conditions for proposed process designs.
1 Micro-Scale Natural Gas Liquefaction Processes: Basis of Design
1.1 Purpose of Design
Due to limited research being available on micro-scale LNG production plants and mixed preferences from different sources on which process technology is most suited for small- or “micro”-scale liquefaction, this work aims to establish what type of process (single mixed refrigeration (SMR) cycles or nitrogen expansion cycles) is most feasible considering production capacities between 13.5 and 275 tonnes per day (0.25 and 5 million gigajoules per annum). The design approach to be implemented is illustrated in Figure 1.
Figure 1: Block flow diagram illustration of the design approach
Optimal process models based on previous work has been identified for SMR and NEC cycles. Process models will be formulated using Aspen HYSYS. Process optimisation will then be applied with the main focus on minimising energy consumption. A costing analysis in terms of operational and capital expenditures will follow to feed into a detailed economic model.
Financial and economic feasibility evaluations will be performed to determine optimal solutions for production scales considered. Feasibility will be measured in terms of process simplicity, net present value (NPV) and accounting rate of return (ARR) on an identical basis.
Basic engineering design of the most feasible process for a 55 tonnes per day (approximately 1 million GJ per annum) facility will then be conducted. This will form the basis of a generic design that can be adjusted, detailed and applied in the industry within the proposed production scale range.
1.2 Unit Conversions
The below-tabulated units will form the basis for all conversion factors to be used in the design.
Table 1: Unit Conversion Table
Value | Mass (tonnes) | Energy (mmBtu) | The volume of LNG (m3) | The volume of NG (m3) | The volume of NG (ft3) | Energy (GJ) | |
Mass (tonnes) | 1.00 | 1.00 | 53.38 | 2.22 | 1 300.00 | 45 909.00 | 56.32 |
Energy (mmBtua) | 1.00 | 0.02 | 1.00 | 0.04 | 24.36 | 860.10 | 1.06 |
The volume of LNG (m3) | 1.00 | 0.45 | 24.02 | 1.00 | 585.00 | 20 659.00 | 25.34 |
The volume of NG (m3) | 1 000.00 | 0.77 | 41.10 | 1.70 | 1 000.00 | 35 310.00 | 43.36 |
The volume of NG (ft3) | 1 000.00 | 0.02 | 1.17 | 0.05 | 28.30 | 1 000.00 | 1.23 |
a million British thermal units
1.3 Micro-Scale Natural Gas Liquefaction Processes: Scope of Work
1.3.1 Battery Limits
The battery limits for the design are the inlet to the liquefaction process (known as the “cold box” of an LNG plant) and the outlet of the cryogenic heat exchanger(s)/expansion valve(s) before LNG storage. Figure 2 illustrates the design battery limits within an LNG plant block flow diagram context.
The battery limits were selected because the liquefaction process design does not change significantly with a change in feed gas conditions. The plant pre-treatment facilities ensure that the feed gas is “cleaned” to meet required inlet conditions. Only when heavier hydrocarbons are becoming significant in the feed gas must cryogenic distillation be designed for the liquefaction processes.
Figure 2: Process design boundary conditions (battery limit illustration)
1.3.2 Design On-Stream Time and Turndown Ratio
The capacity of the designs will be based on a total available production time of 24/7, 330 days per calendar year (with an allowance of one-month maintenance per annum).
The equipment and valves will be designed for a turndown ratio of 50%, allowing for flexibility during a change in market demands and initial ramp-ups.
1.4 Micro-Scale Natural Gas Liquefaction Processes: Design Data
1.4.1 LNG Product Specifications
The critical design conditions for the LNG product are summarised in Table 2. Required LNG product specifications following British Standard (BS EN 1160:1997 Installations and equipment for liquefied natural gas – general characteristics of liquefied natural gas) are listed in Table 3. Industry standards require the LNG product specifications to be within “lean” and “rich” given ranges.
Table 2: Required LNG product conditions
LNG Product Condition | Unit | Specification |
Flow low rate | Tonnes per day | 13.5 – 275 |
Temperature | ºC | < -162 (or below dew point) |
Pressure | barg | < 16 barg |
Table 3: LNG product specification (as per BS EN 1160:1997)
Composition (mol%) | LNG (Lean) | LNG (Average) | LNG (Rich) | |
Nitrogen (N2) | 0.5 | 1.79 | 0.36 | |
Methane (CH4) | 97.5 | 93.9 | 87.2 | |
Ethane (C2H6) | 1.8 | 3.26 | 8.61 | |
Propane (C3H8) | 0.2 | 0.69 | 2.74 | |
i-Butane (i-C4H10) | – | 0.12 | 0.42 | |
n-Butane (n-C4H10) | – | 0.15 | 0.65 | |
Pentane (C5H12) | – | 0.09 | 0.02 | |
Molecular Weight | (kg/kmol) | 16.41 | 17.07 | 18.52 |
Bubble Point Temperature @ 101.325kPa | (ºC) | -162.9 | -166.5 | -161.5 |
Density @ Bubble Point Temperature @ 101.325kPa | (kg/m3) | 432.7 | 449.5 | 464.9 |
Higher Heating Value | (MJ/Sm3) | 38.22 | 38.76 | 42.59 |
Lower Heating Value | (MJ/Sm3) | 34.43 | 34.95 | 38.51 |
Wobbe Index | (MJ/Sm3) | 50.73 | 50.43 | 53.19 |
1.4.2 Utility Specifications
The following are general design water and electrical specifications for South African industry standards and will be assumed available.
