Escolar Documentos
Profissional Documentos
Cultura Documentos
SPE 29782
mbmwmnla2i9dti Pmnmmimbym
sPEPmgmmcamllMu
--~~wnhhd blmSlmImct
S4wMtodtyttwauthq
o).cllnl,md*~,
-~. titim~wti md~wwmtimw m~hbtims). m~, u~, ad~m~
POdt&mdt&O&lhc’c’cd
PObvlwm
E_, b_, w~. b~a=Qmameh h@~*mhdti~
malwultoan~l)tnalnmwmlme. shb8tbNnnynmb0w. nN~tiammin~~
01 whom md by whom =kma. WI’S, lJbIWUI, SPE, P.O. Son SSSSSS, ~, TX 7S0GSWS, U.SA., Tolmt, 1SS24S SPEUT.
121
2 Development of a Network and Gas LiftAflosation Model for Production Optimisation
the deepest possible point of injection with a given gas pipeline to the production platform RB-PP
gas supply. In normal operation, the depth of gas and then distributed to the 3 wellhead platforms
injection rarely changes, assuming a reasonably and from the wellhead platforms to the wells
constant gas supply pressure and reservoir pressure. through 2 “ flowlines.
The performance of the well (and indeed all other
wells connected to the same system) is primarily The production system consists of 3 well
governed by the rate of gas injection, Unless an head platforms delivering production to a
effectively unlimited source of high pressure gas is central process platform via 12” pipelines. The
available, it is usual to design the-gas-lift system for produced fluid is subjected to a first stage
a field to achieve the economically optimum pro- gasloil separation before being pumped onsho-
duction rate, rather than the maximum possible, i.e. re. Produced oil and gas is then piped onshore
such that any incremental production would not pay via 12” and 20” pipelines respectively. Onshore,
for the cost of compressing the extra gas to lift it. the oil undergoes a second stage of separation
This leads to the commonly encountered situation followed by stabilization, dehydration and
in which the efficient operation of die field requires destilting before being routed to storage prior to
the limited amount of gas available to be optimally export by tanker. The separated gas is dried
allocated between the producing wells. and compressed at the compression plant. See
Figure (1) for the simplified production rduting.
Given the usual uncertainties regarding All 15 wells are gas lifted using 4.5’ tubing with
future well performances and future trends in limited potential for natural flow. The high
production ( e.g. reservoir pressure and water cut), pressure gas lift supply is limited by the com-
sensitivity analysis to such uncertaintities of any pressors capacity and field optimization is
production plan become key issues. Therefore, achieved by the proper gas lift allocation be-
there is a requirement for a methodology which tween wells.
treats the overall production system as an inte -
grated system2. All Ras Budran wells suffer from down-
hole scaling problems either of a carbonate or
RAS BUDRAN FIELD FACILITIES AND NET= sulphate type, the degree of scaling and its
WQRK -MQDEL ---
l) FQf7RIPTIflN
b----- . .. .. e------
f feet nn
.-
w*11 nerf nrmnnm=
---- s--- -.. -—-.
mni+-llinu
-“w-—-e
hsc he-n
-_ “---
taken into account and specific observations
The Ras Budran field lies on the east side of made during the work highlighting well and
the Gulf of Suez 5 km offshore. The reservoir is flowline problems.
contained within a heavily faulted compact struc-
ture of sandstone and shale with its crest at 10000 NETWORK MODEL DEVELOPMENT
feet and the oil- water contact at 12350 feet subsea.
The initial reservoir pressure was 5632 psia and the The development of the network model
reservoir is heavily under-saturated with a bubble for an integrated production system is achieved
point pressure of 1198 psia. The current reservoir by describing accurately each component within
pressure varies between 3200 and 3500 psia for the production system to simulate and match
different layer$ and is 4500 psia for the lowest measured data across the field under study. Thr-
layer. ough the selection and application of proper
correlations for each component in the produc-
The field commenced production in Jan. tion network, the validity and hence all subse-
1983 and gas lift was implemented in 1985. Three quent optimization work will be greatly im -
gas lift compressors were commissioned providing proved. To optimize the system effectively, each
approximately 27 mmscf /d of gas at 82 Bar (1189 component must be evaluated separately and
Psig). A fourth gas lift compressor was added in then in combination with other components to
May 1992 adding an additional 9 mmscfid of gas. evaluate the entire well production system.
The gas from these units is transmitted via an 8“
122
EL- Maaary,Y.and Priec, A. 3
The network model is constructed through A calibration factor (~) was selected for
accurately modeling the following: each PVT property to reduce the difference be-
tween calculated and measured PVT properties
1) Reservoir fluid properties. to a minimum value.
2) Well performance.
