Академический Документы
Профессиональный Документы
Культура Документы
SPE
A Gas-Lift Optimization and Allocation Model for
Manifolded Subsea Wells
by R.Edwards and D.L. Marshall, Amerada Hess Ltd., and K.C. Wade,
Baker-Jardine & Assocs. Ltd.
ThIa~-IftIlINdIll"- .~IIO.lhe~,""""""22·24a-1IIIIO.
ThIa~ _ _ tor _ _ b\''' SPE ~CcImiIWetlllawing_oI_""-1n .. _1UbmiIIId b\'1IIe8Ulhor(.~ ConwMoIthe_
.. ~ _nat --. _ b\'1IIe SociIIyol PwIIoIIum ens;r- _ _ IUbjecIIII- b\'the8Ulhor(.~ lhe-'''~ _nat~reIIKl
iliff poeitian 01 the SociIIyof PeIIIum ~ ilaoltlc8ra or _ _ '"--__'.SPE -.ga _1UbjecI1II JIIIIIiI*ion _ b\' EdIcriII ~ofllle
SociIIyolPwllollum~ I'8rmiaaIanIllCC!ll'is_III .. _oInat_tIwI300_I_"..natbeccp;.d. lhe _ _ Clll'llllin~
_ _"dg.iI_dol _ _ b\'wllamllle~is _ _ _ \MtIa~~ SPE. p.o. BalB33838. ~ TX 75083-3836. r.. 73OII88SPElW..
ABSTRACf
The increasing need to develop marginal performance module of the optimisation model. This
accumulations using subsea equipment and common feature enables the correct allocation of total production
facilities has highlighted the challenge of the allocation to individual wells between well tests. The model will be
and optimum distribution of fluids within these systems. fully employed in the day-to-day management of the field
The commingling of produced fluids requires that an when gas-lift is required to maintain production rates.
auditible procedure is employed to accurately back- In its predictive mode the model can be used in
allocate fluids produced from different fiscal regimes. combination with reservoir simulation data to identify the
Similarly, there is a requirement to optimise lift gas incremental improvements that may be achieved through
allocation to ensure that the available gas provides the gas lift later in the field life and the impact of gas
maximum incremental oil production. A method of deficiency.
optimising lift gas allocation and allocating production
within a manifolded subsea development has been PROJECf BACKGROUND
developed for use on a portable computer. The basic
requirements for the model were that it should be Reservoirs
usable by offshore personnel and allow a quick The Ivanhoe and Rob Roy fields are located in
comparison of different operating scenarios. the UK Sector Block 15/21a, approximately 110 miles
north east of Aberdeen. As development of the Ivanhoe
INTRODUCTION and Rob Roy fields progressed, exploration in the 15/21
Prior to start up of the Ivanhoe and Rob Roy area continued, resulting in the discovery of the Hamish
fields, one of the development scenarios considered field with well 15/21b-21. Hamish was appraised,
early water breakthrough with an immediate need for developed and on production two years after its
gas lift as an artificial lift mechanism. The variable discovery, using the existing IvanhoelRob Roy
nature of the two reservoir units within each field infrastructure (Figures 1 and 2).
dictates that some method of optimising the distribution The Ivanhoe and Rob Roy and fields produce
of lift gas between wells will be necessary to ensure that from two Jurassic intervals, the Main and Supra Piper.
the available gas-lift gas is used to best effect. The Hamish field produces only from the Main Piper.
