Sie sind auf Seite 1von 16

See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/293516566

Refining: Crude column relief design

Article · April 2006

CITATION READS
1 515

4 authors, including:

Cal Depew Ralph-Uwe Dietrich


Schneider Electric German Aerospace Center (DLR)
13 PUBLICATIONS   21 CITATIONS    86 PUBLICATIONS   118 CITATIONS   

SEE PROFILE SEE PROFILE

Some of the authors of this publication are also working on these related projects:

Kraftstoffmodellregion View project

Fuels Joint Research Group (FJRG) View project

All content following this page was uploaded by Ralph-Uwe Dietrich on 05 April 2016.

The user has requested enhancement of the downloaded file.


Under the Auspices of the Kingdom of Bahrain
Petrotech 2003 - The 4th Middle East Refining and Petrochemicals Exhibition and Conference
29th September - 1st October 2003

Plant-data Validated Model predicts Crude


Column Relief Loads
1
Howard B. Jemison, Ph.D.
2
Ralph-Uwe Dietrich
1
Invensys Process Systems
2500 City West Blvd., Suite 1200
Houston, Texas 77042
U.S.A.
2
Invensys Systems GmbH
Heerdter Lohweg 53-55
D-40549 Düsseldorf
Germany

This paper will give you an example of the advantages of using


SIMSCI-ESSCOR’s dynamic simulation tools. With Dynsim, we
built a detailed rigorous model of a crude distillation unit for a
German client and determined the flow out of the pressure relief
valves during different scenarios.

It contains details about a specific business case for building a


rigorous dynamic process model, the scope of work, problems
regarding the model validation, crude tower behavior at emergency
situations and further possible usage of that detailed process model.

INTRODUCTION
OMV Deutschland GmbH in Burghausen, Germany (OMVD), is a 3.5
Mio tonnes crude oil p.a. refinery within the OMV, oil and gas group
in Europe and the largest company quoted on the Austrian stock
exchange ATX.

With the mandatory regularly shut down OMVD introduced changes


in their crude distillation process design and included the results of
a recent column mechanics analysis into their operating procedure.
As a result, the maximum allowable pressure was reduced and the
set pressures of the overhead relief valves had to be lowered. A
new redundant ESD system provides additional safety to the
operation. Before Start Up, the national safety authorities (TueV
Sueddeutschland) requested a recertification of the safety system.

C:\HBJ Admin\Papers\Plant-data Validated Model predicts Crude Column Relief Loads PETROTECH
2003 paper.doc
Page 1 of 15
Under the Auspices of the Kingdom of Bahrain
Petrotech 2003 - The 4th Middle East Refining and Petrochemicals Exhibition and Conference
29th September - 1st October 2003
Conventional methods for safety systems design using steady-state
models predict higher loads, if the difference between operating
pressure and the maximal permissible column pressure gets
reduced. In this case, the result of the conventional calculation was
that the relief system can not handle these higher loads.

Before OMVD had to invest in an expensive upgrade of the relief


system, it further investigated, what relief loads really would occur,
if the column will get depressurized quickly enough to avoid any
pressure overshoot. Conventional methodologies make many
assumptions concerning the nature of the event. These
assumptions often lead to total pressure relief loads that are overly
conservative, predicting relief loads that are in excess of actual
relief loads experiences in refineries. These methods do not account
for the following conditions:
 Strong interaction between heat-integrated equipment, such as
pumparound heat exchangers
 Interactions between different relief devices connected to a
common flare header system
 Time delay to reach the relief pressure after an event, and the
maximum pressure obtained
 Change of relief flow composition after the initial opening of the
valve
 The effect of operating controls is not considered, although some
of the control responses happen automatically (e.g., ESD logic).
Some regulatory control action exacerbates relief scenarios.

In recent years, the use of dynamic simulation to predict more


accurate relief loads has become used and accepted in the
petroleum and petrochemical industries1-5. Some literature sources
believe the overestimation of total relief rates and loads based on
these procedures is 50-200%. An accurate rigorous dynamic model
of the process, the major control loops, and the associated relief
system (including the pressure relief valves) gives a more detailed
and accurate description of the process interactions during a
relieving scenario.

To do so, OMVD chose SIMSCI-ESSCOR, Lake Forest, California, to


perform an engineering study to determine the relief loads of the
crude unit for various abnormal conditions and relief scenarios
because of the following main reasons:
 SIMSCI is well known in this refinery for the delivery and the
ongoing support of its steady-state simulator, PRO/II.
 SIMSCI were developing a dynamic simulator at that time and
had provided relief studies with Dynsim in the past.

