DESIGNING FLEXIBLE LNG PLANTS FOR MARKET ADAPTATION

Transcription

1 DESIGNING FLEXIBLE LNG PLANTS FOR MARKET ADAPTATION Jonathan Berg, William Schmidt, Dr. Fei Chen, Christopher Ott Air Products and Chemicals, Inc. In today s LNG market with the addition of new gas sources, it is often desirable to design a flexible LNG plant to account for changing feed gas composition, changing ambient conditions and even changing markets. Designs might include an innovative front-end to remove varying levels of heavy hydrocarbons (HHCs), and/or innovative back-ends to capture helium or remove varying levels of nitrogen. The ever-changing LNG market may require staged investment and/or the need for efficient deep turn-down. The diurnal temperature variations and potential frequent pipeline composition changes make it necessary to have a flexible HHC removal system that can easily adjust to frequent changes in LPGs and heavier components. This paper discusses the synergies of combining adsorption and distillation into a hybrid system, giving a highly flexible, efficient, and small footprint solution. Pipelines and wells can also see the nitrogen concentration change over time. If the backend design is flexible, then nitrogen can be removed and purified by using efficient, novel integration schemes. These schemes can also be used to produce a crude helium product if helium is present in the feed. Recently there has been some interest in staged investment for baseload LNG plants in which a project developer is faced with one or more constraints that preclude investment in large LNG trains. Some projects are proposing sizing some sections of the plant for higher production and the balance for a staged investment with multiple smaller units. This paper explores the optimization of which equipment to upsize and which to break into multiple smaller units. Jonathan Berg, Air Products and Chemicals, Inc.

2 Technical Choice Basics To produce a marketable product, the manufacturer must meet certain product specifications. For an LNG facility, these product specifications can be dictated by end users or be a requirement for safe storage, transportation, and transfer of the LNG. Typical LNG product specifications include: Maximum carbon dioxide (CO2) content Maximum sulfur content Maximum nitrogen (N2) content Range for higher heating value (HHV) Maximum heavy hydrocarbon content (HHC), often expressed as maximum C5+ content Additionally, removing certain components prevents freeze-out in the liquefaction process that could otherwise create significant operational issues or cause damage to equipment. These include: Carbon dioxide Water, either as a pure component or within a hydrate structure HHC s, particularly C6+ and aromatics such as benzene, toluene, and xylenes (BTX) Mercury (Hg) Figure 1 is a simplified block diagram of an LNG liquefaction facility. Following the feed gas flow path, most baseload LNG facilities utilize an amine-based Acid Gas Removal Unit (AGRU) to remove CO 2 and sulfur components to acceptable concentrations. Due to the nature of such an AGRU process, feed gas exiting the AGRU is saturated with water. A temperature swing adsorption (TSA) process reduces the water content of the feed gas below 1 ppm, followed by adsorption beds to remove any mercury present. While there are obviously many technical considerations within this simple blue block in Figure 1, this paper will focus on the remaining separations required, including their integration and synergies with liquefaction to ensure an adequately flexible design than can meet the needs of today s evolving LNG market. Figure 1 Simplified Liquefaction Facility Block Diagram Fuel Endflash Feed AGRU, Driers, Hg Removal NGL &HHC Removal Liquefaction Endflash Drum Storage NGL & HHC One remaining separation is the natural gas liquids (NGL) and HHC components. These condense at warmer temperatures than methane, so their separation either occurs upstream of or is integrated with the liquefaction process. This separation must be performed in almost every facility for at least one of the following reasons: Adjust the LNG HHV Recover valuable NGL byproducts Prevent freezeout/precipitation of HHC s during liquefaction or in the LNG Produce the liquefaction process refrigerants

