Over the last thirty years, liquefied natural gas (LNG) refrigeration cycles and their respective gas turbine drives have evolved to meet the needs of plants. Despite the technology’s capability to support larger LNG trains and higher refrigeration capacity, the optimum plant size is currently regarded as 5.0 to 6.0 mmtpa per train.

Turbomachinery trains continue to evolve through validation of larger compressor drivers and more versatile and efficient cycles with greater turndown capabilities. In modern LNG plants, the refrigeration compressor designs confront significantly higher flow coefficients and gas densities compared to other industrial applications, challenging designers to evenly distribute power per stage and control the Mach number limit. As the train size has leveled off and material costs continue to rise, many operators have “standardized” on two types of refrigeration cycles for the typical 5.0 mmtpa compression train: Air Products and Chemicals, Inc. Split Mixed-Refrigerant cycle (APCI Split MR) or the ConocoPhillips Optimized Cascade cycle. These cycles effectively dictate the size and the number of refrigeration compressors and drives.

Standardization, however, will never be truly possible for LNG plants since each facility must address unique site issues, electrical power options (or lack of), modularization strategy and its own project economics. Many other design issues remain including redundancy, driver efficiency, overall plant thermal efficiency, degradation rates, emissions and plant availability. We will review turbomachinery and process cycle selections for modern plants by discussing evaluation criteria including thermal efficiency, reliability, turndown, degradation, emissions and optimal size for cost-effectiveness.

LNG design factors

In recent years, the capital costs for greenfield LNG plants, on a per ton basis, have risen dramatically, calling into question the “economies of scale” argument for larger train sizes above 5.5 mmtpa. Figure 1 shows plant capital costs in dollars on a cost per ton basis versus train size. The data are organized by start-up year into three decade periods in order to understand cost inflation. The trend shows a decreasing cost in the early years of LNG developments for all plants built and online prior to 2000. The linear trend suggests that cost reductions from economies of scale were possible, where a 1.0 mmtpa train averaged $500 per ton per annum (tpa) compared to a 3.8 mmtpa train, which averaged about $320 per tpa. After 2000, the LNG plant installed costs are more scattered although it could be argued that costs leveled off at around $450 per tpa for train sizes between 3.5 and 5.0 mmpta. A flat trend is also in line with the Qatargas expansion at $400 per tpa.

Unfortunately, the current decade indicates costs significantly above previous years. The cost increase is primarily due to more difficult green-field site development, and increased material and labor costs. Many of these sites required a modular construction plan which incurred additional costs in modularizing the plant. Cost escalation is not due to the compression train itself. However, this new capital cost environment alters project structure and plant design considerations.

Greater capital costs mean that the net present value (NPV) evaluation and its sensitivity to cash flow variations may be affected more dramatically by operational aspects of the plant. For example, availability in the plant’s initial operating years significantly affects NPV since the project must quickly transition from negative to positive cash flow. Arguably, availability and operational flexibility can be more important to modern LNG project economics than thermal efficiency. To this end, it becomes clear that efficiency can be over-emphasized when competing designs are within several efficiency points of one another and the availability between these designs differs significantly.

Another consideration is the cost of the feed gas and the reserve size. If plants are fed by relatively inexpensive gas (~$1/MMBTU) and reserves are considerable, the compressor train availability and degradation rate will affect project NPV more than relatively small thermal efficiency gains of around 3% to 5%. Electrical power rates also enter into the evaluation when comparing electric motor drives to gas turbines. Since the true efficiency of the electric motor is difficult to appraise given the electrical power line losses and initial generation sources, the real comparison should be an economic one: the cost of electric power vs. the cost of natural gas power.