Measured Propertyva
3) Pipeline performance. Kc= ............ (4)
Calculated Propertywa
RESERVOIR FLUID PROPERTIES
A calibrated Lasater’s correlation was
Reservoir Fluid properties play an important then used to compute PVT’ properties at any
role in calculating pressure drop in wellbores, flowl - pressure and temperature in the current study.
ines and transmitting pipelines in single and mu- Table (1) shows the final calculated PVT prop-
ltiphase flow. It is necessary to obtain numerical erties using Lasater’s correlation.
values for fluid properties such as oil formation VOI-
ume factor J30,gas solubilit y Ra, Z-factor, as well as WELL PERFORMANCE AND HISTORY
POand Pg, oil and gas viscosities. These parameters MATCHING
may be obtained from laboratory PVT analysis, but
often it is necessary to estimate them, especially at The well profile of each of the 17 wells
temperatures different from reservoir temperature. was divided into a number of nodes. Each node
During the model setup, the PVT properties were represents a change in tubing inner diameter
calculated based on a previous simulation study of (I.D.), deviation angle, the presence of gas lift
the Ras Budran field carried out in August 19873. mandrels or a change in flowing temperature.
The different PVT properties at reservoir tempera- The total depth for all wells was taken as the
ture were compared with the PVT data calculated middle of the perforations in each well, which
from four different PVT correlations namely: was assumed to represent the well flow entry
Standing,4 Lasater,5 Glaso,6 and Vaaquez-Beggs7 point. The reference point for all depth and
correlations. Laaater’s correlation was found to be distance calculations was the mean sea level.
the best correlation for the Total Absolute Error
percentage (TAE 9.) of the calculated property Pressure survey data were collected for
deviation from the actual value calculated at differ- all wells. This data includes two consecutive
ent pressures as shown in Table (1) , where : surveys for 10 wells (A4, AS, A6, A7, A8, B2,
B6, B7, B8, B9), 3 surveys for one well (RB-C4)
............... (1) and one survey for 5 wells (B1,C2,B5,B3 and
TAE%
B4) during the year 1993. Of the wells which
were matched using single survey data,. 3 of
where : n = number of the measured property. them (B1,C2 and B5) were surveyed only once
E . = error in measured PVT property at point i. in the last two years, therefore old survey data
Fo; example for & property : matches would not be valid.
123
.
performance model was the reservoir pressure and The failure of correlation(s) in predicting
the bottom hole pressure was calculated using a pressure losses at any station could be attribut-
straight line productivity index (P.I.)8. ed not only to the correlation performance but
also to error in measured pressures and/or
The multiphase flow simulator used con - missing parameters in describing well perfor-
tained a number of different multiphese flow mance. Thus, evaluating each station separately
correlations for calculating pressu~~ ~osses, holdup will eliminate the station calculated error de-
and flow- regimes in verticai fiow-’-. These correc- pendance on each other and wiii highfight the
tions are summarized in Table (2). Only 7 correla- shortcomings in measured pressures and well
tions were used for the performance history match - performance data.
ing part of the study.
Another” type of error measurement
Having specified the correlations to be used method will enhance evaluating pressure loss
in the well performance history matching, the calculation for each station separately and will
pressure losses from the mid point of perforation increase the confidence in the selected correla-
up to the well-head were calculated. Various tion. This error is assigned “delta pressure
correlations were evaluated by applying two error percentage error’ or “station error 90”.
measures, namely :
In addition to the above, two error mea-
{D -*V \
t
‘~”’ Xlcto-loo ....... (s) sures for evaluating the best correlation for
~Emr% = ;i;-Pw-) simulating well performance, the final governing
selection parameter is that the closer the given
(AP--AP~)
De&aReas.ihw% (SaxionEmu%)= (6) correlation is in predicting the well-head pres-
‘p+ sure the more preference is given to that corre -
lation whether it is the best or the 2nd best
where: correlation, provided that the cumulative error
P~ = bottom hole flowing pressure. is always kept less than 5 %. The accurate
Pwi = bottom hole pressure at station i. calculation of well-head pressures using the well
A P-= pressure losses between two consecutive performance model , will provide the inlet
pressure measurement stations. pressure for the pipeline, and will therefore
enhance pipeline manifold pressure calculations
The cumulative error is the difference assisting network model convergence. Fig. (2)
between the calculated and the measured pressures shows examples of some of the well perfor-
with respect to a reference point (e.g. middle perfo- mance history matches.
ration), used to evaluate the different correlations
in calculating pressure losses between the reference It was concluded that 11 wells were best
point and the solution node (well- head or produc- modelled using Orklsaewskt’s correlation, 4
tion manifold). wells were best modelled using Ifegedorn and
Brown’s correlation, and only ene well was best
In this type of error calculation the calculat- modelled using Mukherjee and Brill’a correla-
ed error at the first station or any other following tion.