3 In some cases, the expected nitrogen content of the feed gas over the life of the plant will exceed the LNG production specification. Because nitrogen condenses at a colder temperature than methane, any separation required will occur after liquefying the bulk natural gas stream. The endflash drum depicted in Figure 1 is a very simple means of rejecting nitrogen from the LNG. If helium is present in the natural gas, it would also be removed post liquefaction, as it condenses at a much colder temperature than methane. Helium, while perhaps only present as a few hundred ppm in the natural gas, is a very valuable byproduct and usually can be recovered from the LNG liquefaction process with minimal additional capital investment and no measurable process efficiency penalty [1]. NGL and HHC Removal Options and Considerations There are numerous technologies and configurations available to separate NGL and HHC components from natural gas. Typical options considered include: NGL extraction process upstream of the liquefaction unit Scrub column integrated with the liquefaction unit Partial condensation integrated with the liquefaction unit (can be combined with a stripping column) TSA adsorption process upstream of the liquefaction unit Combined partial condensation and TSA adsorption process integrated with the liquefaction unit Careful consideration should be given to the specific project objectives and constraints to choose the best solution. The technology map presented in Figure 2 can help guide which process(es) are appropriate for the expected feed gas composition, NGL recovery objectives, and HHC removal requirements [2]. For a flexible plant design, variability of the feed gas composition, both on a short-term and long-term basis, must be evaluated. Figure 3 illustrates how relatively small changes in composition can have a dramatic effect on the resulting phase envelopes. Because the operating point for any vapor-liquid separation process (e.g. NGL extraction, scrub column, partial condensation) must be within the two-phase region of the envelope, such composition changes may require an equally dramatic adjustment to the operating point. It is important that the process supplying refrigeration for any rectifying section condenser be able to provide that refrigeration at the required temperature. As an example, using a pure component propane refrigerant circuit to cool a scrub column condenser cannot cool the column overhead below approximately -35 C before the propane compressor suction must operate below atmospheric pressure the boiling point of propane at 1 atm is -42 C. This level of refrigeration may be acceptable for a rich feed gas but can become problematic for a lean feed gas, or one that becomes leaner over time. Figure 2 NGL Recovery and HHC Removal Technology Map Rich Integrated Scrub Column more economical Feed Gas Partial Condensation Partial Condensation + TSA NGL Expander Process less economical Lean TSA Only Low NGL Recovery + HHC Removal High

4 Methane Concentration (mole%) C6+ Concentration (ppm) Figure 3 Phase Envelopes of Various Feed Gas Compositions For processes integrated with the liquefaction unit (i.e. integrated scrub column), any pressure or temperature adjustments required to achieve the proper separation will disturb the liquefaction process. Due to the large heat transfer duties associated with liquefaction, the system also has a significant amount of thermal inertia. If rapid and/or unpredictable composition changes are anticipated, such as the pipeline feed concentrations illustrated in Figure 4 and Figure 5, it may be impractical to quickly adjust the operating point. In this case, either a robust and flexible process is necessary to accommodate a wide range of feed gas compositions, or alternatively, the initial operating point must be selected to work across the range of expected feed gas compositions. Inherent to the latter solution is that for most feed gas compositions the liquefier will not operate at its peak efficiency. For example, increasing the feed pressure generally improves liquefaction efficiency, as can be seen in Figure 6 [3]. Operating an integrated scrub column at 45 bara to accommodate all compositions in Figure 3 means there is unrealized LNG production (or excess power consumed) during the periods with richer feed gas compositions. An upstream, stand-alone NGL recovery unit allows liquefaction to occur at any operating pressure, while a combined partial condensation and TSA unit also offers flexibility to handle wide feed gas variations [2]. An economic analysis should be performed to evaluate if the additional LNG production obtained justifies installation of such technologies. 100 Methane Figure 4 Pipeline #1 Feed Composition C Time (days) Methane C6+ (ppm)

5 Relative Concentrations Figure 5 Pipeline #2 Feed Composition 2.0 Relative Feed Concentrations N2 C2 C4 C5 C Day Figure 6 Effect of Feed Pressure on LNG Production and Overall Liquefaction Specific Power Nitrogen Removal Considerations Including Helium Recovery Due to their low normal boiling points (NBP), both helium and nitrogen condense at colder temperatures than methane. It is, therefore, logical to use vapor-liquid-equilibrium (VLE) to separate them from LNG either with or after the natural gas liquefaction process. Helium NBP = C Nitrogen NBP = C Methane NBP = C