Software has been developed to allocate the lift The Main Piper sands are thicker and generally more
gas in order to maximise productivity by balancing productive than the Supra Piper. There are thiee
incremental oil production per unit of injected gas for distinct crude types within the fields (Table 1).
each well. The gas lift optimisation model can then be
operated as a decision support system whereby the System Architecture
current and optimum production rates can be compared The field architecture is shown in Figure 3
to recommend operator action according to a set of depicting the commingling of the fluids at the Ivanhoe
decision criteria. The production asllocation model can and Rob Roy production manifolds through the riser
be used to monitor production by utilising the wellhead base manifold (RBM) and then onto the floating
pressure performance curves created by the system
Page 535
2 GAS UFT OPTIMISATION AND ALLOCATION FOR MANIFOLDED SUBSEA WELlS SPE 21)979
production facility (FPF) for processing. The FPF was These curves are then sent offshore (on floppy disk) to
built using a converted semi-submersible construction be installed in the data base of the gas lift optimisation
vessel. A total of nine production and six injection and allocation model.
wells have been drilled and completed sub-sea during After a well has been placed on test, the
development of the fields. Of these, half were resulting performance data is used by the engineer
completed prior to first oil. The wells were not drilled onshore to check and recalibrate the well model, and if
through a conventional template but instead were required, produce a new set of system performance
positioned in clusters around the two production curves to be sent to the pla.tform.
manifolds. While the user on the platform is primarily
The two well clusters are located two miles interested in the current state of the field, the onshore
apart on the seabed in approximately 460 feet of water. engineer will be concerned with future well performance.
Each subsea production well has a remotely operated Models of possible future scenarios can be built to
subsea choke controlling production. Similarly, the gas include new wells, or different reservoir and production
lift line to each production well has a remotely operated performance characteristics. The same optimisation and
choke system which controls the volume of gas delivered allocation models used on the platform can then be used
for gas lift. This gas is metered for each well using to examine the system in detail and give, for example,
orifice plate meters located on the subsea manifold estimates of future gas lift requirements. These future
upstream of the choke. The ability to remotely control models can be stored in the data base, onshore or
and measure the gas volume delivered to each well offshore, and compared with operational data as the
makes a gas lift optimisation programme feasible. performance of the field and the wells changes. The
Accurate production allocation however, is storage of the data will provide a history map of the field
complicated by the manifolding of the production wells. and allow its performance to be monitored.
Each manifold incorporates hydraulically operated
valves and chokes which control the commingling and Data base interrogation
distribution of oil production. The manifold system is The interrogation section can be run in three distinct
remotely controlled via retrievable control modules. mode of operation, namely;
Connection to individual wells is by flexible flowlines
and electro-hydraulic control lines. ... Gas lift optimisation,
Both production manifolds produce to the • Production allocation and
RBM located midway between the two fields. The • Well monitoring.
RBM is remotely controlled via a retrievable control
module and also acts as a distribution point for injection Once the new data had been incorporated into
water and lift gas to the production manifolds. A total the data base the platform user would follow a
of eight flexible risers connect the RBM to the FPF procedure for assessing the possibility of resetting the
(Reference 1). gas lift chokes to optimise production from the system.
This procedure consists of;
COMPUTER MODEL
1: Using the allocation model to predict the
System design current flowing rates for each well.
The gas lift and allocation model has been
designed for use in two modes of operation, namely; 2: Using the optimisation module to predict the
optimal flowing conditions, given any current
... Onshore by an experienced engineer and field constraints.
... Offshore, as a day-to-day prediction tool.
3: Assessing the overall gain (or loss) of moving
The model is designed to be used in these two the current choke settings controlling the
modes of operation by users with very different amount of lift gas to each well.
computing skills. The analytical part of developing the
well models has been kept separate from the day-to-day The models have been developed using a menu
modelling of the system. Thus the overall model based user interface. The results are displayed
consists of two distinct sections; a data base gathering graphically in a form that corresponds to the layout of
section and an interrogation section. the field infra-structure. The user can gain immediate
access to the data base and all calculated results by
Data base gathering selecting individual system components directly from the
The onshore engineer uses a multi-phase flow graphical screen presentation using the mouse. Any of
simulator to generate a set of system performance the performance curves generated can be immediately
curves for each well and flowline system in the field. recovered and displayed on the screen for inspection. A
Page 536
SPE 20979 D L MARSHALL, R J E EDWARDS and K C WADE 3
tabular summary is also available through the same than the unconstrained optimum, as defined above, then
interface. the problem reduces to allocating the available lift gas
For application in a predictive "real-time" mode most efficiently between wells.
