Вы находитесь на странице: 1из 5

Originally appeared in:

September 2006 issue, pgs 69–75


Used with permission.
www.HydrocarbonProcessing.com

Article copyright © 2006 by Gulf Publishing Company. All rights


reserved. Not to be distributed in electronic or printed form, or posted
SPECIALREPORT
on a Website, without express written permission of copyright holder.

Analyzing the value of first-


principles dynamic simulation
Use it to test new control schemes before implementing them in a plant
R. GONZÁLEZ, Petronor, Repsol YPF Group, Bilbao, Spain;
and J. M. FERRER, AspenTech, Barcelona, Spain

C
ontrol engineers can make practi-
cal use of first-principles dynamic �����
� ��������� ����� ����
modeling technologies, which are ���� �������� �����
now becoming mature enough to provide ������
�� ������� ������� ����
� ��������� �������
significant value for control purposes. The ��
���� ����� ��������
modeling efforts used in this case study have ��� ������ ��������

�� ���������� ����
been analyzed from the control engineer’s ����� ���
perspective, including a discussion of best �������� ��� ��� ��� ����
���
�� ����� ��� ����
practices to optimize the use of internal ���� �����
resources. ���� ����� ����� �������
A first-principles dynamic model of a ������� ���
� � � ��
depropanizer column is used to test and ����� ����� � ���
�������� � ������� ����
to confirm a new proposed operation and ����
����� ���� �����
control mode that improves the steam con- ���������
���� ��� ������� �� ��
� �
sumption and the stability of the column. ���� ������
���� �
The previous column operation mode was �������� �� ���� ���
��
��� ��
not suitable for the planned advanced pro- ���������� ���
�������
cess control (APC) project in the area of
this unit, so a new basic control scheme FIG. 1 Depropanizer and associated controllers.
was proposed, simulated and validated in
the rigorous dynamic model. The new con-
trol scheme was tested and implemented in
the plant confirming the predicted results �����
� ��������� ����� ����
�������� �����
of the model. ����
������ �������
�� ������� ����
� ���������
Introduction. This study was prompted �� ����� ��������
��� ������ ��������
by the modeling problems found in the �� ����������

����
new FCC unit APC system. After some ����� ���
�������� ��� ��� ���
step tests, the identified multivariable ��� ��� ����
�� ����� ��� ����
controller models showed some problems �������� �����
���� ����� ����� �������
related to the heat and material balance in ������� ���
the FCC gas unit depropanizer column, � � � ��
����� ����� � ���
specifically associated with the head pres- �������� � ������� ����
����� ���� �����
sure and bottom and top qualities. ����
���������
A golden rule in APC projects is to ���� ���
� ������� �
�� ��
solve all basic control problems first, 1 ���� ������
���� ���� �
�������� �� ���
since step testing is usually a major exer- ��� �� ��
�������
cise that no one will want to repeat until ���������� ���
process changes require it. Therefore, a
more detailed study of the depropanizer FIG. 2 New control scheme.
was required.
HYDROCARBON PROCESSING SEPTEMBER 2006
SPECIALREPORT REFINING DEVELOPMENTS

• The 10 –12% imbalance could all result from flow through


����������������� ������������ the balance line.
• Difficulty in controlling the column pressure (PC-20) can
������������� be explained by a significant flow through the balance line.
�������� ���������� �������
������������
����
����� ����� ����� This balance flow is out of control since it is determined by the
upstream (column) and downstream (D-10) pressures, which are
������������� ������������ affected by the ambient conditions. For example, if a rainstorm
��� ��� ������������� decreases the column pressure, then the balance flow is lower
����������� ������� ������������� and pressure in the feed drum (D-10) also decreases. This in
����������� ���������� turn causes the load to the column to be reduced, which again
reduces the column pressure.
��������������� • To improve the column control it is necessary to look for a
��� different pressure source, so that it is possible to completely close
the balance line.
FIG. 3 Dynamic modeling methodology.