6 Nitrogen separation technologies typically considered for LNG facilities include [4]: LNG flash drum (note, this could include an endflash drum and/or flash directly in the LNG storage tank) Nitrogen stripping column LNG flash drum with nitrogen rejection unit (NRU) Nitrogen stripping column with NRU Nitrogen rectifier column Because the vapor stream from a single stage of separation (i.e. LNG flash drum) or even a nitrogen stripping column will still contain a large fraction of methane, this stream often satisfies the facility fuel demand. While somewhat counter-intuitive, sourcing fuel gas from end flash (rather than warm, high pressure feed gas) actually improves the overall liquefaction efficiency [5,6]. Therefore, the technology selected for nitrogen removal is generally a function of the expected range of nitrogen concentrations in conjunction with the facility fuel demand. Figure 7 illustrates appropriate nitrogen removal technologies as a function of these parameters. It is worth noting that the selection of the refrigerant compressor driver(s) is very important, because it sets both the amount and quality of the fuel gas required. Electric motors supplied from a local power grid do not require any fuel gas, whereas gas turbine (direct drive and indirect drive through onsite electricity generation) and steam turbine drivers require a significant amount of fuel gas. The allowable nitrogen concentration in the fuel gas also depends on the driver type. The process efficiency of the liquefaction technology selected then has a compounding effect on the fuel gas demand. Due to the interactions, synergies, and constraints among various technology choices within an LNG facility, it is important to consider both upstream and downstream impacts when choosing technologies. Figure 7 Nitrogen Removal Options While typically present in much smaller quantities than nitrogen, helium is another impurity often found in natural gas, particularly in certain geographic locations. However, due to its low concentration, inert nature, and the specialized analytical techniques necessary to definitively detect its presence, helium is sometimes overlooked. If present in sufficient quantities, helium recovery presents a significant opportunity to improve an LNG opportunity s economics. The economic evaluation usually is quite simple and reduces to the quantity of helium available to recover effectively the facility throughput multiplied by the helium concentration in the feed gas. Generally, MMSCFY is sufficient to justify recovery, although this threshold may be as low as 50 MMSCFY for U.S. facilities [1]. This equates to feed gas concentrations of only a few hundred ppm for large baseload LNG facilities. With its very cold boiling point of C, helium is easily separated from liquefied natural gas using a single stage flash drum (note, this helium recovery flash drum would be in addition to the LNG flash drum previously discussed), and much of the world s helium supply is obtained in this manner. By adjusting the pressure of the helium recovery flash drum (typically in the 2-5 barg range), helium recoveries of >90% are usually achievable. The crude helium stream is upgraded and purified in a separate processing unit for export as a valuable byproduct. Any contained methane in the crude helium stream can then be returned to the LNG facility as fuel gas.

7 Despite the relative ease of separating helium from the bulk natural gas stream and minimal capital investment required to do so (single flash drum and associated valving and piping), upfront planning and coordination is critical. In addition to helium being overlooked in the feed gas, the extremely large scale of LNG facilities can make helium recovery an afterthought during the initial design phase. That same extremely large scale also complicates retrofit efforts, as any downtime required for implementation would have huge financial ramifications. Proper analysis of the expected feed gas composition is essential to determine if any helium recovery infrastructure (or even tie-ins for future recovery) should be installed. Special consideration should be given to facilities using pipeline feed gas, as it may be supplied by many different sources that change over time. Liquefaction Process Flexibility There are a few fundamental decisions that must be made when selecting a liquefaction process. These very basic decisions include: Facility and/or liquefaction train size Liquefaction refrigerant type o Pure component, including N2 and C1 refrigerant processes o Mixed refrigerant (MR) Precooled or single refrigerant processes Figure 8 Air Products Liquefaction Process Offerings [7] As noted in Figure 8, MR processes can offer a significant efficiency benefit compared to vapor expansion cycles, however, MR processes require the extraction or purchase of those refrigerants as well as on-site storage. Likewise, precooled processes offer an efficiency benefit at the expense of the additional equipment associated with a separate precooling refrigeration circuit. By shifting a portion of the refrigeration duty to a separate refrigeration circuit, precooled processes also can accommodate higher production for a given main cryogenic heat exchanger (MCHE) size. Additionally, the improved efficiencies of both MR and precooled processes enable even larger LNG train sizes, as the amount of circulating refrigerant per unit of LNG produced decreases. The overall project objectives and constraints generally guide these basic decisions. Train capacities over 1-2 mtpa typically justify the additional capital investment required for MR and precooled processes, while lower capacities or refrigerant sourcing/storage concerns might push an owner to choose refrigerants containing N2 or C1 only. Once the type of liquefaction process is selected, another major decision is the driver to be used for the refrigeration compressors. Traditional drivers in the LNG industry have been steam turbines, gas turbines (both frame and aero-derivative), and electric motors. The configuration and number of drivers required will then be a

