UniSim Design 用于LNG 液化过程的模拟
tube bundles. The liquid from the HP sepa-rator passes through the first (warm) bundle of the MHE, where it is sub-cooled. It is then flashed into the shell at the warm bun-dle top, joining with the refrigerant from the top (cold) bundle to provide refrigera-tion. Vapor from the HP separator passes through both bundles where it is partially condensed. It is then flashed into the shell to provide refrigeration for the top bundle. As the mixed refrigerant progresses down the shell toward the compressor suction, the liquid becomes heavier in composition and boils at higher temperatures, provid-ing evaporative cooling at a continuum of temperatures. The last amount of liquid is vaporized in the bottom bundle and the resulting mixed refrigerant vapor is super-heated before reaching the compressor.
Alternatively, the MHE can have three tube bundles rather than the two bundle configurations, as illustrated in Fig. 2, that shows a high-level flowsheet for dynamic simulation of an LNG plant. With the three-bundle configuration, the bottom bundle serves as the condensing heat exchanger for the fractionation (scrub) column, rather than using the precool-ers for this purpose. Vapor (almost pure natural gas) from the reflux drum of the scrub column is re-introduced into the
main heat exchanger at the bottom of the
middle bundle where it is cooled further.
Also, the natural gas pressure is reduced
through a Joule-Thomson valve before final
cooling against the low-pressure refriger-
ant in the top bundle. Product purity is
adjusted using liquefied petroleum gas,
which is cooled and at least partially con-
densed in the bottom and middle bundles
prior to being mixed with the natural gas
at the bottom of the top bundle as it enters
the bottom bundle of the MHE.
Main heat exchanger. A multi-tube,
spirally-wound heat exchanger is made
up of tubes that are spirally wound on a
mandrel, as thread or cable is wound on a
spool.4 As shown in Fig. 3, a layer of tubes
is wound (left to right) on the mandrel and
spacers (bars, wire, etc.) are attached to
them. This is followed by a second layer
of tubes wound in the opposite direction
(right to left) and then a third layer (left
to right again), each layer complete with
its own set of spacers. This procedure is
repeated until the required number of tubes
has been wound onto the mandrel.
The longitudinal distance between the
tubes in a layer and the tube inclination
are kept constant for all layers. For the
large exchangers used in LNG plants, the
tube diameter ranges from 3?8 in to 3?4 in
and the tubes are applied to the mandrel
with a winding angle of approximately
10°. The tubes are connected to tubesheets
at each end of the heat exchanger and each
layer contains tubes from all the differ-
ent streams so the shell-side duty is uni-
form. The heat exchanger operates in
total counter-flow, with evaporating fluid
flowing downwards on the shell side and
high-pressure, condensing fluid flowing
upwards on the tube side.
For the multi-bundle exchangers used
in natural gas liquefaction processes, the
bundles are housed within a single shell.
Additionally, there is a reservoir for each
bundle within the mandrel to collect and
redistribute the liquid phase of the refriger-
ant over the annular rings within the shell
of the tube bundle.
M o d e l i n g t h e m a i n h e a t
exchanger.It is evident from the process
description that the basic unit operation
required to model the MHE is a spirally-
wound shell-and-tube heat-exchanger bun-
dle having multiple tube streams and a sin-
gle shell stream. Although numerous papers
LIQUEFIED NATURAL GAS DEVELOPMENTS
have been published and/or presented at
conferences that discuss modeling of LNG
processes on a qualitative basis, there are few
publications that discuss these modeling
processes, in particular modeling the main
heat exchanger, on a quantitative basis.
A simplified model of a spirally-wound
tube bundle will not predict the expected
dynamic process behavior over the range of
operation for which dynamic simulation is
required. For example, a simplified model
will not accurately predict startup dynam-
ics, when, during initial startup, volumetric
capacitance influences the refrigerant charg-
ing procedures and compressor suction
conditions are influenced by the refrigerant
supply as a function of the exchanger duty.
Simplified modeling of heat exchangers also
produces irrational temperature profiles
with crossovers at segment boundaries and
between individual shell-and-tube streams.