C:\HBJ Admin\Papers\Plant-data Validated Model predicts Crude Column Relief Loads PETROTECH
2003 paper.doc
Page 2 of 15
Under the Auspices of the Kingdom of Bahrain
Petrotech 2003 - The 4th Middle East Refining and Petrochemicals Exhibition and Conference
29th September - 1st October 2003
 Quick agreement during the problem discussion to define the
solution pathway, of how to get the required answers and to get
the governmental authorities approval as well.
 Commitment of both partners to investigate a process in depth
and get reliable conclusions out of this study.

SCOPE OF WORK
A crude unit is a relatively safe process to operate in a refinery.
The feed and products are in the liquid phase under normal
conditions, and vapor must normally be added to the overhead
system in order to maintain a positive pressure. Therefore,
thermodynamics and normal states of the crude oil components are
working in our favor.

However, a large amount of energy must be added through the


furnace in order to achieve proper separation. The management of
this heat is the key to maintaining safe operation. Once this heat is
added to the material in the crude tower, it must be properly
removed to provide efficient separation. The heat energy is
removed through the pumparounds, the overhead condensers and
the removal of product. To minimize fuel consumption, the majority
of this heat is recycled to preheat the feed to the furnace. This
strategy works well, unless the feed flow is stopped. That removes
the primary sink for heat in the CDU. The heat will then accumulate
in the tower and vaporize additional material, thus increasing
system pressure and the amount of energy traveling up the tower.

Often the same feed flow travels equally through the preheat
exchangers and the furnace, but in the case of OMVD, a preflash
tower decouples the flow of crude oil through the feed exchangers
and the furnace. Therefore, it is possible to have flow going
through the furnace and adding heat to the tower, while the heat
removal flow is not available.
The feed flow through the preheat exchangers is manipulated by the
preflash tower level controller. During an upset, the controller could
close this valve and temporarily stop the flow of feed through the
feed exchangers. This behavior happens several minutes after the
initial cause of the upset, due to the lags in the system and the
dynamic response of the controllers.
The new ESD system provided that the feed flow through the
furnace was quickly reduced under certain conditions. This furnace
flow came from the preflash tower, so the reaction of the preflash
tower level controller would cause the feed flow through the
exchangers to temporarily stop completely. The tuning of this
controller is therefore important to the safety of the unit, and this
project assumed that the controller tuning provided large swings in
the level during these cases.

C:\HBJ Admin\Papers\Plant-data Validated Model predicts Crude Column Relief Loads PETROTECH
2003 paper.doc
Page 3 of 15
Under the Auspices of the Kingdom of Bahrain
Petrotech 2003 - The 4th Middle East Refining and Petrochemicals Exhibition and Conference
29th September - 1st October 2003

OMVD’s crude distillation unit has it’s own personality, but it’s a
fairly typical crude unit. There are not many situations that will
cause a dramatic system pressure increase. Increased pressure
results from the vaporization of the light components of the feed or
the external introduction of light hydrocarbons (required for
pressure control). Any vaporization requires the addition of energy
(heat), and efficient separation requires the efficient removal of this
heat to condense the products for removal. Of course,
thermodynamics tells us that an increase in pressure will cause
some vapor components to condense. The dynamic energy
distribution, material inventories, multicomponent thermodynamics,
and control responses must all be accurately calculated in order to
provide useful information about the actual process behavior.

The importance of these relationships requires a detailed model of


the tower internals, sidestrippers, preheat exchangers, and
overhead system. SIMSCI’s rigorous thermodynamic methods has
been the approved standard in this refinery for all steady-state
simulation tasks for years, and DYNSIM uses these same methods
to provide accurate dynamic results, as well.

For simulation speed and other practical reasons, some modeling


assumptions are required. However, all of these assumptions were
made to be as pessimistic as possible to produce the maximum
amount of vaporization and pressure.

The maximum possible relief load estimation needs detailed


understanding of that particular plant behavior and the prediction of
what the worst operation scenarios would be to create the highest
pressure in the overhead section of the tower. OMVD’s vast
experience on this unit and SIMSCI’s experience with dynamic
simulation on other units in similar studies produced a list of five
scenarios to be considered in the study:
1. Total loss of power,
2. Loss of Reflux Pump,
3. Loss of Naphtha Condenser fin fans,
4. Failed Feed Pump,
5. Valve supplying fuel gas for overhead pressure control fails open
(opposite of fail position).