8 function of the available driver power, train size, liquefaction process selected, and the corresponding process efficiency for the given feed gas pressure and ambient conditions [8]. Over the last 30 years, LNG trains have increased in size from under 3 mtpa to nearly 8 mtpa. As the technology improves and advances, it naturally makes sense to exploit all available economies of scale. As a very simple example of fundamental economies of scale, the volume of metal required for a given length of pipe at a given design pressure and constant fluid operating conditions (i.e. density) to maintain the same pressure drop is calculated by the following equations as a function of mass flow. (1) P = (f L ) ρ v2 = (f L ) ρ ( m D i 2 D i 2 ρ A )2 = (f L m 4 ) ρ ( D i 2 ρ π D i 2) 2 = (f L (2) P 1 = P 2 = C 0 ( m 2 1 D i1 5) = C 0 ( m 2 2 D i2 5) (3) D i2 = ( m ) D i1 m 1 (4) D o = D i + 2t (5) required pipe thickness, t = P D o 2 (SE+PY) = P (D i+2t) 2 (SE+PY) = C 1 (D i + 2t) (6) t = ( C 1 1 2C 1 ) D i = C 2 D i D i 5) ( m 2 ρ ) (2 π )2 = C 0 ( m 2 D i 5) (7) metal volume, V = π (D i + t) t = πd i t + πt 2 = πc 2 D i 2 + πc 2 2 D i 2 = π(c 2 + C 2 2 )D i 2 = C 3 D i 2 (8) V 2 = C 2 3D i2 V 1 C 3 D 2 = ( D i2 ) 2 = [( m ) i1 D ] = ( m 2 i1 m 1 ) 0.8 m 1 (9) splitting fixed flow into N parallel pipes, V tot = N V N = N ( m 0.8 N) V m 1 = N 0.2 V 1 where, P Pressure drop, assumed constant P Pipe internal design pressure, assumed constant f Friction factor, assumed constant (developed flow) S Material allowable stress, assumed constant L Pipe length, assumed constant E Material quality factor, assumed constant D i Pipe internal diameter Y Coefficient of wall thickness, assumed constant ρ Fluid density, assumed constant V Pipe metal volume v Fluid velocity N Number of pipes m Mass flow rate of fluid C 0 Intermediate (lumped) constant A Pipe internal cross-sectional area C 1 Intermediate (lumped) constant D o Pipe outside diameter C 2 Intermediate (lumped) constant t Required pipe wall thickness C 3 Intermediate (lumped) constant Table 1 summarizes the relative pipe metal volume for N pipes in parallel using Equation 9. Because the amount of metal required increases at a slower rate than that of the mass flow, this means that 1 pipe will always require less metal than 2 thru N pipes. Table 1: Relative Pipe Metal Volume for N Pipes in Parallel N pipes Relative Metal Volume, % 100% 115% 125% 132% 138% 158% 172% 182% 197%

9 While a fundamental scale-up factor (exponent) of 0.8 has been derived for volume of pipe metal, Figure 9 illustrates the relative total cost for N equipment in parallel considering various scale-up factors. Though a scale-up factor of 0.6 has been traditionally assumed for LNG trains, Kotzot suggests economies of scale start to deteriorate at very large LNG train sizes [9]. Note that a scale-up exponent less than 1.0 represents favorable economies of scale, and a scale-up exponent greater than 1.0 represents diseconomies of scale. Figure 9: Relative Total Cost as a Function of N Parallel Units at Various Scale-up Factors Returning to the pipe example, if the resulting pipe size for a single pipe is easily manufactured, procured, shipped, installed, etc., and the only cost factor is weight of metal, then a single pipe is the most cost-effective solution. However, at some point a very large, thick pipe will become difficult to roll, weld, ship, install, etc. and result in longer pipe lengths as equipment becomes larger and farther apart. Similarly, other equipment can approach practical constraints or diseconomies of scale with larger LNG trains. One example is the required refrigerant compressors aerodynamics, such as tip (or peripheral) Mach number and inlet flow coefficient, which may exceed well-referenced designs as the driver power increases to accommodate larger LNG trains [10]. Parallel driver arrangements can help alleviate such constraints to debottleneck the capacity attainable in a single LNG train. In addition to the technical capabilities of individual components, today s evolving LNG market, with new gas sources being developed by some owners in new types of geographic locations, is presenting new project challenges and constraints that must be considered. Certain business models require staged investment, efficient turndown operation, or condensed project cycle times. Certain sites are plot space constrained or not easily accessible. Certain climates have larger ambient temperature swings than the tropical climates of many previous baseload facilities. Certain regions have a shortage of skilled, affordable labor. These constraints have generated significant interest in smaller, parallel trains, possibly built in modules that are constructed offsite by a better skilled, more established, and/or less expensive labor force. To maximize an LNG project s profitability, ideally these alternatives are applied when necessary to meet project constraints or objectives or when diseconomies of scale exist. Economies of scale would then be exploited wherever possible. There may even be some synergies in using smaller, parallel components for some functions and single, larger components for others. Parallel components or trains certainly have the potential to offer efficient turndown or accommodate staged investments, but only if all components in the train can run stably and efficiently at lower rates. Expected operating ranges for equipment often used in LNG liquefaction trains are as follows [11,12,13]: Compressors The power consumed depends on the specific machinery configuration. With a single compression string, the power decreases with turndown to approximately 60-80% of design capacity. Below this, the compressors will recycle and there will be no further decrease in power. For parallel compression strings, power reductions down to 20-30% of nameplate typically can be obtained. Flash drums includes partial condensation and end flash drums Simple flash drums typically can be operated stably and efficiently anywhere between 0 and 100%