Consequently, a first-principles math-
ematical model for a tube bundle of a
spirally-wound heat exchanger, employing
rigorous physical property calculations and
thermodynamic flashes, was developed as a
dynamic unit operation of a process model-
ing package. This unit operation, called the
spirally-wound tube-bundle module, when
used in a flowsheet with the standard unit
operations of process modeling, reflects
the behavior of natural gas liquefaction
processes with the fidelity, reliability and
robustness necessary to yield meaningful
results over the range of process operations
typical of dynamic simulation studies and
simulation-based training of process opera-
tors. The spirally-wound tube-bundle mod-
ule predicts:
? Exit flow, temperature, pressure,
vapor fraction and composition for each of
the outlet streams
? Phase change within each of the tube
streams and the shell stream
? T ube and shell wall temperatures
? Intermediate temperatures along the
heat exchanger
? Thermal profiles in the shell wall and
insulation.
Fig. 4 shows the standard views of the
spirally-wound tube-bundle module of the
process modeling package, illustrating a
great detail of what is captured in the model.
In large-scale, real-time and faster-than-
real-time dynamic simulations typical of
dynamic studies and simulation-based
operator training, fidelity and calculation
speed are always competing objectives.
Simplifying assumptions, such as using a
representative tube winding for each tube
stream and lumping the shell-side annular
rings into a single shell stream, were made
when formulating the mathematical model
so as to balance these objectives.
The model formulation incorporates
an axially distributed model for the mate-
rial flows in the multiple tube streams and
the shell stream, and an axially and radi-
ally distributed model for the heat flow
through the tube walls and the shell wall
and insulation. T o predict phase change in
the tube streams and the shell stream, the
model for the material flows incorporates an
isobaric-isenthalpic (PH) flash at each grid
point. The solution of a spatially distrib-
uted model incorporating flash calculations
for a multiple-tube stream countercurrent
flow configuration is very challenging from
a computational perspective —stability,
robustness and speed. Solution stability is
addressed by employing the equations-ori-
ented solution architecture that solves all the
modeling equations for the unit operation
simultaneously. Solution robustness and
calculation speed are addressed by replacing
the highly nonlinear PH flash equations by
first-order Taylor series expansions whose
coefficients are updated by exception as the
solution moves through the operating space
and by employing a multilayer grid for the
process streams, calculating some quantities
on a course grid and projecting values for
these quantities onto the finer solution grid.
The model formulation and solution
methodology employed in the spirally-
wound tube-bundle unit operation is
proven technology, having been successfully
deployed in dynamic simulation models of
more than 10 LNG plants.3
The power of dynamic simulation.
The key value of dynamic simulation is
the improved process understanding it
provides.6 After all, plant operations are
by nature dynamic. Realistic dynamic
models can be used to enhance the design
of the control system, improve basic
plant operation, and train both opera-
tors and engineers.
Plant life cycle—early stages. In
the design phase, dynamic simulation mod-
els can help identify operability and control
issues and influence the design accordingly.
They serve as valuable tools for designing,
testing and tuning control strategies prior
to startup. They can also be used for recon-
ciling trade-offs between optimized steady-
state design (targeted at minimizing capital
expenditures and operating utility costs)
and dynamic operability. In addition, such
models often assist in the development
of operating procedures. However, using
dynamic models for training plant opera-
tors before commissioning is, by far, the
most well-known application of dynamic
simulation.7 With a good understanding
of the production process and knowledge
of the control procedures applicable to nor-Spirally-wound heat exchanger
with four streams.5
FIG. 3
training.8 Analysis has shown that approxi-mately 90% of plant incidents are prevent-able and that the majority of incidents—by some estimates the vast majority—result from the actions or inactions of people. Because people will always play an integral role in plant operations, continuous train-ing of plant personnel is crucial to achieving safe, reliable and efficient operation.
Dynamic simulation has the power to create significant value throughout the life cycle of a project, from initial investigation of the processing concepts right through to plant operation. Although this value is described here in broad terms without specific reference to LNG projects, it can certainly be realized in LNG projects, as shown by the following case study.
Case study—Ras Laffan LNG—Train 3. A precommissioning dynamic simulation study (DSS) was undertaken for T rain 3 of the Ras Laffan LNG facility to confirm operational readiness of key plant assets.3 The dynamic model encompassed the liquefaction process (feed dryers, feed pre-coolers, scrub column and main cryo-genic heat exchanger) and the refrigeration process (closed-loop mixed-refrigerant and propane compression system).