Figure 2 shows LNG plants by initial start-up year from left to right against the capital cost in 2012 dollars. On the left is the oldest operational LNG export plant built in 1970 at Marsa El Brega. The color coding represents the type of turbomachinery selected for the refrigeration cycle. The chart highlights the success of past innovations in LNG processes and advancement of the process and equipment selections. Throughout the years, several projects have accomplished gains in capacity, lowered specific energy consumption and improved the balance of train power. Many of these projects validated new refrigeration train drivers and refinements on the process cycle while at the same time achieving successful execution schedules and capital costs in line with more traditional designs of the time.

For example, the 1989 Northwest Shelf LNG project used multiple Frame 5 gas turbines instead of steam turbine drives within the APCI propane pre-cooled mixed refrigerant (C3MR) cycle. The capex cost for the plant was only ~$100 per ton more than comparable steam projects. This project provided the basis for future gas turbine driven LNG plants with a smaller footprint.

Atlantic LNG began operation in 1999 and later expanded to accommodate four trains by 2005. The significantly lower capex cost (<$300 per ton) is likely a result of the duplicate trains and a somewhat replicated design. Proximity to the Gulf coast (for reduced shipping costs of modular equipment) and access to a trained labor force also helped this project reduce capital and gain economic advantages. The fourth train used the CoP Optimized Cascade cycle, similar to the three previous trains, but was designed for higher capacity (5.2 mmtpa instead of 3.3 mmtpa) with additional refrigeration compressors. Aeroderivatives were initially considered for the fourth train but operators chose three Frame 5D gas turbines to match the other trains. Efficiency gains were made by re-wheeling the compressors and adding side stream cooling to the methane compressor.

The ExxonMobil RasGas III project was designed with installed costs below the historical average ($300-$400 per ton). The earlier RasGas Trains I and II used the standard APCI C3MR process and Frame 7E gas turbine drives. RasGas III was the first use of the APCI Split Mixed Refrigerant cycle to effectively utilize all available turbine power and incorporate identical Frame 7 gas turbines by shifting load to the propane precooling train. Darwin LNG started operations in 2006 with aeroderivative gas turbines implemented in the CoP Optimized Cascade cycle. The cascade cycle had been used previously at Atlantic LNG and Egypt Idku with industrial turbines. Although all were successful examples of the cascade cycle, Darwin LNG accomplished this new design for only $40 more per ton ($370/ton instead of $330/ton in Egypt).

Carbon factors

With evaporative cooling to boost power output on a per compressor basis and cogeneration via steam turbines (Meher-Homji et al, 2011 Middle East Turbosymposium), this plant maintains high thermal efficiency and a reduced CO2 footprint. This is especially important in Australia given a carbon tax at $23/ton. Statoil’s Snohvit LNG plant was the only major export plant to use electric motor drives in 2008. It incorporated three 65 MW load-commutated inverter drives for start-up and variable speed control in addition to synchronous motors using the Linde mixed fluid cascade process. Although overall capital cost was higher than average projects at the time, it faced other hurdles (small site plot, site access and a stick-built barge which impacted the construction schedule) apart from the driver selection. The design of the variable speed drive system was a major extension of the technology and required extensive testing to ensure torsional natural frequencies did not coincide with integer and non-integer harmonics (Hutten et al, 2011 Middle East Turbosymposium).

Future LNG plants can now consider electric motors with variable speed drives (up to ~70 MW) as viable options which can be appealing in environmentally sensitive areas and in the cascade cycles which require multiple smaller compressors and associated drivers (~32-50 MW). Variable speed drive technology has not yet been proven beyond 70 MW.

Qatargas’s expansion trains were the first design to accommodate the AP-X cycle. The AP-X cycle expanded the APCI C3MR cycle by adding a sub-cooling step with a nitrogen expander to increase train capacity up to 7.8 mmtpa. The project validated the use of Frame 9E gas turbines for LNG applications. Costs were in line with other projects at the time but the project likely benefitted from some co-location benefits with the original Qatargas and the addition of two identical trains in the expansion project.

During the development time period (2004-2008), many other projects were conceptually evaluated to continue “pushing the envelope” on train capacity. However, current LNG projects with larger train capacities have not realized reductions in project costs. Train sizes between 5 and 6 mmtpa are now believed to be the most favorable choice for potential savings from economies of scale.