10 Distillation columns includes scrub columns, NGL fractionation, and N2 stripping columns Turndown will be limited by the stable operating limit of the distillation column, which depends on the type of trays or packing selected. Typical turndown limits are 20-80% of nameplate. Temperature swing adsorption A TSA process can turn down to any feed rate from 0 to 100%, by extending the online time. The regeneration time can also be extended to reduce the regeneration stream, which is typically used as fuel. Air / water cooled heat exchangers includes those used for compressor intercoolers and aftercoolers These heat exchangers can be operated stably and efficiently anywhere between 0 and 100%. Operation below 100% should improve efficiency by reducing the approach to ambient temperature. Alternatively, air cooler fan banks can be turned off, if necessary. Kettle type heat exchangers includes those used for precooling in the AP-C3MR TM process These heat exchangers can be operated stably and efficiently anywhere between 0 and 100%. Core type or brazed aluminum heat exchangers (BAHX s) Efficient turndown operation is achievable, however, means to maintain flow stability and equal distribution must be considered to prevent thermal stresses if multiple cores are operated in parallel particularly with condensing and boiling fluids. Coil wound heat exchangers (CWHE s) CWHE s have been proven to operate stably at very low rates, even down to 5% of nameplate, as the CWHE configuration offers inherent flow stability on both the tube and shell sides. Air Products CWHE design also has inherent modular characteristics with multiple tube sheets that can be headered internally. This design ensures equal distribution to minimize thermal stresses. From this list, most equipment has a very broad operating range, except single string refrigerant compression. Therefore, parallel compression strings are likely required for efficient operation at significant turndown, while economies of scale can be exploited within project constraints for other equipment. As noted previously, parallel compression strings can also help debottleneck compressor aerodynamic constraints. Recent advances in compressor driver arrangements have enabled parallel compression strings to be utilized more readily. Probably the most notable example is the arrangement with all compression services on a single shaft and then duplicated in parallel. Figure 10 depicts a 2x50% arrangement that has been implemented in four trains using the AP-C3MR TM process [14,15]. While such arrangements increase the number of compressor casings, an important benefit is seamlessly shifting power between precooling and liquefaction compression services. This flexibility is particularly useful in climates with wide ambient temperature variations that result in large swings in the required precooling duty, as it allows for increased utilization of the overall available power installed [16]. Additional methods to increase power utilization over a broad range of ambient temperatures, such as compressor inlet guide vanes (IGV s), are discussed by Ott et al. [5]. Figure 10: 50% Parallel AP-C3MR LNG Process Parallel compression strings within a single liquefaction train offer a unique, synergistic benefit. Provided common modes of failure are minimized and a robust parallel compressor control scheme is implemented to prevent sympathetic trips among the compression strings, there is a very high probability that one compression string will always be operational. Given a 95% availability for a single compression string, Table 2 lists the probability that at least 1 compression string will be operating as a function of the number of compression strings in parallel. For a process that chills a product to -160 C, this allows equipment that operates at -160 C to continue operating at C, even when most compression trips occur. Keeping equipment cold and operating eliminates costly time (and possibly flaring) required to cool it down or even warm it up and then cool it down in the controlled manner that is otherwise necessary following a trip.

