Performing engineering studies with dynamic simulation is
becoming a critical requirement for new liquefied natural
gas (LNG) plants. Typically, the major target for such a
study is the investigation and review of critical plant design
elements, such as compressors, flares and boiler systems. Dynamic engineering
studies can identify design changes that will significantly
improve plant performance and the safety and reliability of plant operations.
Furthermore, if such design changes are identified early, they
can be implemented at a low cost and provide significant
savings during a plants lifetime.13
plant design is based on steady-state process simulation. This
approach typically does not take into account rotating
equipment characteristics, holdups or actual pressure drops.
However, the use of a dynamic process simulator is required to
understand actual plant transients and dynamics, to examine and
verify control schemes, and to review plant procedures. The
dynamic model allows for the calculation of process variables
as a function of time (i.e., as a movie instead of a series of
snapshots). Moreover, it is possible to examine process upsets,
including process startups and shutdownsa critical
functionality not offered by the steady-state simulator.
This article presents experiences obtained from a detailed
engineering study for Sonatrachs GNL3Z project, a grassroots baseload LNG
production plant in Algeria. A brief description of the LNG
production process is provided, along with a number of test
cases demonstrating the design challenges faced and their
resolutions. Lessons learned during this study with regard to
project execution and project management are offered, and the
benefits obtained through engineering studies are
Systems examined and simulation test cases
All compressor systems in the GNL3Z plant have been
examined, and a series of simulations has been performed to
verify the process design. In this section, the approach used
in engineering studies and in the overall liquefaction process
is described. Next, details are provided on the compressor
systems and on the test cases reviewed.
Engineering study approach. A dynamic
engineering study involves a number of steps before delivery of
the final results and recommendations, as outlined below:
- Exact scope definition, which includes the modeling
boundaries and the test cases to be examined
- Careful data collection and reduction to a form usable by
the simulator; these data should be easily accessible,
reusable and verifiable, which, in practice, means
maintaining specific project folders and spreadsheets and
ensuring version control
- Calculation of pipework volume and resistance to flow
through analysis of isometric drawings
- Model building as per piping and instrumentation diagrams
(P&IDs) and agreed scope, incorporating plant logic
(e.g., cause-and-effect charts) and isometrics; elements
typically included are:
a. Compressor system (gas/steam
turbines, electric motors, heat exchangers, vessels,
b. Compressor control, including
the anti-surge, hot-gas/cold-gas bypass and bleed
c. Regulatory control, emergency
shutdown logic and related valves
d. Startup and normal shutdown
- Integration of sub-models and
preparation of the final model, aligned at the agreed heat
and material balance (H&MB)
- A first report to the final user clarifying data and
general approaches used, comparing the model with the agreed
H&MB, and supporting the model review
- Model update, which can include client comments,
incorporation and review of data, preparation of repeatable
scenarios for test runs, defining test run duration and
specific integration time steps, etc.
- Execution of the test run, which involves collecting
data, graphically reviewing the results (i.e., process
variable as a function of time) for easy analysis, and
discussing first results with the client
- Delivery of the individual test run and reporting of
observations, conclusions and possible recommendations
- Delivery of the final report, including all individual
run reports and final recommendations.
LNG production process description. Natural
gas is first compressed in a feed gas compressor and then sent
to the mercury-removal unit. The gas is further treated in an
acid gas absorber to remove CO2 and H2S,
if present. The sweet gas is dried in molecular sieve beds and
further processed in the natural gas liquids (NGL) recovery
unit to remove the C2+ compounds. The chilling
requirement for the treatment is supplied by auxiliary propane
The treated natural
gas is then compressed in a residue gas compressor and
cooled by C3 chillers. Afterward, it is fed to the
bottom of the main cryogenic heat exchanger (MCHE), where it is
liquefied by mixed refrigerant (MR). This is the typical
liquefaction technology used in projects based on
a propane/MR process. The LNG produced is sent to the helium
recovery and nitrogen-stripping sections. The nitrogen- and
helium-free LNG is then sent to the storage tanks. The boil-off
gas (BOG) from storage is recompressed and sent to fuel gas.