The study employed a rigorous dynamic model of the Crude Tower


and associated equipment, containing 2 Pumparounds, 4 side
strippers and the overhead cooling and reflux system. 3 model relief
valves represent a number of equally sized valves each, that open
at it’s specific set pressure and close at a certain reset pressure.

C:\HBJ Admin\Papers\Plant-data Validated Model predicts Crude Column Relief Loads PETROTECH
2003 paper.doc
Page 4 of 15
Under the Auspices of the Kingdom of Bahrain
Petrotech 2003 - The 4th Middle East Refining and Petrochemicals Exhibition and Conference
29th September - 1st October 2003
The crude oil feed is heated by a number of heat exchangers, that
cool the product or pumparound streams on the other side.
The hot feed gets flashed in a Preflash column and enters the
furnace before passing to the crude tower. A list of major
equipment included for the CDU model is given in Table 1.
Table 1: Modeled Equipment
Crude Tower Naphtha Condenser
Reflux Accumulator Reflux Exchangers
Naphtha Accumulator Reflux Pumps
Crude Heater (Furnace) Naphtha Condenser
Side Strippers Tower Relief Valves (2 Pressure Levels)
Preflash Tower Condenser Relief Valve

Dynsim accurately calculates mass and energy balances and


rigorous thermodynamic calculations for all equipment. The dynamic
model includes the instrument and control logic of over 30
controllers, 50 transmitters, and a variety of logic modules.

MODEL VALIDATION
Model validation is the key for trusting the results and predictions of
any simulation model. This statement is especially true if safety
conclusions are made. No model is perfect at all circumstances, but
for these purposes, it must be certain that any model discrepancies
with actual plant behavior provide the more pessimistic estimate of
the safety of the process. An experienced plant engineer can
compare model results with plant data and find important
discrepancies with the model or any inconsistencies in the plant
data itself.

The first validation test was a comparison of the steady-state plant


data combined with PRO/II simulation results of that case with the
Dynsim ones at steady-state conditions. The correspondence was
excellent, since similar components and identical thermodynamic
methods were used:
 All dynamic model product flow rates matched the flow
measurements
 Minimal pressure discrepancies existed between model and plant
measurements
 The product temperatures as well as the internal column
temperature profile were met within <1% difference – lower than
the measurement standard deviation
The differences lied within the measurements standard deviation, so
that both models just delivered different solutions of the data
reconciliation of the same set of plant data.

The DYNSIM model contains more detailed representations of each


of the individual column trays. It provides far more additional
information about the dynamics of the process than is available
from the instrumentation or laboratory measurements. This
C:\HBJ Admin\Papers\Plant-data Validated Model predicts Crude Column Relief Loads PETROTECH
2003 paper.doc
Page 5 of 15
Under the Auspices of the Kingdom of Bahrain
Petrotech 2003 - The 4th Middle East Refining and Petrochemicals Exhibition and Conference
29th September - 1st October 2003
includes information about tray dumping, flooding or dry out during
process upsets.

Secondly, a dynamic validation test was designed to insure that the


dynamic results from the model were consistent with actual plant
behavior. Ideally, a dynamic model should be validated against
actual plant data that is similar to the actual operating conditions of
interest. For plant relief studies, it is impractical and unwise to
expose the plant to a relief scenario just to get dynamic plant test
data for validation of the dynamic model. Thus, a more practical
and safer plant test was applied. A carefully monitored process
response of a well defined process change of the reflux rate was
applied for detailed validation of the dynamic model behavior. The
dynamic responses of key measurements in the overhead section of
the Crude Tower were recorded. Careful test planning and
operations execution is necessary to keep all other process
conditions under control during the brief test period. The same step
change in the simulation model reflux flow rate by 15% was
performed, and the responses were compared to the plant response
data.

The comparison of step change plant measurements and model


results produced an excellent match of the majority of parameter
trends. Even minor discrepancies needed to be analyzed in detail,
ensuring that any model discrepancy has to be more pessimistic
than the real process behavior. Having slightly more and lighter
material in the models overhead system provides a more
conservative estimate of the relief loading during abnormal
conditions. The model indicates f.e. a quicker increase of
temperature at the Kerosine draw tray than the measurement
shows, missing some thermal capacitance at the model due to
thermal inertia in the metal and transmitter lags in the plant.
Following the rule that any model discrepancy has to be more
pessimistic than the real process behavior, conclusions will stay on
the safe side by producing significantly faster energy addition to the
top of the tower in the model, as indicated by a relatively rapid
temperature increase.