11 Table 2: Probabilities of Operation for N Units in Parallel (Each with 95% Availability) N parallel units (95% availability each) Probability of at least 1 unit operating, % Probability of at least 1 unit down, % Probability of at least 2 units down, % N/A Total trips per year (assuming 6/unit/year) Average trip interval, days Table 2 can be applied to parallel liquefaction trains as well as parallel compression strings. The following learnings from Table 2 should be considered in the economic evaluation of using many smaller units to overcome potential economies of scale of larger units: Probability that 1 string/train is operating nearly 100% of the time is achieved with only a few parallel units. If there are many units operating in parallel, it is very likely at least 1 is not operating meaning the facility is in constant startup mode, including downtime to cool down equipment if utilizing fully parallel liquefaction trains. Such activity usually requires operator attention (both in the control room and in the field), so the required staffing level of the facility is expected to be higher. For many trains in parallel, common failure modes (e.g. facility power interruption) require restarting many trains at the same time. This results in a longer time until all units are back online or requires significant extra staff to restart several units in parallel. Case Study Examples The following two case studies illustrate how the above technical choices may be applied collectively to achieve robust designs for two generic LNG opportunities containing several of the evolving LNG market needs. A third case study will explore the optimal number of trains for a staged investment strategy. Project A Requirements A large baseload facility of 24 mtpa is be constructed. However, investment is to be made in 6 mtpa (720 tonne/hr at 95% availability) increments. The plot space has few restrictions, and a precooled mixed refrigerant process is preferred to achieve the best possible efficiency considering the large plant size. Specific project constraints and a conducive labor market should allow for economies of scale to be exploited wherever possible. Based on site conditions and an inlet feed gas pressure resulting in a liquefaction inlet pressure of 55 bara, the average expected process efficiency (in the form of specific power) is 300 kw-hr/tonne LNG. Ultimately, this results in a required power of 216 MW for each 6 mtpa. The project preference is to realize the reduced auto-consumption and highpressure restart benefits associated with 2 large high-efficiency, multi-shaft gas turbine drivers per train each with a site rated, available gas horsepower of 100 MW after appropriate margins. There are significant diurnal temperature variations. The feed gas is supplied by pipeline with N2 content just above 1%, 300 ppm of helium, and a relatively lean but variable composition. Nonetheless, refrigerant recovery is still desired. Due to the project business model, efficient turndown may be necessary for extended periods of time. Project A Selections Based on these requirements, an upstream NGL removal unit likely provides the best HHC removal solution for this opportunity. It offers the ability to handle feed gas variability while also providing increased NGL recovery, which considering the lean feed gas, will be necessary to recover adequate refrigerant make-up components. In addition, an upstream NGL removal unit has some synergies with the liquefaction driver selection. With 200 MW (or 200,000 kw) of power available and an expected specific power of 300 kw-hr/tonne, only 667 tonne/hr of LNG can be produced without adding helper motors. Note that large starter motors are not required to restart the multi-shaft gas turbine drivers. Utilizing an upstream NGL unit, the feed gas pressure to liquefaction can then be increased to ~90 bara to achieve the target LNG production of 720 tonne/hr without adding helper motors (see Figure 6).

12 Both the AP-C3MR and AP-DMR processes would be viable precooled MR liquefaction process selections with comparable efficiencies. Due to the small footprint of CWHE s [13], utilizing them to perform the precooling duty in the AP-DMR process can save valuable plot space. However, plot space is not restrictive in this example, so the additional space required for the C3 kettles in the AP-C3MR process is acceptable. Therefore, AP-C3MR will be selected for its well-referenced technology. Compressor aerodynamic constraints, particularly on the C3 compressor, necessitate parallel compression strings for 6 mtpa LNG production. By boosting the feed gas pressure after the NGL removal unit to eliminate the need for helper motors, the driver configuration in Figure 10 is more easily utilized as fewer bodies are required on the driver shaft. The Figure 10 configuration is also a good fit to accommodate the significant diurnal temperature variations, as power can be shifted seamlessly between the C3 and MR circuits. Considering current manufacturing capabilities, inherent modular nature in design, and ability to operate at very low turndown rates, it is very feasible to use a single CWHE for a 6 mtpa LNG production using the AP-C3MR process. A single string of C3 kettles has adequate turndown capability, such that parallel kettles are not required. Having two parallel compressor strings, each with 95% availability, and a single MCHE allows for efficient turndown across a broader range [11] and means at least one compression string should be operating 99.75% of the time per Table 2. Full liquefaction train restarts, including time to cool down equipment, should be very infrequent. The low N2 content of the feed gas allows a single flash stage to meet a 1% N2 specification in the LNG. This flash can be done in a separate vessel or directly in the LNG storage tank, and the resulting flash gas can be used to supply fuel to the gas turbine drivers to improve liquefaction efficiency. The helium content of the feed gas is high enough to justify recovery, so a separate flash drum will be added upstream of any end flash. Project B Requirements As part of a staged investment model focused on CAPEX, a mid-size plant of ~1 mtpa is desired, and it is expected economies of scale can be exploited for a single train of this size and specific project constraints. The feed gas composition is variable with a HHC tail but is too lean to source refrigerant make-up components and has a high N2 content. The facility location has access to an economical, robust electrical power grid and possesses the ability to import refrigerant components. While the local skilled labor force is limited, the site is easily accessible. Project B Selections Considering the ability to import refrigerants and focus on CAPEX, the AP-SMR process can be selected to balance process efficiency and CAPEX at this train size. As mentioned previously, a combined partial condensation and TSA unit offers the best flexibility to handle wide feed gas variations and offers a robust, flexible design for this application particularly given that NGL removal is not necessary for refrigerant make-up or to adjust the LNG HHV. The access to an economical, robust electrical power grid makes the use of an electric motor as the refrigerant compressor driver quite appealing. However, using electric motors results in a very low facility fuel demand, and combined with the high N2 content in the feed gas, means a N2 stripping column combined with an NRU will be required for N2 removal per Figure 7. There may be some optimization to determine if utilizing a gas turbine driver with a higher fuel demand could eliminate the need for an NRU and provide an overall project benefit. Lastly, a modular design built by a more established labor force will be utilized to minimize local construction costs/risks. The site accessibility should allow transportation of any large modules. Project C: Evaluate the optimal number of LNG trains for staged investment and ramped demand Beginning 1 year following a final investment decision (FID), constant facility demand growth of 1 mtpa each year for a total facility LNG capacity of 5 mtpa after 5 years (6 years after FID) Capital expenditure (CAPEX) costs of $1,000 / tpa LNG assumed for a 5 mtpa LNG train (MM$5,000 total). For model simplicity, capital outlay occurs as a single charge in investment month 0 for each train. Scale-up costs for multiple parallel trains are based on Figure 9. Installation time for a single 5 mtpa train is assumed to be 48 months. A scale-up factor of 0.4 has been assumed for schedule so that 30 parallel trains can each be constructed and installed in 12 months to exactly match the demand ramp period. Because some common site infrastructure is still necessary, this assumption may be optimistic. The investment interval is then optimized to maximize the net present value (NPV) of investment. Discount rate of 12% and annual inflation rate of 2% are assumed. Revenue is based on $7/MMBtu, 1050 Btu/SCF (@ psia and 60 F), and LNG MW = 17 lb/lbmole. Note, this is likely an optimistic assumption, particularly for a tolling fee, and would favor being able to match the demand curve closely with multiple parallel trains.