The gas coming from the helium-recovery and nitrogen-stripping
units is compressed in the end-flash gas compressors and sent
to the fuel gas header.
The following compressor systems are considered in the
current dynamic simulation scope:
- Feed gas compressor
- Residue gas compressor and turboexpander
- Auxiliary propane compressor
- Refrigerant (C3/MR) compressors
- End-flash gas compressors
- Boil-off gas compressors.
Several test cases are reviewed in the following
Feed gas compressor startup. The natural
gas is fed to the feed gas compressor suction knockout drum to
remove any entrained or condensed liquids. The overhead vapor
is directed to the feed gas compressor. The compressed gas is
cooled and then directed to other LNG
plant units for further treatment. A simplified process diagram
is shown in Fig. 1.
Fig. 1. Feed gas compressor
simplified process flow diagram.
During startup, the feed gas compressor pressurizes other
downstream units to the residue gas compressor. The volume of
the downstream system is very large (about 3,000
m3). Thus, during a typical startup, when the
downstream pressure is low, our simulations show that the
compressor would operate in the stonewall region until the
downstream system is pressurized, as shown in Fig.
2 (red line).
Fig. 2. Compressor maps
for the feed gas
compressor system, with the movements of
the operating point during compressor
To avoid operating at high compressor flows, it was decided to
isolate the compressor from the downstream system during
startup. A valve bypassing the discharge on/off valve and
controlling the downstream pressure was used to smoothly
pressurize the downstream system. Results are shown in
Fig. 2 (blue line). In this case, the
anti-surge controller was allowed to control pressure while
increasing the compressor speed from minimum to operating
A further refinement to the procedure was implemented to
allow the anti-surge system to control when the compressor
reached the minimum speed. Results are shown in Fig.
2 (green line). The compressor startup procedure was
optimized, achieving smooth system pressurization while the
compressor was protected from surge.
End-flash gas system: Motor trip case.
Nearly 85% of the fuel gas is made by the end-flash gas coming
from the helium-recovery drum overheads and the
nitrogen-stripping column overheads. Before combining together,
both streams pass through the end-flash gas exchanger for heat
recovery. The combined stream is then compressed to about 32
bara. There are two compressor trains with three stages each. A
simplified process diagram is shown in Fig. 3.
Fig. 3. End-flash gas
process flow diagram.
The performance of the anti-surge system during motor trip
was studied extensively. It has been observed that the
first-stage compressor was surging upon trip (Fig.
4, blue line). To solve this problem, two options were
examined to improve the system design:
1. Use a check valve at suction (Fig.
4, green line)
2. Use a hot-gas bypass (HGB) valve (Fig.
4, cyan line).
Fig. 4. First-stage
fuel gas compressor map
during motor trip.
Both options appear valid. However, the results demonstrate
that, if an HGB valve is used, high temperatures occur at the
compressor discharge. This can be detrimental to the compressor
seals. Based on these findings, it was recommended to modify
the design and to use a check valve at the compressor
Liquefaction section: Refrigeration compressor
cases. Refrigeration for this process is provided by
two major systems: the C3 system and the MR system,
containing nitrogen, methane, ethane and propane. The
low-pressure MR is compressed in three stages of MR
compressors. The high-pressure vapor from the discharge of the
compression train is cooled and partially condensed by four
C3 chillers in series. The partially condensed MR is
fed to a high-pressure separator, where heavy MR and light MR
are separated before entering the MCHE.
The C3 refrigeration system utilizes
C3 evaporating at four pressure levels to supply
refrigeration to the natural gas feed and to the MR circuit.
The chilled natural
gas feed enters the MCHE, where it is liquefied.