Figures 3 and 4 highlight changes in LNG turbomachinery selection and process cycles planned in the designs of future plants. The 2016 forecast shows a distinctly different landscape than the plant population in 2003. The widespread use of Frame 7 heavy duty gas turbines and aeroderivatives demonstrates that new projects are staying within the same train capacity at around 5.5 mmtpa and looking for gains in thermal efficiency through cogeneration and more efficient compressor designs rather than new process cycles or validating new drivers.

Since capital cost is driven by many factors, it is difficult to draw firm conclusions based on the process cycle or turbomachinery selection. Viewed as a whole, however, the following turbomachinery trends in Figures 3 and 4 are noteworthy:

Level of Risk Averseness: Although the selections have evolved to accommodate larger train capacity, LNG plants starting up after 2012 are designed using one of two gas turbines: aeroderivative or industrial frame 7E; and one of two primary cycles: CoP Optimized Cascade or APCI Split MR. The use of the proven equipment and refrigeration cycles for new plants is indicative of the risk averseness of new greenfield plant builds due to higher capital costs.

Ambient Temperatures and the Need for Power Augmentation: Many of the plants under construction are located in Southeast Asia or Australia (9 of 11) and will use aeroderivative gas turbines. These locations are characterized by relatively mild ambient temperatures and high labor costs favoring a modular construction approach.

The ambient temperature effects in this area will cause more of a power drop for aeroderivatives compared to industrial turbines (1% vs. 0.7% for every 1 degree C), but the inverse effect is also true: power augmentation will provide a greater power boost compared to heavy duty industrial units: 5%-20% for aeroderivatives compared to a boost of 2%-10% for industrial units.

Growing Emphasis on Driver Efficiency: One of the major reasons to use aeroderivative gas turbines over heavy-duty units is higher efficiency. Aeroderivatives are typically between 38% and 42% efficient compared to heavy-duty gas turbines at 33% to 36%. Many of the new projects have chosen the cascade cycle, which requires multiple lower-horsepower turbines and can use smaller aeroderivatives.

Some of the newer LNG projects will likely incorporate waste heat recovery systems to reduce the specific CO2 footprint. Without a government requirement for high efficiency or a mandated carbon tax, it may be difficult to justify the higher capital cost and potential for lower availability that comes with the combined cycle system.

Train Sizes: Frame 9E gas turbines were used for the Qatargas expansion project but will likely not be needed in future LNG plants as train sizing has leveled off at around 5-5.5 mmtpa. This change was not anticipated five years ago when the majority of technical papers discussed moving beyond 5.0 mmtpa per train.

Project experience has shown that larger capacity trains are more difficult to start-up and do not necessarily offer economies of scale. As such, industrial turbine users will probably continue to select Frame 7 gas turbines for the C3MR cycle. Smaller aeroderivatives will continue to be used in the CoP Optimized Cascade cycle.

This historical review of LNG plants has shown that each project is unique with its own set of upstream costs, layout considerations, operational factors and governing economics. LNG plant design has reached a turning point in technology evolution where cost savings from economies of scale have flattened out and new process cycles can offer reductions in overall plant energy consumption. However, cost savings are less likely to be found in “going bigger” and more likely will come from innovation and smaller, more modularized LNG designs that are suited to project needs and location.

In Part 2 of this series, we examine two primary design factors for modern LNG plants, specific energy consumption and availability. In assessing turbomachinery and process cycle selection for an LNG project, the primary design criteria often compete against one another. We will analyze the impact of specific energy consumption and availability on a project’s net present value.


Marybeth Nored is a Project Engineer at Apache Corporation in the Worldwide Projects group. Marybeth.nored@apachecorp.com

Dr. Klaus Brun is the Director of the Machinery Program at Southwest Research Institute. Klaus.brun@swri.org