Table 4: Electrical availability
Volt | Phase | Frequency |
6.85kV | 3 | 50 Hz |
525V | 3 | 50 Hz |
Table 5: Conditions for available cooling water
Cooling water | Value |
Mechanical design pressure (kPag) | 700 (at grade) |
Nominal supply pressure (kPag) | 450 (at grade) |
Vacuum design pressure (kPag) | -85 |
Maximum available exchanger ΔP (kPag) | 100 |
Supply temperature (oC) | 27 |
Max. allowable exchanger outlet temp. (oC) | 40 |
Table 6: Conditions for other available water streams
Stream | Min Press @ Grade, (kPag) | Mech. Design Press, (kPag) | Low/High Temp’s, (oC) |
Potable | 400 | 1000 | 10/40 |
Hydrotest water | TBDa | TBD | TBD |
Firewater | 800 | 1500 | 10/40 |
a TBD – to be determined
Table 7 specifies the air and inert gas (nitrogen) specifications that are industry standard and easily accessible. These are assumed to be available.
Table 7: Inert gas and air industry standard specifications
Fluid | Noma Pressure (kPag) |
Normb Tempd (oC) |
Mechc Design Pressure (kPag) |
Mech Design Temp (oC) |
Dew Point (oC) |
Instrument Air | 700 | 35 | 1000 | 60 | -20 |
Utility Air | 700 | 35 | 1000 | 60 | 35 |
Nitrogen (LPe) | 300 | 35 | 1000 | 60 |
a Nominal; b Normal; c Mechanical; d Temperature; e Low Pressure
1.5 General
Design regulations, codes and standards are project-specific and must be addressed during the detailed design of a plant. Site information concerning environmental and climatic conditions are critical factors and must also be addressed during detailed design. Control and instrumentation design strategies are excluded from this work.
2 Micro-Scale Natural Gas Liquefaction Processes: Process Simulation Strategies
2.1 Previous work
2.1.1 Single Mixed Refrigerant Cycles
The SMR process, patented as the Prico Process (Fan & Oelschlegel, 1986), was selected as one of the investigated processes due to its simplicity (low number of equipment units and process variables) and popularity for small-scale LNG plants.
Previous work has been focused on the design and optimisation of SMR cycles. It was found that key decision variables are: flow rate and composition of mixed refrigerants, partition temperature and operating pressures before and after compressors. This is because the most important unit operations’ (compressors and heat exchangers) designs are determined by given variables.
A simplified process schematic of the process under discussion is illustrated in Figure 3.
Figure 3: Simplified process schematic of the SMR cycle (Austbo, Wahl & Gundersen, 2013)
In general, there is always a trade-off between the size of the heat exchanger(s) and the size of the compressor(s). If the heat exchanger size is increased (increase in capital cost), the power requirements in the compressors decrease (decrease in operational and capital costs). If the size of the heat exchanger is decreased (lower capital cost), larger compressor(s) is required (increase in operational and capital cost). (Asepulund et al, 2009)
Asepulund et al (2009) developed a gradient-free optimisation-simulation for Aspen HYSYS. Processes were simulated and the tool was applied to an SMR process to find the total refrigerant flow rate, composition and the refrigerant suction and condenser pressures that minimise the energy requirements.
Austbo et al (2013) simulated and optimised an SMR process using adaptive simulated annealing techniques. A constraint handling function utilising process characteristics was proposed and compared with static penalty function formulations. The results showed that the best solution found exceeded the previously published results.
Chang, Lim & Choe (2012) investigated the effect of multi-stream heat exchangers on the performance of natural gas SMR liquefaction cycles. The liquefaction performance was estimated with a method of minimum temperature approach (using Aspen HYSYS) and found that the Figure of Merit (FOM: the ratio of minimum work to actual work) value from Aspen HYSYS is an over-estimate, but can closely be achieved in practice with a proper heat exchanger design.