offshore, the algorithm within the model must execute The definition of the constrained optimum is
quickly. Once the data base has been established the that the gradients of all producing wells are equal, ie. all
system developed can optimise all the wells in the wells would increase their flowrate by the same quantity
Ivanhoe, Rob Roy and Hamish fields in about eight if an extra incremental amount of gas was injected. Thus;
seconds (real time) on an 20 MHz, mM compatible, 386
PC. d!,.(X1) _ dJ,.(X;> _ ... _ df.,(x,,) _ G (3)
dx dx dx
Gas lift optimisation model
The gas lift optimisation model is designed to where G (bbl/d per mmscf injected) is the gradient of all
allocate lift gas resources amongst wells in order to the wells (m) in the system. This optimum can be
maximise total field oil production while taking account obtained for all wells providing that each well, j;
of operational and reservoir constraints. The model also
has the capability to allow "what-if! scenarios by closing 1: Can flow at the specified gas lift value, X, and
in any number of wells.
A multi-phase fluid flow simulator was used to 2: Is not bound by any other overriding constraints.
generate a system performanCe curve, artificial lift
quantity versus liquid flow rate, for each well. The There are many numerical techniques available
system performance curves for eight wells, operating for solving linear system of the form;
under current conditions, are shown in Ftgure 4.
Examination of these data points showed that although Max: !t(x1) + h(X;> + •.. + I.(x,,)
the gross liquid is a function of many variables, the Xl + Xz + ••• + XIII S X (4)
underlying trend is dependent upon the total quantity of x,O!:O
gas injected. The optimisation model utilises this set of
system performance curves as its data base. these range from sophisticated optimisation methods to
From this data base the performance curve for simple iteration procedures. However, the introduction
a well, j, is selected and a mathematical function of the of more complicated non-linear constraints ego
form; maximum flow down any given flow line, etc, makes the
setting up of more constraints mathematically complex.
''(x) - L" bj,l x I (1) Thus an iterative technique is employed to provide
i-l greater flexibility and speed of solution.
Page 537
4 GAS UFf OPTIMISATION AND AlLOCATION FOR MANIFOLDED SUBSEA WEllS SPE 20979
where the ~/s are coefficients determined by the ll: A possible future case, consisting of higher
method of least squares, q is the total liquid flow rate water cuts, lower productivity indices and the
and m is the number of wells. The fitting of a function same reservoir pressures. (It is intended to
to the base data enables any minor irregularities to be maintain reservoir pressures at their current
"smoothed out" and a "well-behaved" function obtained. level).
In this case a 2nd order curve was deemed most
appropriate. Thus well head pressure becomes a For both scenarios wells RH16 and RE31 are producing
function of flow rate by; down one of the Rob Roy flowlines and HC40, RM25
and RL24 down the other. All Ivanhoe wells, IE32, m27
(6) and ID19, are producing down the Ivanhoe production
line (Figures 2 and 3).
for each well.
The algorithm devised requires the use of the Gas lift allocation and optimisation
rate of change of flow rate with respect to pressure F'tgure 4 shows typical system performance
(gradient). By fitting a smooth function through the curves for the eight currently flowing wells at operational
points this derivative is now readily available as the conditions, (scenario I). It is clear from these that
inverse of certain wells, ego RL24, would produce less oil if gas
was injected, as the curves of production rate decreases
dp. with increased gas lift quantity. Other wells, ego ID19,
_J _ a + 2a'j"q (7)
dq 'j.1 ... would clearly benefit, although only marginally, from gas
lift.