The overall very good match between dynamic plant measurements


and simulation model results indicate that the system volumes,
thermodynamics and controller tuning are very accurate and allow
the conclusion that our study results will be valid.

RELIEF SCENARIOS
Plant experience is required to decide which failures are the most
harmful in terms of pressure. Process understanding and operating
experience are necessary to consider all of the possibilities. Certain

C:\HBJ Admin\Papers\Plant-data Validated Model predicts Crude Column Relief Loads PETROTECH
2003 paper.doc
Page 6 of 15
Under the Auspices of the Kingdom of Bahrain
Petrotech 2003 - The 4th Middle East Refining and Petrochemicals Exhibition and Conference
29th September - 1st October 2003
equipment or groups of equipment might fail and cause sudden
pressure increases in the unit, sometimes at very different locations
in the system. The following scenarios were selected for this
analysis as the expected worst cases, that needs to be analyzed for
any crude unit. These are typical fractionation relief cases with
obvious effects like:
1. Total loss of power - All pumps and fin fans fail simultaneously
2. Loss of Reflux Pump - stops liquid flow to the CDU top tray
3. Loss of Naphtha Condenser fin fans - stops cooling and
condensation
4. Failed Feed Pump - stops cooling in the Reflux and Pumparound
heat exchangers
5. Valve supplying fuel gas for overhead pressure control fails open
(opposite of fail position) - 50 times more fuel gas than normal is
forced to the overhead system.

The flexibility of the model allows even the study of very theoretical
events like CASE 5, where 2 control valves fail simultaneously and
move into the opposite direction of their fail position.

The new ESD logic was included in the dynamic model, as well as all
the necessary regulatory control for normal plant operation. In each
scenario the Furnace and steam shutdown logic is active. The
Furnace will trip when the Crude Tower pressures reach their limit
values. If the total feed or the reflux flows drop below the allowed
low limits or the Crude Tower temperatures reaches an upper limit
value, the Furnace will also trip.

Several conditions incorporated into the model insure a very


conservative estimate of the relief load:
 The crude composition in the test cases is slightly lighter than
the normal feed to the Crude Tower.
 Slower cooling of the Furnace will add more heat to the Crude
Tower in the test scenarios.
 The pressure control valve that opens to the flare system was
left closed for each case, leaving no way for vapor to the crude
unit except through relief valves or by condensing. Additional
calculations with this controller operating estimated more
realistic maximum column pressures.
 The overhead temperature control in Manual won’t increase the
reflux flow.
 No credit for heat loss to ambient is taken in any equipment,
except in fin fans.
 No pump flow is allowed when the power is off or the pump is
failed, regardless of the pressure drop.
 No corrective action of any kind is taken to recover stable unit
operations, except the ESD logic.

C:\HBJ Admin\Papers\Plant-data Validated Model predicts Crude Column Relief Loads PETROTECH
2003 paper.doc
Page 7 of 15
Under the Auspices of the Kingdom of Bahrain
Petrotech 2003 - The 4th Middle East Refining and Petrochemicals Exhibition and Conference
29th September - 1st October 2003

The model results containing every process parameter trend (even


if it is not measured in the plant) is far more, than this paper can
fully describe. A summary of main simulation results at each
scenario with the focus on safety relevant conclusions and
unexpected process interactions is given in the following section.

CASE 1 - Total Loss of Power


Operators know, that in that case all electrical driven equipment
stops working, at a crude unit it’s the pumps and fin-fans. The ESD
logic of the plant stops steam flow to the Crude Tower and to side
strippers after two seconds, but doesn’t stop steam flow to the
Furnace.

The Furnace outlet temperature increases immediately due to the


thermal inertia in the Furnace and the decreased flow when the
topped crude flow is stopped. Over the next 15 minutes the Furnace
outlet temperature drops as the Furnace cools to the steam
temperature.

If the feed pump stops working and no crude oil is pumped to the
preheat exchanger, the overhead flow is not cooled any more.
Sharp rise in the overhead temperature, loss of condensation
overhead and a quick overhead pressure grow up are the immediate
effects. The First Level tower relief valves open the first time about
2 min after the power failure. The important finding of the dynamic
simulation is, that after opening these First Level relief valves once,
relief flow will decline the overhead pressure within 10 seconds
below the relief valves reset pressure until they close. The pressure
will be up again a half minute later, causing the First Level tower
relief valves to quickly open and close within the very short interval
again. During each of that short period, the First Level tower relief
valves release about 2 kg vapor. The valves cycle for about forty
minutes, until the lightest material has been released. The other
relief valves like the second level tower relief valves (1 Bar higher
set pressure than the First Level ones) and the Naphtha condensate
receiver relief valves will not open during a crude unit power loss.