Nogal et al (2008) developed a new design methodology for mixed refrigerant systems with multi-stage heat exchangers to optimise key decision variables. The method has successfully been applied to an SMR process to minimise shaft power demand for compression. The results showed 8% improvement in power demand compared with previously published work.
Nogal et al (2008) also described and illustrated the application of a new optimisation strategy for the synthesis of mixed refrigerant cycles. It considers single- and multi-stage refrigerant compression, full enforcement of the minimum temperature difference along with temperature profiles, simultaneous optimisation of variables, incorporation of capital costs in the objective function and the use of stochastic optimisation (genetic algorithm) to overcome local optima. The results illustrated improved and feasible solutions.
Jacobsen & Skogestad (2013) studied the optimal operation of an SMR LNG process considering disturbances in natural gas flow rate and ambient conditions. The compression power was minimised compared with previous work using UniSim for process modelling and fmincon in MATLAB for optimisation. The authors found that the compressor characteristics set limits on the variations in ambient conditions handled by the process.
Khan & Lee (2013) investigated an SMR process with four compression stages. The power consumption was minimised and optimised employing the particle swarm paradigm using MATLAB. UniSim was used for process modelling and simulation. The authors found that better results were obtained than when using a deterministic approach for the same process.
Khan, Lee & Lee (2011) minimised the total compression power for an SMR process with four compression stages. UniSim was used for process simulation and the sequential quadratic programming (SQP) method fmincon in MATLAB was used for optimisation. A stochastic search method was applied to the same problem represented by Khan & Lee (2013).
Khan et al (2013) developed a knowledge-based decision-making algorithm based on thermodynamic knowledge and insight to improve the energy efficiency of mixed refrigerant LNG process concepts. The methodology proposes a simple and practical method for selecting the appropriate refrigerant composition. The method was applied to an SMR process and proved optimised results.
Marmolejo-Correa & Gundersen (2012) discuss different exergy efficiency definitions and applied this to an SMR process. The authors found significantly different results for various proposed definitions and stipulate the importance of decomposing the exergy into its components (temperature-based and pressure based). This is important to develop exergy efficiency formulations that properly evaluate the thermodynamic performance of different process designs.
An SMR process using standard refrigeration components was proposed by Neksa et al (2010) for re-liquefaction of boil-off gas from LNG tankers. Exergy analysis was performed and process irreversibilities minimised by manually improving the refrigerant composition. PRO/II was used for process modelling and experimentally verified in laboratory scale and full scale. Also stated is that process optimisation has not been performed and indicates room for improved results.
A genetic algorithm was utilised by Mokarizadeh & Mowla (2010) to minimise the compression power in an SMR process with two-stage compression. MATLAB was used for simulation and optimisation, while Aspen HYSYS was used to validate the results. An exergy analysis was also conducted to calculate the irreversibility of the individual components.
Xu et al (2012) investigated the influence of ambient temperature on the optimal refrigerant composition for an SMR process assuming constant pressure levels. Aspen Plus was used for process simulations using a genetic algorithm to minimise the total power consumption. When not adjusting the composition to account for changes in the ambient temperature, it was found that a limited temperature range can be covered without violating the design constraints. Using linear regression of the optimised results, a model for the optimal refrigerant composition was proposed.
He & Ju (2016) developed a dynamic model applicable to a multi-stage refrigeration cycle and investigated the dynamic behaviours of the process for disturbances in feed natural gas temperature, composition, pressure, flow rate and water cooler temperatures. LNG temperature and compressor power requirements were key criteria investigated. Results indicated that compressor power requirements varied with disturbances, where the LNG temperature returned quickly back to set point.
2.1.2 Micro-Scale Natural Gas Liquefaction Processes: Nitrogen Expansion Cycles
He & Ju (2011) proposed two different precooling cycles including propane and R410a to a simple NEC natural gas liquefaction process to improve process performances. These processes are illustrated in Figure 4, Figure 5 and Figure 6.
Figure 4: Nitrogen expansion cycle (NEC) liquefaction process (He & Ju, 2013)
Figure 5: NEC-Propane precooling liquefaction process (He & Ju, 2013)
Figure 6: NEC-R410a precooling liquefaction process (He & Ju, 2013)
Unit energy consumption was optimised in terms of several operating parameters. An exergy analysis was also performed and found that by introducing R410a and propane pre-cooling reduces the base case NEC process by 22.74% and 20.02%, respectively.
He & Ju (2014) introduced a novel conceptual design of parallel nitrogen expansion liquefaction process for the small-scale plant in skid-mounted packages. A thermodynamic analysis was conducted from where the process was optimised using a genetic algorithm.