Case 1: Verification of compressor valve
size. One of the objectives of the simulations
performed was to verify the sizing of the compressor anti-surge
valve. For the majority of the runs performed, it was noted
that, for the low-pressure C3 compressor anti-surge
valve, a valve size (calculated as CV) of 1,600 was adequate to
protect from the risk of damage due to surge. However, while
performing startup of the C3 system, the valve size
proved to be insufficient. Results of the test cases examined
are presented in the following figures.
Fig. 5 shows a map of the C3
low-pressure (first-stage) compressor with anti-surge valve CVs
of 1,600 (red line) and 2,600 (blue line), while starting up
the system with propane. Using a CV of 1,600, the operating
point appears to be in the surge region for a significant part
of its path. The higher CV (2,600) is needed to start the
compressor safely, with normal composition and pressure of 1.5
Fig. 5. First-stage
C3 compressor map
during startup with propane (Liquefaction
Case 1). The pink line shows the
curves at the highest and lowest speed and
the surge line.
However, when starting up the C3 system with defrost
gas or N2, the low-pressure C3 anti-surge
control valve requires an even higher CV to protect from surge.
A CV of 3,000 is needed to start the compressor safely with
N2 or defrost gas, together with a starting pressure
increase to 1.52.5 bara. The case of defrost gas is shown
in Fig. 6.
Fig. 6. First-stage
C3 compressor map
during startup with defrost gas
Case 1). The pink line shows the
compressor curve at the nominal speed
the surge line. The blue, green and pink
demonstrate the operating point using
different CVs. A high CV is required for
operation when defrost gas is used.
Compare with Fig. 5.
To start the compressor with different compositions
(C3, N2 and defrost gas), a
semi-automated approach was developed, based on the
aforementioned results. The plant operator knows the current
operation case (e.g., startup with defrost or propane gas) and
presses the appropriate button on the distributed control
system. The operator action ensures passing the necessary
information to the compressor control system. The compressor
control then adjusts the value of the first-stage C3
compressor anti-surge valve maximum opening/CV through a
software clamp, and it also defines the startup pressures and
The different approaches used in valve design by vendors are
worth mentioning. The typical valve design is based on a number
of distinct/discreet operation cases at steady-state
conditions. Each case may be based on different gas conditions
at the suction and discharge of the valve (e.g., gas
temperature, pressure, density, etc.).
With dynamic simulation, it is possible to explore a
continuous set of plant conditions. For example, during a
simulated compressor startup, all plant conditions from the
settle out/trip to the final steady-state conditions are
visited. These include intermediate conditions with increasing
compressor speed, varying suction/discharge compressor
temperatures, ΔP across the anti-surge valve,
In brief, using dynamic simulation, the engineering study
team was able to evaluate design modifications and to capture
cases where the original valve design was not satisfactory for
Case 2: MR compressor HGB valve design.
Another objective of this dynamic simulation study for the
low-pressure MR compressor was to confirm the requirement for
an HGB valve and its stroking time. Note that the low-pressure
MR compressor is provided with a bleed valve and a discharge
pressure control valve to flare.
A number of sensitivity test runs were conducted. It was
determined that the bleed valve opening time could be reduced
to 1 sec. Furthermore, it was shown that, by opening the bleed
valve quickly and the discharge valve to flare relatively
slowly, the HGB valve size could be reduced, or the valve
itself could be removed. Results are demonstrated in
Fig. 7. To compensate for any uncertainties,
it was decided to maintain the original design (i.e., maintain
the HGB valve with a small size).
Fig. 7. Low-pressure MR
upon MR compressor trip (Liquefaction Case
2). The pink line shows the compressor
curves at the highest and lowest inlet
vane positions and the surge line.
Case 3: Valve opening on compressor trip. As
soon as the C3/HP MR compressor driver was tripped,
the compressor speed decreased, as expected. However, the
simulation run showed a pronounced effect on the MR
compressors. The first stage of MR compressor was entering the
surge region upon C3 compressor trip.
A series of simulation test cases revealed that, to protect
the MR compressor from surge upon C3 compressor
trip, both the low- and medium-pressure MR compressor
anti-surge valves must open shortly after the C3
trip, following a feed-forward signal. The performance map of
the low-pressure MR compressor with the aforementioned
feed-forward signal to the low-pressure MR anti-surge valve can
be seen in Fig. 8.