The peak size of possible relief flow (compared to the relief system
capacity) is a key parameter for safety evaluation of a plant. Using
a rigorous dynamic process model, it could be shown to be far
below the valve flow capacity and as well far below the steady-state
method prediction. Relief flow at CASE 1 occurs between the 2nd
and the 30est minute after the failure, until sufficient material is
released to decrease the pressure down into safe areas.

C:\HBJ Admin\Papers\Plant-data Validated Model predicts Crude Column Relief Loads PETROTECH
2003 paper.doc
Page 8 of 15
Under the Auspices of the Kingdom of Bahrain
Petrotech 2003 - The 4th Middle East Refining and Petrochemicals Exhibition and Conference
29th September - 1st October 2003
The simulation provides trends of important process parameters like
steam flow, overhead vapor and overflash flows to get additional
understanding of the particular process response. The model
calculates the amount of overall relief load, relief load time, relief
valve opening period etc. In fact, every calculated process
parameter can be trended, cause-effect relationships become
obvious and quantified. The whole unit behavior becomes
understandable during abnormal situations, that the plant operator
hopefully never will see.

The most appropriate human interactions can be determined as


well. To do so, it might be necessary to include pressure controller
action, as it usually works even after the loss of electrical power.

The validated dynamic model shows, that the column pressure will
increase so rapidly, that the overhead pressure controller action
does not make any difference at the beginning of that failure
response. After a number of short relief valve openings the extra
pipe opened by the pressure controller allows enough vapor relief,
that the relief valves action period becomes shortened to about 10
minutes. The trends of the overhead temperature and the overflash
flow are identical with the power failure CASE1, because they don’t
get affected by pressure controller action. The overhead vapor flow
out of the column shows the instability of the column traffic after
flooding.

As for the case before, a detailed description of the process


behavior for that case is part of the study delivery. Pressure
controller action reduces the load of the relief valves but can not
remove all vapor material quickly due to the controller tuning.
Different PID parameters might change this time frame, and the
model can help to determine what the optimal controller parameters
would be.

CASE 2 - Loss of Reflux Pump


The Reflux Pump failure stops the flow of chilled liquid to the top
tray of the column, upcoming vapor within the column has no
interaction with liquid anymore. The ESD acts when the reflux flow
drops below its minimum limit. The Furnace shuts down, steam
flow to the column and side strippers stops. Feed flow valves drop
to the 10 percent position. Internal traffic in the column drops
dramatically without reflux, upper trays dry out within about five
minutes, temperatures increase significantly and product flows are
lost.

The Overflash liquid flow at the feed tray of the column increases
shortly and becomes 0 flow after 4 min. (WHY?) The Overhead

C:\HBJ Admin\Papers\Plant-data Validated Model predicts Crude Column Relief Loads PETROTECH
2003 paper.doc
Page 9 of 15
Under the Auspices of the Kingdom of Bahrain
Petrotech 2003 - The 4th Middle East Refining and Petrochemicals Exhibition and Conference
29th September - 1st October 2003
pressure decreases for the first 2 minutes into safe areas. It seems
that this case keeps innocuous, before the pressure turns back
sharply and causes the relief valve to open after 3.5 minutes, a
pretty unexpected effect during this scenario. The explanation can
be found by watching more parameter trends in the model: ESD
logic reduces the crude flow to the Furnace, affecting the Level
Controller action on the Preflash Tower. This controller briefly stops
the crude flow to maintain the Preflash Tower level reducing the
cooling source about three minutes after Reflux Pump failure.
Overhead vapor is no longer condensed and pumparounds aren’t
cooled.

As understood in CASE1 already, crude feed flow rate has a


tremendous impact on the column OVHD and product cooling. Every
time, the feed flow drops below a certain rate, the pressure
increases above the relief valve set pressure. First Level tower relief
valves open about three minutes after the Reflux Pump failure and
at this time they remain open for one minute due to Preflash Tower
Level Controller action. A total of 1 t relief material gets released
through the First Level tower relief valves, all other relief valves of
the crude unit stay closed after the loss of reflux pump. Pressures
increase again about eight minutes after the reflux flow failed,
because the Preflash Tower Level Controller almost closes the crude
feed flow again. A reduced response of that level controller could
prevent the relief valve from opening. The simulation shows that
the relief valves will close soon after the level Preflash Tower Level
Controller resumes the feed flow. The frequency of the cause (feed
flow rate) exactly meets the frequency of effects (OVHD vapor flow,
OVHD pressure and relief flow.