Although there are some publications on design and optimisation of NEC liquefaction processes, only a few exist regarding micro-scale processes due to the high energy consumption of the compressors. The Claude cycle refrigerator is commonly used in boil-off gas re-liquefaction applications. This cycle was not investigated as it comprises a three-stage compression unit, turbo expander and three heat exchanger units resulting in a complex and expensive process. (Hoseyn, & Babaelahi, 2010; Moon et al, 2007)
2.1.3 Micro-Scale Natural Gas Liquefaction Processes: Process Comparisons
Cao et al (2005) compare two processes, an SMR cycle and a methane-nitrogen expander cycle, by optimising the specific power requirements for small-scale LNG processes. Aspen HYSYS with the built-in original optimiser was used for simulation and optimisation. The expander process was found to be most efficient.
Yin et al (2008) optimised and compared two LNG processes, a multi-stage mixed refrigerant and reverse Brayton process, with a focus on minimising the compression power under given conditions. The capital costs associated with the processes were also studied. The authors found that the capital and operating costs were both smallest for the mixed refrigerant process.
A genetic algorithm was applied by Yoon et al (2012) for the optimisation of an SMR process and a dual nitrogen expander cycle for natural gas liquefaction. Both processes use multi-stage compression. Aspen HYSYS was used for process simulation and MATLAB for process optimisation.
He & Ju (2015) four configuration strategies of gas expansion liquefaction cycles (multi-stage expanders, single precooling cycle, regeneration and mixture working fluid) for the distributed scale LNG plant. Sixteen feasible cycles were found using FOM and optimised with a genetic algorithm. A cost analysis was applied to the optimal cases and found that a two-cycle, R410a precooling cycle and parallel nitrogen expansion cycle, was most feasible.
2.1.4 Literature Findings
Previous research work discussed were analysed and key process parameters contributing to this study are compared (listed in Table 8). The focus was placed on searching for resources with processes producing less than 500 tonnes per day LNG (open literature for these micro-scale LNG processes are very limited).
Found, was a general lower specific power consumption for mixed refrigerant cycles compared with gas expander cycles. Also seen is that smaller-scale processes exhibit a higher specific energy consumption compared with the large-scale processes represented by Asepulund et al (2009) and Austbo et al (2013).
Xu et al (2012) and Yin et al (2008) illustrate good specific power consumption for MRC compared with the large-scale processes. Yin et al (2008) and He & Ju (2015) showed good results for gas expansion cycles. Highlighted processes also have LNG production rates which fall well within the scale of this study.
Table 8: Literature findings key parameter comparisons
Reference | Optimised specific power consumption (kWh x 10-2 /kg LNG) |
LNG flow rate for the optimised case (tonnes per daya) |
Mixed Refrigerant Cycles (MRC) | ||
Asepulund et al (2009) | 25.39 | 8640 |
Austbo et al (2013) | 29.53 | 8640 |
Khan & Lee (2013) | 38.07 | 0.024 |
Khan et al (2011) | 42.44 | 86.4 |
Khan et al (2013) | 43.24 | 0.024 |
Neksa et al. (2010) | 49.00 | 20.1 |
Mokarizadeh & Mowla (2010) | 30.33 | 75.24 |
Xu et al (2012) | 28.15 | 86.4 |
Cao et al (2005) | – | 1.824 |
Yoon et al (2012) | 34.82 | 456 |
Yin et al (2008) | 29.75 | 17.1 |
He & Ju (2016) | 38.10 | 42.39 |
Gas Expansion Cycles | ||
He & Ju (2012) | ||
NEC Cycle | 60.70 | 0.456 |
NEC-R410a | 46.89 | 0.456 |
NEC-Propane | 48.54 | 0.456 |
Cao et al (2005) | – | 1.824 |
Yoon et al (2012) | – | 456 |
Yin et al (2008) | 44.91 | 17.1 |
He & Ju (2014) | 67.05 | 34.88 |
He & Ju (2015) | 48.83 | 38.36 |
a Daily operation assumed at 24 hours per day (hourly rates given in most literature cases)
2.2 Micro-Scale Natural Gas Liquefaction Processes: Simulation Process Strategies
2.2.1 Single Mixed Refrigeration Cycle
The basic process for the design will be taken from the work of Xu et al (2012) due to the lowest shaft power consumption found for an LNG production rate below 500 tonnes per day. The two-stage compressor and -the cooling process is illustrated below.
Figure 7: Two-stage compressor SMR LNG process (Xu et al, 2013)
After the first stage of compression and cooling, the mixed refrigerant stream is partly condensed and separated in a phase separation drum. Upon separation, the gas is compressed and the liquid pumped, where after the streams are mixed and passed through the second water cooler before entering the cryogenic heat exchanger. (Xu et al, 2013)
In the cryogenic heat exchanger (or cold box), the high pressure mixed refrigerant and natural gas are hot streams, while the throttled mixed refrigerant derived from the high pressure is the chill stream and supplies the cold energy for liquefaction (Xu et al, 2013).