Page 538
SPE 20979 D L MARSHALL, R J E EDWARDS and K C WADE 5
Similar curves are produced, Figure 5, for basis, ie. multiply the production from each well by the
scenario II. Under these conditions most wells would field factor. Thus, even though the model under predicts
benefit from lift gas being injected. Indeed, some wells, the production by between two and five percent, the fmal
HC4O, RL24, etc will not flow unless gas is injected. allocation will match.
This is indicated by no liquid values for a gas lift
quantity of zero. CONCLUSIONS
The optimisation module of the model was used The above results show clearly that the
to optimise the oil production rate from the fields under allocation models developed for use on the Ivanhoe, Rob
the two scenarios. It was assumed that the optimisation Roy and Hamish fields can effectively predict the current
would give Ivanhoe, Rob Roy and Hamish equal priority performance of the wells given certain operational data
in the allocation of the 11 mmscfd available (priorities available from well tests. Even if crucial data is
can be given to either reservoir). Tables 2 and 3 unavailable, as will often happen in the day-to-day
summarise these optimised rates for both scenarios. running of the field, the model can still be used to
The first column in each table shows the production rate predict performance.
if no gas is injected into the wells and the final column The gas lift optimisation module will be used
the gain in bbVd of injecting the specified amount of lift when the wells begin to cut significant volumes of water.
gas. The scenarios dicussed, indicate that within a subsea
The model predicts that for scenario I, [Table manifolded system, the optimisation of field production
2], there would be a gain of 6,974 bbVd over current will only be achieved by simulation of the entire gas lift
production rates (approx. 8% rise) if a total of 11 system.
mmscfd was injected in the amounts given.
If the situation modelled in scenario II is ever NOMENCLATURE
reached then production will fall to 25,140 bbVd, [Table G Incremental change
3], if no gas was injected. Clearly, under these Any particular well in the
conditions gas must be injected to maintain a reasonable system
production rate. By injecting the available 11 mmscfd, m Number of wells in the system
the production can be increased to 54,446 bbVd (approx. q Flow rate
116% rise). ~(q) General function for well j
PI Pressure for well j
Production Allocation PI Well head pressure for well j
The aim of the allocation model is to predict PI.test Pressure from well test, wellj
the individual well rates given the wellhead pressures. 01 Required flow rate
Table 4 summaries the main quantities used and <I),test Flow rate for well test, well j
calculated in the model. The flowrate and the wellhead "t Gas lift quantity for well j
pressure, in the first two columns, are taken from the X Total gas lift quantity
available test data. The rate of change of flow with a,b Polynomial coefficients
pressure (dQ/dp) is shown as an indication of the n Order of polynomial
stability of the well. A large dO/dp value indicates that
a small change in pressure will result in a large change ACKNOWLEDGEMENT
in flow rate. The next two columns show the actual
wellhead pressure and the predicted flow, from equation The authors wish to thank Amerada Hess
(8). Finally the difference (%) between the test flow Limited, Deminex UK Oil and Gas Limited, Kerr-
rate and the predicted are shown along with the overall McGee Oil (UK) pIc and Pict Petroleum pIc for their
contribution to the total measured production. permission to publish this paper.
The wellhead pressure for well RM25 was
unavailable and was estimated using the multiphase flow REFERENCES
simulator. For wells without a well head pressure. value,
the test line values are automatically used. 1. S Jewell, P Collins, D L Marshall "The
A commonly used statistic to evaluate the IvanhoelRob Roy Fields: Operational Innovation
performance of a mathematical model is to calculate the on a Major Subsea Development" presented at
field factor. This is defmed as the ratio of the field's the European Petroleum Conference, The
production rate to the mathematical models prediction Hague, October 21 - 24 1990.
of the production rate. Using this criteria the model
gives a field factor of 1.0238 for 14/05/90, the field
factors for neighbouring dates are shown in Table 5.