This case demonstrates, how even unexpected effects in critical


plant behavior can be found by studying the dynamic model
response, and that the model behavior itself helps to increase the
process understanding and to find explanations for these effects.
Recommendation could be derived for a better setup of the control
regime to not allow the preflash level controller to disturb the heat
transfer regime as extensive.

CASE 3 - Loss of Naphtha Condenser Fans


In this test the fin fans in the naphtha condenser are blocked, but a
small amount (ten percent) of cooling occurs due to natural
convection. Naphtha Accumulator pressure increases sharply and
equalizes with that reflux drum, but then the pressures in the
overhead system and crude tower increase very slowly. After
approximately 38 minutes, the overhead pressure reaches it’s ESD
logic trip point and ESD logic trips the topped crude flow, furnace
and steam flows. The tower pressures drop after the ESD action.

C:\HBJ Admin\Papers\Plant-data Validated Model predicts Crude Column Relief Loads PETROTECH
2003 paper.doc
Page 10 of 15
Under the Auspices of the Kingdom of Bahrain
Petrotech 2003 - The 4th Middle East Refining and Petrochemicals Exhibition and Conference
29th September - 1st October 2003
About two minutes after the topped crude flow is reduced by the
ESD logic, the Preflash Tower Level Controller stops the flow of raw
oil to maintain the level. As in the case before, that level controller
action causes via raw oil flow to stop the heat removal in the
overhead condenser and the reflux exchanger. The following effect
is a sharp rise of pressures in the tower and the First Level tower
relief valves opening. These valves continue to open and close for
about 1.5 minutes, until the raw feed flow is restored by the
Preflash Tower Level Controller. The relief valves open again after
another three minutes when the Preflash Tower Level Controller
again reduces the raw oil feed flow rate. Overall, less than 1 ton of
material was released. The other relief valves did not open during
this test. The cone pressure stays below the allowed maximum.

CASE 4 - Loss of Raw Oil Feed Pump


When the raw oil feed pump is failed, the flow through the overhead
condenser and reflux exchangers stops. The cooling duty of the
overhead condenser immediately drops to zero, and the overhead
vapor to the reflux drum no longer condenses. The overhead
pressure increases sharply and within thirty seconds the ESD
system trips the steam and feed flows at 1.5 barg. First Level tower
relief valves open once very briefly, and the pressures quickly
decrease as the vapor flow up the tower decreases. The pressures
increase again after another two minutes due to the loss of tower
inventory and the increasing temperatures. The First Level tower
relief valves open about 3.5 minutes after the feed pump is lost.
The flow of material overhead is not sufficient to maintain the level
in D-002. About four minutes after the feed pump is failed, the
Reflux Drum loses the level and the reflux flow stops. The internal
traffic is lost and soon the Kerosene draw, upper reflux flow, Gas Oil
1, Gas Oil 3 and overflash flows are lost. About six minutes after
the feed pump is failed, the level in the preflash tower is lost and
the topped crude flow to the crude tower stops completely. The
steam flow is maintained through the furnace, but the overhead
vapor flow rate decreases substantially when the ESD logic is
activated. Approximately ten minutes after the feed pump is lost,
the lightest material has been released and the pressures drop
enough for the overhead pressure controller to add fuel gas to
maintain the pressure.

The First Level tower relief valves open once about thirty seconds
after the feed pump is failed and open again about three minutes
later, cycling for about five minutes. A total of about 3.2 tons is
released during this time, and the cone pressure reaches a
maximum of 1.7 barg. The other relief valves do not open during
this scenario.

CASE 5 - Overhead pressure control valve fails open


C:\HBJ Admin\Papers\Plant-data Validated Model predicts Crude Column Relief Loads PETROTECH
2003 paper.doc
Page 11 of 15
Under the Auspices of the Kingdom of Bahrain
Petrotech 2003 - The 4th Middle East Refining and Petrochemicals Exhibition and Conference
29th September - 1st October 2003
This rather theoretical case was studied to have a look at the overall
relief valves capacity.