Process constraints found for safe and stable operations are listed in Table 9. Initial design process parameter specifications (taken from Xu et al (2013)’s optimised results) to be applied are shown in Table 10. Other process constraints will be specified as lower and upper bounds during optimisation.
Table 9: Process constraints for the SMR process (Xu et al, 2013)
Initial process simulations will be conducted using process simulation software package, Aspen HYSYS. The objective of this exercise will be to establish what affect key decision variables (flow rate and composition of mixed refrigerants, partition temperature and operating pressures before and after compressors) have on other process parameters for given constraints. This is envisioned to give insight for the selection of process parameter search bounds and a better understanding of process operation before applying optimisation tools.
Table 10: Initial process parameter inputs (taken from Xu et al (2013))
Parameter | Unit | Value | |
Feed flow rate | tonnes/day | 86.4 | |
Feed pressure | bar | 50 | |
Feed temperature | ºC | ||
Feed composition | mol% | Nitrogen Methane Ethane Propane Isopentane |
14.59 20.21 32.94 18.00 14.25 |
Compressor outlet pressure | bar | 40 | |
Compressor inlet pressure | bar | 3 | |
Hot MR temperature at HEX outlet | ºC | -160 | |
NG temperature at HEX outlet | ºC | -160 | |
Adiabatic compressor efficiency | % | 78 | |
Superheat temperature at the inlet of first stage compressor | ºC |
greater or equal than 3 |
|
Minimum temperature approach of HEX | ºC | 3 | |
NG pressure drop in HEX | bar | 0.5 | |
Hot MR pressure drop in HEX | bar | 0.5 | |
Cold MR pressure drop in HEX | bar | 0.5 | |
Heat leakage of HEX | % | 0 |
2.2.2 Nitrogen Expansion Cycle
As the basis of process design for this study, the optimised case represented by He & Ju (2015) was selected due to its low specific power requirement and lower number equipment units compared with Yin et al (2008). The R410a-precooling with parallel nitrogen expansion cycle process is illustrated in Figure 8.
Figure 8: Process flow diagram of R410a precooling-nitrogen expansion liquefaction cycle (He & Ju, 2015)
Natural gas flows through two heat exchangers; the first HEX cools the gas down from ambient conditions to -44ºC, where after the second cryogenic HEX produces LNG at -160ºC. Its pressure is reduced to 2 bar upon exit of the throttling valve. Optimal operating conditions found (therefore initial process conditions for this study) is listed in Table 11.
There are two refrigeration cycles. The first, utilising R410a, passes through a two-stage compression cycle with water inter-cooling. Then the refrigerant passes through an expansion valve to reduce the temperature and pressure. The cold
Table 11: Initial process parameter inputs (He & Ju, 2015)
Parameter | Unit | Value | |
Feed flow rate | tonnes/day | ||
Feed pressure | bar | 50 | |
Feed temperature | ºC | 20 | |
Feed composition | mol% | Methane Ethane Propane Iso-butane Na-butane Iso-pentane N-pentane C6-C9b Nitrogen |
92.9 3.54 1.49 0.28 0.34 0.06 0.05 0.03 2.1 |
Pressure drops in HEXs | bar | 0.1 | |
Pressure drop in water coolers | bar | 0.1 | |
Outlet temperature after water coolers | ºC | 40 | |
Adiabatic efficiency of compressors | % | 80 | |
Adiabatic efficiency of expanders | % | 85 | |
LNG storage pressure and temperature | bar / ºC | 2 / -158.2 | |
Minimum temperature approach in HEXs | ºC | 3 |
a Normal; b Hydrocarbons with 6 – 9 carbon atoms
R410a goes through a precooler to provide cooling capacity for the feed gas and nitrogen streams. (He & Ju, 2015)
The second (nitrogen) cycle consists of two compressors (C-1 & C-2) and two water coolers (WC-1 & WC-2), which is then divided into two parts. Each part is booster-compressed (B-1 & B-2), followed by cooling water (WC-5 & WC-6)
to reduce the high-pressure stream temperatures. These streams exiting WC-5 and WC-6 are mixed and cooled in HEX-1 to -44ºC. After that, the nitrogen is divided into two parts again. The major part (207) of the stream enters the first stage expander to reduce its pressure and temperature. The minor part (203) passes through HEX-2 and exits at -110ºC. Then, this low-temperature nitrogen stream enters the expander (E-2) and provides cold energy for HEX-2. Stream 206 from HEX-2 mixes with stream 208. The nitrogen provides cold energy for HEX-2 and HEX-1 subsequently. Compressors B-1 and B-2 are driven by expanders E-1 and E-2, respectively. (Hu & Ju, 2015)
R410a as a refrigerant was proposed by He & Ju (2015) and selected for two reasons: R410a is a common refrigerant, widely available and the dew point of R410a is lower than that of propane (also common refrigerant). This means an R410a-precooling cycle can provide more cold energy for expansion in high-temperature ranges than propane (He & Ju, 2015).