Proposed solution. Feed drum D-10 is working in equilib-


rium, so the pressure is determined by the feed composition. The
Process and problem description. The depropanizer only way to set a desired pressure is to inundate it completely, and
column (T6) with the feed drum (D-10) and condenser drum use the feed drum as a filled pipe segment (Fig. 2).
(D-11) is represented in the DCS screenshot in Fig. 1. The flow control valve FV-69, previously used by FC-69 and
The study of the unit focused on two specific areas of concern: LC-26 (D-10 level), is now used by the new PC-19 pressure
mass balance and column pressure control. controller. It is not needed to control the feed flow to the column
The steady-state mass balance equation: since it is imposed upstream by the debutanizer column pressure
controller.
column feed (FC-69) = bottom extraction (FC-74) + distillate (FC-14) It is also important to note that with the drum D-10 totally
full, the capacity to absorb load variations is lost. But D-10 geom-
cannot be matched with the plant data. Other significant streams etry and flow conditions reveal that the real residence time is
must need to be considered in the mass balance equation since three minutes, so the dumping capacity was, therefore, almost
there is a 10 –12% mass imbalance. negligible.

The column pressure (PC-20) is controlled through the distil- Simulation platform. To validate the proposed new control
late flow controller (FC-14), which varies the condensing capac- scheme, a first-principles dynamic model of the unit was cre-
ity of the condenser by partially flooding the shell side, since the ated. Building a model has traditionally been considered very time
condenser drum (D-11) remains always totally full of liquid. The consuming and requiring expert skills to achieve valuable results.2
usual response time of this kind of system (pressure vs. distillate Depending on the chosen simulation platform, however, this effort
flow) should be less than one hour, but the step tests revealed that can be reduced significantly.3,4 Ideally companies should consider
the column needed three hours to stabilize the pressure after a the following criteria when selecting simulation technologies:
change in the distillate flow. This affected control performance, • Reusability of steady-state process models for the dynamic
causing oscillations to the propylene-propane splitter feed. simulation
• Consistent and easy conversion to dynamic models
Problem analysis. These two problems indicated that some- • Rich set of process control objects, including real APC con-
thing different is happening in this column compared to standard trollers
binary distillation columns. The difference in this case is a pres- • Mature and widely proven technology
sure balance line from the top of the column to the feed drum • Intuitive, quick-to-build and easy-to-use, requiring little
(D-10) (Fig. 1). training
The original purpose of this balance line was to keep a certain • Fast simulation execution.
pressure in the feed drum (D-10) to assure enough NSPH in pump Some software-independent factors that will contribute to
P-15 under all operating conditions. The initial design estimated a make the dynamic modeling task easier are:
flow of 1 m3/h through this balance line to keep the right pressure, • Quality of the steady-state models from process engineering:
but in real operation, it was not possible to pressurize drum D-10 Most of the steady-state models handled by process engineering
when the column load was 90 m3/h. This fact suggests that the real departments are updated from the design conditions to the real
flow through the balance line could be well above 1 m3/h. operating conditions, but since they are not needed, they will not
The lack of pressure control in drum D-10 limits the attainable contain all the valves, bypasses, pumps, etc.
feed flow. The flow control valve FV-69 bypass (FC-69) has to • Access to equipment design engineering data: A quick and
be opened to reach the right flow, and even the preheater (E-21) easy electronic archive of all datasheets and design details will
bypass has to be opened as well to reduce the heat and thus reduce simplify data collection.
the backpressure. • Access to real plant process data: A good process information
Therefore, it was concluded that: system and historic database will facilitate access to process data
• To improve column controllability, proper feed control is for calibration.
needed, and this requires the right pressure in the feed drum • Instrumentation level and quality: A reliable and well-instru-
(D-10). mented plant will make model calibration easier.
HYDROCARBON PROCESSING SEPTEMBER 2006
REFINING DEVELOPMENTS SPECIALREPORT

FIG. 5 Operation change sequence.