To have all relief valves opened, the assumption was made, that the
pressure control valve, that provides fuel gas to maintain the
column top pressure, stays 100% open, what it the opposite of it’s
fail position. Feeding about 50 times more than normal fuel gas to
the column, opens the Naphtha Condensate Receiver relief valves
immediately. These relatively small relief valves take very little of
the added fuel gas amount, not sufficient to avoid further rapid
increase of pressure in the OVHD system. The pressure increase
moves backwards to the top of the tower, until the First Level tower
relief valves open after 0.7 minutes. First Level tower relief valves
opening drops the pressure immediately below the valves reset
pressure, later these relief valves will oscillate as we have seen in
the cases before.

After 3.5 min the First Level tower relief valves stay open longer
term and can’t avoid, that the column pressure rises even further.
After about 10 min after malfunction, the Second level tower relief
valves open frequently. This additional relief flow is sufficient to
avoid further pressure increase in the tower above the second level
set point. After 5 min oscillating, the Second level tower relief
valves stay closed, and the First Level tower relief valves take over
the pressure preservation again.

The peak flow rate with all relief valves opened at the same time is
calculated as 150 t/h. Allowable overpressure for that system isn’t
exceeded. This is the opposite of the result of applied standard
steady-state methods for relief load estimation. Storage capacity of
the whole column and partial condensation of column inventory will
reduce the arising pressure at that abnormal condition as well.

STUDY CONCLUSION
Dynamic simulation can be an effective tool for crude column relief
analysis. OMVD was able to conclude from the study results that
their relief system is still adequate to handle the safety tasks
around the crude unit after the revamp. German authorities (TueV
Sueddeutschland) agreed in that conclusion.

The OMV D Crude Unit is inherently a fairly safe process, as are


most crude units. At ambient temperatures and pressures the raw
oil and all of the products are in the liquid phase and stable. The
single source of heat for the crude unit is the furnace. The heat
added by the furnace is removed in a controlled mechanism that
provides efficient separation of the different products. Stripping
steam added to the process also adds some heat, but this heat is

C:\HBJ Admin\Papers\Plant-data Validated Model predicts Crude Column Relief Loads PETROTECH
2003 paper.doc
Page 12 of 15
Under the Auspices of the Kingdom of Bahrain
Petrotech 2003 - The 4th Middle East Refining and Petrochemicals Exhibition and Conference
29th September - 1st October 2003
small relative to the heat from the furnace and is usually at a lower
temperature than the hydrocarbons. The steam provides a
stripping effect and the high heat capacity of steam carries the heat
and lighter hydrocarbons up the crude tower. The pre-flash tower
upstream of the crude tower removes most of the lightest material
from the raw oil feed. The lightest components in the topped crude
feed to the crude tower pass overhead. The condensers in the
overhead section of the crude tower totally condense all of the
hydrocarbons and water. As in most crude units, fuel gas must be
added, or the vapors compressed, in order to maintain a positive
pressure and not create a vacuum in the overhead receivers when
the vapors are condensed.

The SIMSCI dynamic model proved to match the steady-state and


dynamic behavior of the OMV D crude tower well. The steady-state
values matched extremely well, with only minor adjustments to
account for any energy losses in the actual plant. The dynamic
validation showed that the pressures, levels and flows agree very
well with the plant measurements, but the temperatures in the
model respond somewhat faster than the actual unit. The
composition lags in the equipment and piping, the thermal lags from
the metal heat capacity, and the inefficiency of mixing and mass
transfer in the upper part of the tower may contribute to these
small discrepancies in the dynamic behavior. The model provides a
conservative estimate of relief loading during the tests: the faster
temperature increase, slightly higher overhead temperatures, and
slightly lighter composition result in slightly higher calculated relief
loads during emergency conditions. Conclusions out of that
calculation are always on the more conservative and safer side.

The test scenarios illustrated that the crude tower is fairly safe, and
the instances when the relief valves opened were generally brief,
easily explained, and attributed to very few causes.

The main reason that the relief valves at the top of the crude tower
opened, was the lack of raw oil flow through the reflux heat
exchangers. This flow provides cooling in the overhead condenser
and the reflux exchanger. Therefore, the loss of the feed flow stops
the vast majority of the heat removal in the crude tower. The
system response to that loss of cooling is the pressure increase to
open the First Level tower relief valves only. These relief valves
alone are able to reduce the overhead pressure enough below the
relief valves reset pressure with a couple of short openings.
Also, the pressure at the crude tower reduction zone (cone) is
maintained within the safety limits (1.7 barg). A maximum total of
4.7 tons was released through the relief valves to the flare. Active
pressure controller reduces the maximum total of 2.9 tons.