The process has several constraints that ensure safe and stable operation. Key constraints to be applied are summarised in Table 12. Initial process simulations will also be conducted using Aspen HYSYS to investigate the affect key process decision variables have on other process parameters for given constraints. This will give a better understanding of what parameters need to be fixed and which need to be varied before formulating the optimisation framework.
2.3 Micro-Scale Natural Gas Liquefaction Processes: Optimisation Strategies
Various optimisation methods have been applied for the design and operation of LNG plants. In most published work, the optimisation of LNG processes is either solved using local search methods for a rigorous process model or using an advanced global optimisation algorithm combined with a simplified process model. (Austbo, Lovseth & Gundersen, 2014)
Table 12: Process constraints for the R140a-NEC (He & Ju, 2015)
Figure 9 is a representation of optimisation problem formulation methods and the different algorithms used. Among non-deterministic search methods, also known as stochastic search methods, genetic algorithms are most commonly applied for LNG processes. Hybrid approaches have also been developed where a combination of deterministic and non-deterministic search methods is used.
Figure 9: Breakdown of optimisation problem formulation methods and algorithms
Alternative approach analysis methods (exergy and pinch analysis) based on thermodynamic principals have also been applied to improve designs or to reduce the search space for optimisation approaches (Austbo et al, 2014).
The optimisation will be applied to proposed SMR and R410a-NEC processes, first conducting a Genetic Algorithm (GA) search, followed with a pinch- and exergy analysis. Heuristics developed for the selection of MRC composition will also be applied.
2.3.1 Genetic Algorithm Search
GA is a global optimisation method. It utilises a search heuristic that mimics the process of natural selection. This heuristic is routinely used to generate useful solutions to optimisation problems. The proposed liquefaction processes to be optimised are highly non-linear programs with many local optima. (He & Ju, 2015)
Figure 10 illustrates the optimisation process framework with applying GA search. The method starts to work by producing an initialisation population. There are several constraints during optimisation that ensures the process to operate safely and stably. GA decides whether the constraints are satisfied or not. If the constraints are satisfied, GA calculates the objective function, else the penalty function is applied. When the objective function reaches the maximum or minimum values, the best population is found. Tuning parameters of the GA, given by He & Ju (2015), to be used are listed in Table 13.
Key variables that need to be optimised are compressor in- and outlet pressures, outlet pressures of expanders/throttling valves, refrigerant flow rates and compositions, split ratio of mass flow rate in parallel expander (NEC cycle). Lower and upper bounds of variables will be developed upon running first process simulations for given initial process parameters. After understanding the process operation and affect change in manipulated variables have on process outcomes, search bound criteria will be established.
The objective functions to be optimised are minimum power/utility consumption and maximum exergy efficiency for a given range of LNG production capacities (13.5 – 275 tonnes/day). Objective functions will be developed during the optimisation. Simulation criteria will be similar for both SMR and R410a-NEC processes to fairly compare results.
Figure 10: Process optimisation framework with GA (He & Ju, 2015)
2.3.2 Exergy and Pinch Analysis
An exergy- and pinch analysis will also be applied to proposed processes to establish lowest power consumption cycles. Theoretic background on these principals has been discussed in an earlier separate submission (Literature Study).
Table 13: Tuning parameters of genetic algorithm
Tuning Parameters | Value |
Population size | 100 |
Maximum numbers of generation | 300 |
Probability of crossover | 0.5 |
Probability of mutation | 0.05 |
Numer of crossover points | 2 |
Selection method | Tournament |
Tournament size | 4 |
2.3.3 Heuristic Approach
A knowledge-based decision-making method for the selection of mixed refrigerant systems for LNG processes developed by Khan et al (2013) will be applied as an alternative approach in selecting refrigerant compositions. The effect of variations in refrigerant compositions on the process will be investigated.
2.3.4 Optimisation Software
Aspen HYSYS (modelling package developed by Aspen Technology), one of the most rigorous simulators for chemical plants and oil refineries, will be used to perform optimisation. This software utilises extensive thermodynamic libraries and robustness in property calculations and is most commonly used in the LNG industry for process modelling and optimisation (Austbo et al, 2014).
Aspen HYSYS also has built-in optimisation tools with global convergence characteristics and better robustness to starting point dependence in comparison to other schemes available (Hatcher, Khalilpour & Abbas, 2012).