by the process engineering group don’t need to incorporate all


the minor equipment and devices like valves, pumps, filters,
FIG. 4 Depropanizer dynamic model. pipe segments or bypasses. To replicate the plant conditions, it
is, therefore, necessary to introduce most of these items into the
same steady-state model. Most of the data like volumes or con-
• Understanding the process: The experience of the control trollers will not be used in the steady-state mode, but it is entered
engineer with the unit being modeled is an advantage when build- once to simplify the process and keep the entire data model in
ing a dynamic model. one file case.
• Unit complexity: Distillation, compression and heat transfer 3. Calibrate the model with process data: This is the most criti-
systems are well know and require less time to achieve reliable cal phase of the process since the model has to be manually tuned
results. Reactors, however, require more knowledge of the kinetics, and calibrated to reflect the true conditions in the real plant. The
which are not always well known. user has to equilibrate the differences by changing boundary con-
ditions by a small amount: delta P and OP of valves, efficiencies,
Modeling methodology. The depropanizer steady-state heat losses, pressure drops, etc. Sometimes this task is not trivial,
model was provided by the Petronor process engineering group, and frequently other hidden equipment or device problems are
but additional efforts were required to build the dynamic model.5,6 found when the model resists matching the plant data. Once
This involves a series of tasks that could be jointly executed by the the model is calibrated, all the heat transfer coefficients or flow
process engineering group and process control groups. The steps resistance factors are calculated, and all pressure-flow relations
followed are shown in Fig. 3. are correctly configured, so that the steady-state model is ready
Percentages shown in the figure represent the estimated spent to switch to dynamic mode when required.
time for every step; for the depropanizer case the four steps took 4. Switch to dynamic and stabilize: The solver is switched to
about ten days for a control engineer with no previous experience dynamic mode and the integrator started. The unit should then be
in dynamic modeling. This can be normally reduced to three or stabilized (note that static heads are not considered in steady state)
four days once the control engineer gets more familiar with the and all the controllers can be put in auto. The model is then ready
tool and methodology. The steps were: to be used for the operating changes and control studies (Fig. 4).
1. Collect design and plant data: First, all equipment data of
the depropanizer and associated equipment were collected: Testing the model. The new proposed controller design was
• Valves: Cv , characteristic, delta P, from valve datasheets incorporated in the dynamic model, and the same operation change
• Relief valves: type, setting pressures, from valve datasheets procedure was programmed. The sequence is shown in Fig. 5.
• Pumps: head and efficiency curves, from pump test The tested sequence began with switching off the old control-
certificates lers (LC-26, FC-69) and activating the new pressure controller
• Vessels: volume, height, nozzles, level taps, from design (PC-19). Then, the bypasses were closed and the balance gas line
datasheets is closed with a ramp. This transient is shown in Fig. 6.
• Column: tray spacing and type, diameter, weir height and The results show how the D-10 pressure (red trend line)
length from design datasheets rises to 11.3 Kg/cm2_g when the FV-69 and preheater bypasses
• Elevations: from plant floor elevations sheet are closed. Then, the balance line valve starts to close and the
• Controllers: PV, OP, SP, range and tuning gains from DCS. flow goes from 8.3 m3/h to zero (brown trend), and logically the
For the process data, a relatively short period (one day) was column feed decreases the same amount (from 88.3 to 80 m3/h,
selected when the process stayed at steady conditions and then the green trend).
average of all available instrumentation devices (pressures, tem- At the same time, the vessel is filled (blue trend) since it is not
peratures, flows, level, controllers’ PV, OP, SP ) was calculated. in equilibrium; the D-10 liquid temperature is decreased 12°C
Laboratory composition analyses from the feed and products were (from 44°C to 32°C, pink trend) because the heat source from the
also taken for that day. It was important to ensure that the process hot gas balance line has been closed. It is also observed that, even
didn’t suffer major changes during the selected period, otherwise it with the D-10 cooler liquid content at 32°C, the column pre-
would have been difficult to calibrate the model later. heater outlet temperature rises 3°C (from 69°C to 72°C, orange
2. Introduce engineering data into the model: Models used trend) because there is less feed flow to heat.
HYDROCARBON PROCESSING SEPTEMBER 2006
SPECIALREPORT REFINING DEVELOPMENTS

FIG. 6 Simulation: gas balance line closing and D-10 filling.


FIG. 7 Simulation: Reflux and steam reduction with same
qualities.

FIG. 8 Real plant: Gas balance line closing and D-10 filling. FIG. 9 Real plant: Reflux and steam reduction with same qualities.