13 Table 3: Assumed installation schedule and calculated CAPEX for N parallel trains N parallel trains Assumed installation time per train, months Train investment interval, months Total CAPEX, MM$ CAPEX/train, MM$ (scale-up factor = 0.6) Total CAPEX, MM$ CAPEX/train, MM$ (scale-up factor = 0.8) N/A ,000 5,000 5,000 5,000 6,598 3,299 5,743 2,872 7,759 2,586 6,229 2,076 8,706 2,176 6,598 1,649 9,518 1,904 6,899 1,380 12,559 1,256 7, , , , , , , While the CAPEX values shown in Table 3 represent the cost in month 0, staged investment must consider the time value of money. By delaying some CAPEX investment, remaining money (if available) can be invested in profitable opportunities elsewhere, and the resulting proceeds will help defray the eventual CAPEX required. As an example, the NPV of the CAPEX required for the 2 nd of 3 trains purchased in month 12 (or year 1.0) can be calculated as follows assuming a discount rate of 12% and a scale-up factor of 0.8. NPV = year CAPEX/train (1 + inflation rate)investment MM$2,076 ( )1.0 (1 + discount rate) investment year = ( ) 1.0 = MM$1,891 Note, the MM$1,891 result is less than the MM$2,076 CAPEX/train in Table 3 and high discount rates will tend to favor more trains, as a higher investment return will be realized on any delayed CAPEX. The revenue stream can be calculated on an NPV basis in an identical manor. A summary of the schedule for LNG demand, installed capacity, CAPEX, and revenue streams is given in Table 4 for the 3-train example using a scale-up factor of 0.8. Time (months) Table 4: Schedule and NPV calculations for 3 trains with a scale-up factor of 0.8 LNG Demand Facility Capacity CAPEX, NPV basis Revenue, NPV Basis (mtpa) (mtpa) (MM$) (MM$/month) Total NPV (MM$) , , , , , , , , , , As presented in Table 4, the discounted payback period for the NPV to become positive is 115 months, or 9.6 years. Figure 11 illustrates the discounted payback periods as a function of N parallel trains across a range of scale-up factors to help determine the optimal number of trains for the given project constraints. In this example, installation of more than 1 or 2 trains in a staged investment strategy only provides an economic benefit for scaleup factors approaching (or exceeding) 1 in other words, the total CAPEX required for 30 parallel trains is identical to that of 1 large train. While turndown efficiency must be considered in the operating expenses (OPEX) for larger trains that would need to operate at reduced rates for an extended period of time, the payback period for many trains in parallel can be delayed significantly if economies of scale (i.e. scale-up factors of 0.6 to 0.8) are available