Fig. 8. Low-pressure MR
on C3 compressor trip (Liquefaction
Lessons learned and execution methodology
It is challenging to perform a meaningful dynamic simulation
study. To ensure useful results, some principles must be
understood and basic guidelines must be followed:
- An engineering study is a dynamic process. It requires
continuous interaction between the EPC contractor, the
dynamic simulation provider, the equipment vendors and the
client at all levels (managerial, technical, etc.). Excellent
cooperation must be achieved to turn the study into important
findings and appropriate process design changes.
- Small teams of engineers with good process understanding
and the ability to cooperate will drive the project to success. The team
should remain the same throughout the project execution.
- The plant data used for the study must be accurate and
consistent. After some point, data changes should be
minimized to have consistent and comparable results, while
avoiding a series of costly sensitivity-simulation runs.
- For all test cases, detailed procedures are required and
should be communicated in writing. This step ensures common
understanding, targets and methodology.
- The simulation models must be of high fidelity. Special
attention should be devoted to all critical parameters that
could possibly invalidate the study results (isometrics,
valve sizes and timings, compressor and driver inertia,
compressor curves, etc.).
- Models should be approved before proceeding to results
generation. The following should be reviewed and
a. Model topology and match with
b. Main data used
c. Match with heat and mass
e. Compressor curves and inertia
- After the simulation models are prepared, it
is critical to debug the models by running a series of test
cases. These can reveal model weaknesses. At the same time,
early findings can be communicated so that the direction of
investigation can be set.
From a project management point of view, a dynamic
simulation study follows a relatively standard schedule that
can be changed according to project-specific needs and
requirements. The main elements are:
- The kickoff meeting is where the studys targets are
confirmed and data are provided.
- When the model is prepared, a report describing the match
of the models to the data and the heat and mass balance is
delivered. The model validation test follows. The delivered
models and associated reports are examined to ensure the
match with the actual plant design. This model review can be
an offline activity. However, face-to-face meetings are
advisable to improve the communication of findings. Such
meetings present ideal settings to finalize the procedure for
every simulation test case.
- After the model review meeting(s) are conducted, the
dynamic models are updated and retested, the individual cases
are examined, and the individual run reports are delivered.
Early findings are communicated, and the direction of
investigation is discussed.
- A number of review meetings should be arranged to closely
monitor the project execution and to discuss important
findings or challenges that must be addressed to avoid delays
and project risks.
- The dynamic study completion is accompanied by the
appropriate deliverables, including the final project report.
This report includes:
a. An executive summary
b. The study objectives
c. Model scope and
d. Stream comparison
e. Individual studies and
The work performed for the GNL3Z project illustrates that
dynamic simulation can be an excellent tool for the support and
verification of process design during the engineering,
procurement and plant construction (EPC) stages. The
advantages of dynamic simulation are becoming more important
and valuable, as evidenced by the fact that dynamic simulation
is increasingly included as a fundamental engineering step.
Experience has revealed several attractive attributes of
- A dynamic simulation study is a useful tool for process
verification and optimization of operating procedures, along
with control and protection systems. Through critical results
analysis and comparison with project constraints, it is
possible to improve the quality of process design. In turn,
the early implementation of improved solutions reduces total
- Dynamic simulation can reproduce the behavior of a real
plant and offer process insight that cannot be obtained with
traditional steady-state simulators. Using dynamic models and
studies through all of the engineering phases improves
knowledge of process dynamics. This understanding is
fundamental during plant startup and for the identification
and resolution of process bottlenecks.
- The evaluation of sequences and procedures involving
rotating equipment and important packages during the
engineering stage increases process controllability and reliability. Furthermore, the
machine/package vendor can verify and implement suggested
improvements in control functionality.