When the hot gas balance line is totally closed and D-10 is In this example, it was possible to test the changes in the real
completely full, the reflux flow and the steam to the reboiler are plant, however, this is not always the case. Dynamic simulation
adapted to the new configuration to achieve the same product has even greater potential value for changes that cannot be tested
qualities. Fig. 7 shows how the reflux flow (green trend) and the in normal operation and which necessitate a physical plant change
steam (red trend) are both reduced by 10%. that requires a planned shutdown.
The dynamic simulation results confirmed that the new opera- To improve internal efficiencies, steady-state models provided
tion scheme was feasible and operable with better control of the by a process engineering group could already be calibrated with
D-10 pressure, less steam consumption and more stable product the plant conditions, making it easier and faster for control engi-
qualities. neers to build the dynamic model. Cooperation between the two
groups is required to establish an optimum modeling workflow
Plant tests. The new proposed scheme was tested and then methodology.
implemented into the plant, where similar results were obtained. In addition to performing basic control studies, dynamic models
Figs. 8 and 9 show real plant data, with the same trend lines and could be used as a support tool in certain APC projects.7,8 HP
colors than in Figs. 6 and 7.
Plant results in Fig. 8 confirm the same reduction in feed flow LITERATURE CITED
1 King, M., “How to lose money with Advanced Controls,” Hydrocarbon
(green trend) and achievable pressure control in D-10 (red trend).
Processing, June 1992.
D-10 temperature (pink trend) and preheated feed (orange trend) 2 Mahoney, D. P. and B. D. Tyréus, “Applications of Dynamic Simulation,”
evolved as predicted in Fig. 6. Proceedings of the Chemical Engineering Chemputers II Conference, March 1994.
Plant results in Fig. 9 reduce the reflux (green trend) and steam 3 Stanley, P., “Dynamic Simulation for Insight,” Chemical Processing, December

(red trend) with a less aggressive ramp than in Fig. 7, but the plant 1999.
4 Feliu, J. A., I. Grau, M. A. Alós and J. J. Macias-Hernández, “Match your
reached similar quality levels.
HYDROCARBON PROCESSING SEPTEMBER 2006
REFINING DEVELOPMENTS SPECIALREPORT

process constraints using Dynamic Simulation,” Chemical Engineering 8 Alsop, N. and J. M. Ferrer, “What dynamic simulation brings to a Process
Progress, December 2003. Control Engineer—Applied Case Study to a Propylene/Propane Splitter,”
5 McMillan, G. K. and R. A. Cameron, Models Unleashed, Virtual Plant and ERTC Computing, London, May 2004.
Model Predictive Control Applications, Chapter 1, 2004, ISA, ISBN 1-55617- 9 Mahoney, D. P., “HYSYS An integrated system for process engineering and

857-3. control,” Hydrocarbon Asia, November/December 1994.


6 Luyben, W. L., “Plantwide Dynamic Simulators in Chemical Processing and 10 Contreras, J. and J. M. Ferrer, “Sagunto LNG terminal dynamic simula-

Control,” 2002, Marcel Dekker, Inc., ISBN 0-8247-0801-6. tion—Design’s verification of Emergency Shutdown system and transient
7 Trivella, F. and G. Marchetti, “Integration for innovation,” Hydrocarbon studies,” Hydrocarbon Engineering, May 2005.
Engineering, November 2004.

R. González Martín holds a BSc degree (University of Mon- J. M. Ferrer Almazán holds an MSc degree in electrical engi-
dragon) and an MSc degree (University of Manchester Institute of neering (University of Zaragoza, Spain). He worked four years for
Science and Technology) in control engineering. He has 14 years’ Dow Chemical at the Tarragona site as process control engineer,
experience in advanced control; four years developing advanced three years at Electronic Data Systems in automatic warehouse sys-
control applications for the Machine Tool Division of the Mon- tems and five years at Aspentech developing applications of dynamic
dragon Cooperative and 10 years designing and implementing advanced control simulators in advanced control, ESD design, controllability studies, DCS checkout and
applications for Repsol-YPF in the Petronor refinery, Bilbao, Spain. Mr. González also operator training. He can be reached at josemaria.ferrer@aspentech.com
lectures on process control systems at Mondragon University. He can be reached at
rgonzalezm@repsolypf.com.

Article copyright © 2006 by Gulf Publishing Company. All rights reserved. Printed in U.S.A.
Not to be distributed in electronic or printed form, or posted on a website, without express written permission of copyright holder.

Вам также может понравиться