2.4 Micro-Scale Natural Gas Liquefaction Processes: Financial & Economic Modelling
2.4.1 Capital Cost Estimations
Fourteen scenarios will be considered in determining the capital costs for the designs:
Table 14: Cases to be considered for economic modelling
Case | Production Capacity (TPD) | Production Capacity (mmGJ/aa) | Process |
1 | 13.5 | 0.25 | SMR |
2 | 25 | 0.5 | SMR |
3 | 55 | 1 | SMR |
4 | 110 | 2 | SMR |
5 | 165 | 3 | SMR |
6 | 220 | 4 | SMR |
7 | 275 | 5 | SMR |
8 | 13.5 | 0.25 | NEC |
9 | 25 | 0.5 | NEC |
10 | 55 | 1 | NEC |
11 | 110 | 2 | NEC |
12 | 165 | 3 | NEC |
13 | 220 | 4 | NEC |
14 | 275 | 5 | NEC |
an mmGJ/a – million gigajoules per annum (330 days per annum operability)
LNG plant equipment suppliers will be approached in determining large process unit costs (including utilities). Budget cost estimations with a +/- 30% accuracy will be determined.
2.4.2 Operating Cost Estimations
Operating costs for given scenarios to be determined will include electricity costs, make-up cooling water, make-up nitrogen and annual chemical fills (mixed refrigerant, lubrication oil, etc.). Labour, maintenance, security, land and other operating costs will be excluded as these are assumed similar for both processes.
2.4.3 Economic Modelling
Preliminary financial modelling and calculation of project returns will be performed based on assumptions listed in Table 15. South African rates will be taken for purposes of this study. NPVs and ACCs will be key criteria in evaluating the most feasible designs between two processes for given scale ranges.
3 Micro-Scale Natural Gas Liquefaction Processes: Basic Engineering Design
Upon completion of financial and economic modelling, the most feasible design will be selected for the 55 TPD scenario. A generic basic engineering package design will follow which will include the following items:
- Basis of design
- Process description and process flow diagrams
- Process control philosophy
- Piping and instrumentation diagrams
- Utility flow diagrams
- Equipment list
- Catalyst and chemical consumption list
- Utility consumption list
- Effluent and emissions list
- Tie-in point list
- Piping line list
Table 15: Economic modelling assumptions
Identifier | Quantity |
Exchange rates | TBD on modelling execution dates |
Inflation (“CPI”) | 6.1% |
Debt repayment period | 5 to 10 years (depending on CAPEXa) |
Plant useful life | 15 years |
CAPEX term (for manufacturing) | 18 months |
Depreciation (100/15) | 6.67% |
Wear and tear percentage – first year | 40% |
Wear and tear percentage – following three years | 20% |
OPEXb | TBD during operating cost estimates |
Annual gas cost escalation | 6.1% |
Annual LNG selling price escalation | 5.6% |
Debt funding ratio | 50% |
Equity funding ratio | 50% |
Cost of debt (pre-tax) | 13.5% |
Cost of debt (post-tax) | 9.72% |
Cost of equity | 24.37% |
Project risk | TBD during design (vary with plant size) |
The weighted average cost of capital | 17.05% |
Tax rate | 28% |
Gas cost | TBD on modelling date (market-related) |
LNG selling price | TBD on modelling date (market-related) |
a CAPEX – capital expenditure; b OPEX – operating expenditure
- Valve list
- Process data sheets
- Instrument list
- Load schedule
- Single line diagram
- Fire protection philosophy
- Overall emergency shutdown philosophy
- Hazard and operability study
- Equipment layout drawings
- Operation and maintenance plan
Above listed can be incorporated into the overall plant design (gas pretreatment and storage facilities) applicable to a specific project in industry. Gas pretreatment designs vary according to gas source compositions (wellhead, pipeline, microbial, biomass, flare gas, etc.). Excluded from this design are overall plant plot layout drawings, civil- and structural drawings, detailed construction schedule, detailed project cost estimates (for the overall plant), health and safety plan and a decommissioning plan which are all project-specific.
4 References
Asepuland, A, Gundersen, T, Myklebust, J, Nowak, MP and Tomasgard, A (2009), “An optimization-simulation model for a simple LNG process”, Computers and Chemical Engineering, 34, 1606-1617.
Austbo, B, Lovseth, SW and Gundersen, T (2014), “Annotated bibliography – Use of optimization in LNG process design and operation”, Computers and Chemical Engineering, 71, 391-414.
Austbo, B, Wahl, PE & Gundersen, T (2013), “Constraint handling in stochastic optimization algorithms for natural gas liquefaction processes”, Proceedings of the 23rd European Symposium on Computer-Aided Process Engineering – ESCAPE 23, June 9-12, Lappeenranta, Finland.