- By maintaining the dynamic model (i.e., ensuring that it
tracks the adjustments applied onsite during the
commissioning and startup phases), it is possible to extend
the life and value of the simulation well beyond the
engineering stage. Any update/improvement of the control
schemes, the creation of what-if scenarios, etc.,
can be easily implemented using the simulation study
In the current project, the use of dynamic simulation
delivered significant estimated value: around 1% of capital
expenditure ($30$60 million) at plant commissioning,
taking into account the minimization of reworks and the
avoidance of prolonged commissioning activities. Increased
process reliability is estimated to yield
around $10 million per year. To capture these benefits and
boost confidence in the plant design, it is critical to have an
appropriate project execution strategy that includes:
- A recognized and validated modeling tool
- An experienced project team with a significant track
- An accurately defined scope of work and a flexible project execution plan
- Maturity of design and availability of actual equipment
data during the early construction stage, which allows
for the timely implementation of study findings into the
Failing to take into account the aforementioned measures
significantly increases the risk that potential benefits will
not be realized. HP
1 Psarrou, S., Y. Bessiris, I. Phillips and V. I.
Harismiadis, Dynamic simulation useful for reviewing
plant control, design, Oil & Gas Journal,
Vol. 105 (30), August 13, 2007.
2 Al-Dossary, A., M. Al-Juaid, C. Brusamolino, R.
Meloni, V. Mertzanis and V. I. Harismiadis, Optimize
Plant Performance Using Dynamic Simulation,
Hydrocarbon Processing, June 2009.
3 Panigrahy, P., J. Balmer, M. A. Alos, M. Brodkorb,
B. Marshall, Dynamics break the bottleneck,
Hydrocarbon Engineering, Vol. 16 (9), September
4 Price, B. C., Small-scale LNG facility
development, Hydrocarbon Processing, January
5 Foglietta, J. H., Consider dual independent
expander refrigeration for LNG production,
Hydrocarbon Processing, January 2004.
6 Harrold, D., Design a turnkey floating LNG
facility, Hydrocarbon Processing, July 2004.
Anton Marco Fantolini is the LNG technology manager at
Saipem. He holds a degree in chemical engineering and
has 14 years of experience in the oil and gas
industry, with concentrations in LNG plant design and
development during the conceptual, FEED and EPC
Pedone is a lead control engineer with
Saipems automation and control department. He
holds a degree in chemical engineering and has 12 years
of experience in the oil and gas industry. His areas of
focus include dynamic and operator training simulators,
management information systems, and advanced process
DOrazi is a process engineer with
Saipem. He holds a degree in chemical engineering and
has six years of experience in the oil and gas
industry, with a concentration in plant
Ramona Prodan is a technical
engineer specializing in control systems and
automation. She has more than five years of
professional experience in the oil and gas industry.
She holds a BS degree in chemical engineering from
the Petroleum Gas University of Ploiesti,
Ankush Sood is a team leader with
Hyperion Systems Engineering in Pune, India. He has seven years
of experience in the oil and gas industry, including
five years of delivering engineering studies for
centrifugal compressors and developing high-fidelity
operator training simulators. He holds a BSc degree
in chemical engineering from Punjab Technical
Girish Bhattad is a team leader
with Hyperion Systems Engineering in Pune, India. He has eight years
of experience in the oil and gas industry, including
four years of developing high-fidelity operator
training simulators and delivering engineering
studies. He holds an MTech degree in process
engineering and design from The Indian Institute of Technology
Dionysios Stavrakas is a senior
engineer at Hyperion Systems Engineering in Athens,
Greece. He has five years of experience in delivering
high-fidelity engineering studies and developing
operator training simulators. He holds an MSc degree
in chemical engineering and energy production and
management from The National Technical University of
Vassilis Harismiadis is the
real-time optimization and training simulation
manager at Hyperion Systems Engineering. He has 12
years of experience in the oil and gas industry, with
a focus on using dynamic process modeling to improve
plant efficiency. Dr. Harismiadis holds a PhD in
thermodynamic modeling of complex systems from NTU