14 and not utilized. Conversely, the benefit of many trains in parallel for scale-up factors of 0.9 to 1.0 is not nearly as dramatic. Figure 11: Calculated discounted payback periods for N parallel trains Conclusions The technology selection for NGL and HHC removal is a function of the feed gas composition and required NGL/HHC removal in combination with the expected feed gas variability. The technology selection for nitrogen removal is a function of the nitrogen content in the feed gas and required facility fuel demand. The facility fuel demand is primarily determined by type of refrigerant compressor drivers selected. The presence of helium should be investigated, as it is relatively simple to recover in an LNG liquefaction process and can significantly improve the project economics. A number of synergies, interactions, and constraints exist among liquefaction, HHC removal, and nitrogen removal technology selections. Parallel refrigerant compressor strings can offer a number of liquefaction process benefits, particularly for large train sizes, as compressor economies of scale deteriorate. Advances in compressor driver arrangements have enabled increased liquefaction unit operating flexibility. Aside from compressors, most other liquefaction equipment provides efficient turndown operation. Based on the case study for Project C, the conditions necessary for many smaller liquefaction trains to provide an overall economic benefit are: o No economies of scale can be achieved (i.e. scale-up factors of 1.0 or above) o Capital is available with a relatively high discount rate o Ramped demand that begins prior to completion of 1 to 2 larger trains but does not reach its final value until sometime after the start-up of 1 to 2 larger trains o The ability of smaller liquefaction trains to be onstream in time to meet demand o Relatively high LNG price to generate significant revenue during ramp up period The evolving LNG market simultaneously presents many opportunities and challenges that demand robust and flexible solutions. Determining the optimal combination of solutions is not always simple, as there are often interactions and synergies among technology choices and specific project constraints. Bigger is not always better. More is not always better. And one size certainly does not fit all. However, close collaboration among the owner, EPC, process licensor, and equipment suppliers will help facilitate the best overall solution for your LNG project.

15 References [1] Nelson, W., Cao, J., Sanderson, C., Ploeger, J., A Fresh Look at Helium Recovery from LNG, LNG 2019 [2] Schmidt, W.P., Chen, F., Ott, C.M., Advances in Removing Heavy Hydrocarbons, Gastech 2018 [3] Schmidt, W.P., Ott, C.M., Liu, Y.N., Kennington, W.A., How the Right Technical Choices Lead to Commercial Success, LNG 16 [4] Ott, C.M., Roberts, M.J., Trautmann, S.R., Krishnamurthy G., State-Of-The-Art Nitrogen Removal for Natural Gas Liquefaction: New Solutions to Meet Market Needs, Gastech 2014 [5] Ott, C., Beard, J., Pearsall, R., Debottlenecking: Getting the Most Out of Your LNG Plant, Gastech 2018 [6] Berg, J., Liu, Y.N., Maximizing LNG Capacity for Liquefaction Processes Utilizing Electric Motors, 2012 AIChE Spring Meeting [7] Bukowski, J.D., Liu, Y.N., Pillarella, M.R., Boccella, S.J., Kennington, W.A., Natural Gas Liquefaction Technology for Floating LNG Facilities, LNG 17 [8] Krishnamurthy, G., Roberts, M.J., Wehrman, J.G., Drive your LNG project to success with optimal machinery selections from Air Products of US, LNG Journal, May 2015 [9] Kotzot, H., LNG Liquefaction Not All Plants Are Created Equal The Sequel, Gastech 2009 [10] Ott, C.M., Schmidt, W., Wehrman, J., Dally, J., Large LNG Trains: Technology Advances to Address Market Challenges, Gastech 2015 [11] Schmidt, W.P., Holstine, M.C., LNG Liquefaction: Uncommon Knowledge Leads to Innovation, LNG 18 [12] Dally, J., Butler, C., Miller, W., Schmidt, W., Styer, J., Coil Wound Heat Exchanger Design for an Evolving Market, Gastech 2018 [13] Dally, J., Schmidt, W., Styer, J., A CWHE design for an evolving market, LNG Industry, November 2018 (page 23) [14] Ott, C., Chasnyk, I., Kaart, S., Bladanet, C., Laflotte, B., Le-Ridant, G., Delva, B., Zucchi, O., Yamal LNG: Meeting the Challenges of Natural Gas Liquefaction in the Arctic, LNG 2019 [15] Bocherel, P., Strohman, J., et al., Converting Dominion Cove point LNG into Bidirectional Facility, LNG 18 [16] Schmidt, W.P., Ott, C.M., Liu, Y.N., Wehrman, J.G., Arctic LNG Plant Design: Taking Advantage of the Cold Climate, LNG 17