Cao, W, Lu, X, Lin, W and Gu, A (2005), “Parameter comparison of two small-scale natural gas liquefaction processes in skid-mounted packages”, Applied Thermal Engineering, 26, 898-904.
Chang, Lim & Choe (2012), “Effect of the multi-stream heat exchanger on the performance of natural gas liquefaction with mixed refrigerant”, Cryogenics, 52, 642-647.
Fan, CT and Oelschlegel, H (1986), “Process and apparatus for the liquefaction of natural gas stream utilizing a single mixed refrigerant”, US Patent 4,901,533, assigned to Linde Aktiengesellschaft, Wiesbaden, Fed. Rep. of Germany.
Hatcher, P, Khalilpour, R and Abbas, A (2012), “Optimisation of LNG mixed-refrigerant processes considering operation and design objectives”, Computers and Chemical Engineering, 41, 123-133.
He, T and Ju, Y (2016), “Dynamic simulation of a mixed refrigerant process for small-scale LNG plant in skid mount packages”, Energy, 97, 350-358.
He, T and Ju, Y (2015), “Optimal synthesis of expansion liquefaction cycle for distributed-scale LNG (liquefied natural gas) plant”, Energy, 88, 268-280.
He, T and Ju, Y (2014), “A novel conceptual design of parallel nitrogen expansion liquefaction process for small-scale LNG (liquefied natural gas) plant in skid-mount packages”, Energy, 75, 349-359.
He, TB and Ju, YL (2013), “Performance improvement of nitrogen expansion liquefaction process for small-scale LNG plant”, Cryogenics, 61, 111-119.
Jacobsen, MG and Skogestad, S (2013), “Active constraint regions for a natural gas liquefaction process”, Journal of Natural Gas and Engineering, 10, 8-13.
Hoseyn, S and Babaelahi, M (2010), “Exergetic Optimization of a Refrigeration cycle for Re-Liquefaction of LNG Boil-Off Gas”, Int. J. of Thermodynamics, 13 (4), 127-133.
Khan, MS and Lee, M (2013), “Design optimization of a single mixed refrigerant natural gas liquefaction process suing the particle swarm paradigm with nonlinear constraints”, Energy, 49, 146-155.
Khan, MS, Lee, S, Rangaiah, GP and Lee, M (2013), “Knowledge-based decision-making method for the selection of mixed refrigerant systems for energy-efficient LNG processes”, Applied Energy, 111, 1018-1031.
Khan, MS, Lee, S & Lee M (2011), “Optimization of single mixed refrigerant natural gas liquefaction plant with nonlinear programming”, Asia-Pacific Journal of Chemical Engineering, 7, S62-S70.
Marmolejo-Correa, D and Gundersen, T (2012), “A comparison of exergy efficiency definitions with focus on low-temperature processes”, Energy, 44, 477-489.
Mokarizadeh Haghighi Shirazi, M and Mowla, D (2010), “Energy optimization for liquefaction process of natural gas in peak shaving plant”, Energy, 35, 2878-2885.
Moon, JW, Lee, YP, Jin, Y, Hong, S and Chang, HM (2007), “Cryogenic Refrigeration Cycle for Re-Liquefaction of LNG Boil-Off Gas”, Cryocoolers, 14, 629-635.
Neksa, P, Brendeng, E, Drescher, M and Norberg, B (2010), “Development and analysis of a natural gas reliquefaction plant for small gas carriers”, Journal of Natural Gas Science and Engineering, 2, 143-149.
Nogal, DF, Kim, J, Perry, S and Smith, R (2008), “Synthesis of Cryogenic Energy Systems”, 18th European Symposium on Computer-Aided Process Engineering – ESCAPE 18.
Nogal, DF, Kim, J, Perry, S and Smith, R (2008), “Optimal Design of Mixed Refrigerant Cycles”, Ind. Eng. Chem. Res., 47, 8724-9740.
Xu, X, Liu, J, Jiang, C and Cao, L (2012), “The correlation between mixed refrigerant composition and ambient conditions in the PRICO LNG process”, Applied Energy, 102, 1127-1136.
Yin, QS, Li, HY, Fan, QH and Jia, LX (2008), “Economic analysis of mixed refrigerant cycle and nitrogen expander cycle in small scale natural gas liquefier”, AIP Conference Proceedings, 985, 1159.
Yoon, S, Cho, H, Lim, D and Kim, J (2012), “Process Design and Optimization of Natural Gas Liquefaction Processes”, Chemical Engineering Transactions, Vol 29, ISBN 978-88-95608-20-4, ISSN 1974-9791.