The calculated production rates are then used
to determine the actual production rates on a pro rata
Page 539
SPE 20979
SCENARIO 2
20000,-------,------....,.....-------,-------,--------,
•
GAS LIFT QUANTITY Immscfdl
•
LEGEND
1. <So ~ RE 31 2. @ ® R M 2 5 3 . ~ ·~HC 40 4. Ar£---.t,,j, IE 32
5 ......- - _ . IB 27 6.""$----~$ ID 19 7.~$----~$ RL 24 8. $ $ RH 16
FIGURE 5
FUTURE WELL PERFORMANCE
1750
~
.:
. .
: :
- : :
- .
1Il00 ••• > ••••• ~ •••••••••• , •••••••••• 0;
<
(ij
e:
w
a:
;:)
en
en
w
a:
C-
O
<
W
X
....I
....I
W 500
~
o 5 w " ~ H so
WELL FLOW RATE IMSTB/DI
LEGEND
1. <So E> RE 31 2.0 ORM 25 3.~ ~HC 40 4. bA---'!>,i, IE 32
5.--~ IB 27 6.""$----~$ ID 19 7. $ $ RL 24 8.-f1:$___~-ffi-$ RH 16
FIGURE 6
WELL PERFORMANCE CURVES
Page 540
SPE 2097 9
c- otl PRODUCTION
IVANHOE
PRODUCTION
......FOI.D
r PROQUCTtON
(PlUS CHEMICAL ..." r PRODUCTION
UrWATEAINJECTION S-OASLIFT
'·OL
==~rnE~~==
PROOUCTlON
r TUT ".. GAS UFT
.·TEST TGASLIF'T
S-GASLtF'T L..,crWATERKJECTJON
1+--::",....- •• WATER
INJECTION
rOL
flOATING PRODUCTION
PRODUCTION
FACILlTY "..GASLon
WEUUYOUf
..DlCAnvE ONLY
FIGURE 3
INFIELD FLOWLINES (FPF TO IVANHOE AND ROBROy)
SCENARIO 1
IOODOr---------------~--------------~----------------~--------------~--------------_,
Q
......
lID
I-
~ 20000
., ............................. ; ............................... ; ............................... ; ..............................................................
Q
5
o
::::;
'5000
4
GAS LIFT QUANTITY ImmecfdJ
• 10
LEGEND
1.~--......€> RE 31 2. <!I ®RM 25 3.6 !!lHC 40 4. A. .t, IE 32
5. _.- - - 18 27 $ ID 19
6.$ 7. $ $ RL 24 8.$ $ RH 16
FIGURE 4
CURRENT WELL PERFORMANCE
Page 541
SPE 20979 15/16
Scapa
- 011 Pipeline
••••• Gas Pipeline
FIGURE 1
IVANHOE AND ROB ROY FIELD LOCAnON
"
•
PRODUCER
INJECTOR
)
E'il3 MAIN PIPER
c:J SUPRA PIPER
FIGURE 2
IVANHOE AND ROB ROY RESERVOIRS
Page 542
SP.E 2097 9
TABLE 5
FIELD FACTORS
Page 543
SEE 20979
TABLE 3
RESUL TS FROM OPTIMISATION MODEL (SCENARIO 2)
TABLE 4
ALLOCATION COMPARISON
Page 544
FORMA TION - UPPER JURASSIC SANDSTONE INITIAL RESERVOIR PRESSURE AT DATUM - 3510 pais
RESERVOIR DATUM - 7700ft TVDSS INITIAL RESERVOIR TEMPERATURE AT DATUM - 175 F
MAIN PIPER SUPRA PIPER MAIN PIPER SUPRA PIPER MAIN PIPER
PorosIty 1%1 23 24 23 23 21
Bubble Point Pressure (pslae 175°F) 1.628 1.628 1.942 3,419 1.942
TABLE 1
RESERVOIR FORMATION AND CRUDE PROPERTIES
TABLE 2
RESUL TS FROM OPTIMISATION MODEL (SCENARIO 1)
Page 545