C:\HBJ Admin\Papers\Plant-data Validated Model predicts Crude Column Relief Loads PETROTECH
2003 paper.doc
Page 13 of 15
Under the Auspices of the Kingdom of Bahrain
Petrotech 2003 - The 4th Middle East Refining and Petrochemicals Exhibition and Conference
29th September - 1st October 2003

The exception case is where the pressure control valves both fail or
are placed in manual in the opposite directions of their normal fail
positions. This dual valve failure case is not very realistic and a
more severe case than will likely exist in the actual plant, but it
gives a worst possible case scenario for comparison. The pressure
at the top of the OMVD crude tower only reached the Second level
set pressure when both pressure control valves are forced in the
opposite direction of their failsafe positions. The relief valve on the
naphtha accumulator opened only when the overhead pressure
control valve adding fuel gas was failed open, where a two inches
diameter valve adds the fuel gas, while a four inches valve vents
the vapors to the flare.

During all of the test cases, the flow through any one of the relief
valves didn’t exceed 38 t/h. The scenario failing both pressure
control valves produced the highest cone pressure (1.72 barg) and
the largest total relief flow (36 tons over 30 minutes).

The results of the examination of the crude unit behavior using a


dynamic model provide significant insight to the operation and
transient behavior of the process as well as what changes can be
made to increase process safety. Steady-state models can not
predict the actual reactions of the crude unit to abnormal
conditions. The conditions that actually cause the relief valves to
open are often indirect to the actual equipment failure, and the
relief valves may open several minutes or more after the initial
failure. As in most cases, the dynamic model predicts a smaller and
much more realistic load on the crude tower relief system than the
steady-state model estimates. The DYNSIM™ model includes the
process lags, transient thermodynamic effects and the influence of
the control system on the process dynamic behavior for an accurate
prediction of the relief load during abnormal conditions.

OMVD was able to conclude from the study results that their relief
system is still adequate to handle the safety tasks around the crude
unit after the revamp. Increasing the size or number of the relief
valves or expanding the flare header will provide no significant
benefits to the safety of the crude unit. However, certain
adjustments in the shutdown logic or process configuration in order
to maintain the feed flow through the preheat exchangers will
provide a significant reduction in the amount of vapor released
during abnormal or equipment failure conditions.

The study reaffirms the refiner’s experience that a crude unit


pressure increase causes additional condensation of column
inventory and reduces the pressure increase and the amount of
relief loads. Its quantification with an accurate, fine-tuned model
C:\HBJ Admin\Papers\Plant-data Validated Model predicts Crude Column Relief Loads PETROTECH
2003 paper.doc
Page 14 of 15
Under the Auspices of the Kingdom of Bahrain
Petrotech 2003 - The 4th Middle East Refining and Petrochemicals Exhibition and Conference
29th September - 1st October 2003
can prevent from unnecessary investments in relief system
upgrades.

In this example, savings from unnecessary upgrade easily justify


the cost of a very detailed rigorous dynamic process model.
Additional usage of that model includes multiple investigations of
changes in the design, in the control and in the operation of that
unit then the study showed. OMVD’s dynamic CDU model could be
taken as the heart of a training simulator that allows improvement
of the operator’s experience in abnormal situation handling.

REFERENCES

1) Heins, L., Jemison, H., and Dietrich, R., “Crude Unit Relief
Loading Study Using a Rigorous Dynamic Simulation,” ERTC
2000 Computing Conference, June 2000.

2) Depew, C. and Dessing, J., “Dynamic Simulation Improves


Column Relief-Load Estimates,” Hydrocarbon Processing,
December 1999.

3) Ernest, J. B., and C. A. Depew, “Use Dynamic Simulation to


Model HPU Reactor Depressuring”, Hydrocarbon Processing,
January 1995.

4) Cassata, J. R., et al., “Modeling of Tower Relief Dynamics – Part


1”, Hydrocarbon Processing, October 1993.

5) Cassata, J. R., et al., “Modeling of Tower Relief Dynamics – Part


2”, Hydrocarbon Processing, November 1993.

C:\HBJ Admin\Papers\Plant-data Validated Model predicts Crude Column Relief Loads PETROTECH
2003 paper.doc
Page 15 of 15

View publication stats

Das könnte Ihnen auch gefallen