The DSS was conducted during the front-end engineering design (FEED) and detailed design stages of the project. Dur-ing FEED, the objective of the DSS was to confirm whether the project specifica-tions and plant design basis were suitable for equipment selection, and whether the control strategies met operability and asset-protection requirements. During this study phase, a simplified control implementation was necessarily employed because the con-tion compressor and the LNG and mixed
refrigerant turbines; and the simplified
control implementation was replaced with
the actual control system, emergency shut-
down logic, gas turbine startup sequences
and compressor anti-surge control. Evalu-
ation of the automation system was critical
to Ras Laffan because its configuration was
new and unique. The simulations performed
during the initial phase of the DSS were
repeated and supplemented by six additional
simulations using the updated and extended
dynamic model.
Generally, the DSS showed that the
control strategies were sufficient to protect
the equipment and personnel during upset
situations and that the new and unique
automation system was effective. A sig-
nificant finding from an operability per-
spective was sensitivity of the compressors
to overload during upset conditions with
high flow rates. However, possibly the
greatest single result of the DSS was the
confidence it provided in readiness for safe
operation through realistic simulation of
the many operating scenarios investigated.
Following the conclusion of the DSS, the
focus of the dynamic model shifted from
engineering to operation. Operating pro-
cedures were prepared and then validated
against the dynamic model, and process
operators were trained on process funda-
mentals and process operation during nor-
mal operation and abnormal situations.
Conclusion.Dynamic simulation has the
power to create significant value through-
out the life cycle of an LNG project, testing
and refining the design, virtually commis-
sioning the control system prior to startup,
training operations personnel both before
and after initial startup, troubleshooting
operating problems and validating pro-
posed changes to plant operations before
implementation. Addition of the spirally-
wound tube bundle module to the pro-
cess modeling package enables this value
to be realized for mixed refrigerant LNG
facilities. This is proven dynamic simula-
tion technology, having been deployed in
numerous dynamic simulation studies and
operator training systems.HP
LITERATURE CITED
1 Edwards, T. J., C. F. Harris, Y. N. L iu and C.
L. Newton, “Analysis of Process Efficiency for
Baseload LNG Production,” Cryogenic Processes
and Equipment, Fifth Intersociety Cryogenics
Symposium, ASME, New Orleans, 1984.
2 L om, W. L., “L iquefied Natural Gas,” Applied
Science Publications, 1979.
3 Henderson, P., H. Schindler and A. Pekediz,
“Dynamic Simulation Studies Help Ensure Safety
by Conforming Operational Readiness of L NG
Plant Assets,” AIChE Spring Conference, New
Orleans, 2004.
4 Crawford, D. B. and G. P. Eschenbrenner, “Heat
T ransfer Equipment for LNG Projects,” Chemical
Engineering Progress, Vol. 68(9), p. 62, 1972.
5 Fredheim, A. and P. Fuchs, “Thermal Design of
L NG Heat Exchangers,” Proceedings for the
European Applied Research Conference on
Natural Gas, T rondheim, Norway, p. 567, 1990.
6 Svrcek, W. Y., D. P. Mahoney and B. R. Yong, “A
Real-Time Approach to Process Control,” John
Wiley and Sons, L td., Chichester, England,
2000.
7 Tang, A. K. C. and G. Stephenson, “L NG
Plant Operator T raining,” Petroleum T echnology
Quarterly, Autumn, 1997.
8 Stephenson, G., P. Henderson and
H. Schindler, “Profit More from Process
Simulation,” Chemical Processing, August, 2009.
eywell Process Solutions. Based in London, Ontario,
Canada, he has worked in the field of process simula-
tion for more than 35 years and has held positions with
DuPont, Atomic Energy of Canada, the University of
Western Ontario’s Systems Analysis Control and Design
Activity (SACDA), and Honeywell. Mr. Stephenson is the
originator of the Shadow Plant dynamic simulator and
is a pioneer of the hybrid solution architecture and its
application to large-scale dynamic simulation. He has an
MS degree in applied mathematics.
from the University of Ottawa. She has hands-on expe-
rience with process simulation and specializes in chemi-
cal engineering thermodynamics. Ms. Wang has also
worked at the National Research Council of Canada as
a research scientist.