station (depth station), if it is high, will shift the
error calculation in the following stations towards Choosing a single multiphase vertical
higher error values es they are all referenced to the flow correlation for Ras Budran was not possi-
same point and the calculated error in any station ble (as dictated by two consecutive measured
will include the previous station error. This type of bottom hole pressure surveys for each well) be-
error calculation was considered sufficient for cause of the requirement to change the wells
general selection of multiphase flow correlations. mechanical and/or flowing parameters to re-
duce the % error between measured and calcu-
124
13L-Ma8cry,Y.and Prica, A. 5
lated pressures. The approach used in the model tween “B” and bPP-B” is taking place within the
was to eliminate these changes and let the change ‘B” platform itself.
in the correlation be the modifying parameter as
this was models the well performance better in the This may suggest a deposition (scale,
predictive mode. aapha!tene;etc i,.,) Q~ u ~rga Qf ~~~~r~$~~~g [@~-
tial closure of a control and/or isrdation valve).
125
6 Development of a Network and Gas LtitAffocation Model for Production Optimization
Network Model Verification (3) For this period of interest we have surface
network measured pressures from 09/09/93 to
In order to verify the network model, all the 15/12/93. Minor changes in flowing parameters
well input data files, including the well profile, on piatform and/or a well basis have occurred,
reservoir data and gas lift quantities, have been (+/- 1,500 STBLPD variance in the total field
updated along with weii test information to repre - production). Therefore, matchtng the network
sent actual field data during the verification period. performance at different dates will not be of
much use.
126
EL-Massry,Y. and Price, A. 7
the overall model performance. artificial lift quhntity versus liquid flow rate, for
each well using the pre - defined multiphase
The comparison between the calculated and vertical flow, correlation. An example is plotted
the measured oil production rates (shown in Table in Figure (5). Examination of system perfor-
4) indicates that the network model is over-predict- mance curve data points showed that although
ing meaimred net oil production by 1980 BOPD the gross liquid is a function of many variables,
(cumulative error % = 4.4) during the validation the underlying trend is dependant upon the total
period (i.e. Dec. 1993). The major gross rate dif - quantity of gas injected. The gas lift allocation
ference between actual and calculated flow rates in model utilizes this set of system performance
the system come from wells B5, B8 and C4. Howev- curves as its data base.
er, injecting the gas shallower in the last two wells
utilizing gas lift calculations improves the situation, From this data base the performance
suggesting that scale deposition had been taken curve for a well, j, is selected and a mathe-
place resulting in a loss in the well P.I. matical function of the form16;
CONSTRUCTING THE NETWORK AND GAS The next stage of optimization is to find
LIFT ALLOCATION MODEL AND MODEL the unconstrained optimum assuming an ynlim -
VERIFICATION ited supply of lift gas. The optimization can be
performed on either the gross or on stock-tank
Historical gas lift optimization in Ras Budran barrels of oil. This is defined as the sum of the
individual wells producing at their local maxi-
Gas lift allocation in the field has, in the mum. Mathematically this is where the gradient
past been conducted in a semi-automated manner, of a performance curve is O. Thus for well j,
via single well modelling of the inflow performance
#if.(Y.)
and verticai flow for each weii and then by combin - ~=(1 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . (12)
ing these results for overall field allocation manual-
ly through a priority ranking14’ls. Quite often Opera- where xj is the optimal gas lift quantity for well
tional experience on a well by well basis has influ- j.
enced decisions on changes made to the gas lift sys-
tem. The effect of the production system network If the total quantity of available lift gas
on the field performance is not possible using this is less than the unconstrained optimum, as
approach. Therefore, in order to automate the gas defined above, then the problem reduces to
lift optimization process, taking into account the allocating the available lift gas most efficiently
impact of the production system, a network model between wells. The definition of the con-
was constructed with the capabilities for gas lift strained optimum is that the gradients of all
optimization included. producing wells are equal, i.e. all wells would
increase their flow rate by the same quantity if
Description of the model an extra incremental amount of gas was inject-
ed. Thu~
A multiphase fluid flow simulator was used
to generate a system performance curve, i.e.:
127
8 Development of a Network and Gas LtitAlfocation Model for Production Optimization
128
EL-Massry,Y. and Price, A. 9
has the capability to calculate “what -if” scenarios by 850 STBOPD (1.3% and 1.5 Yo increase)
introducing new and/or closing- in low priority wells for three and four compressor availability.
and also predicting future lift gas demand.
3- Choosing one multiphase vertical flow
Having identified a production system correlation to describe all Ras Budran wells
bottleneck taking place through the pipeline from was not found possible based upon data
‘B” platform to the process platform “B”, a decision from bottom hole pressures surveys for
was taken to twin this short pipeline (120 feet) by each well. It was concluded that 11 wells
utilizing the existing, unused, water- injection line are best modelled using Orldsaewskl’s
with appropriate pipe- work looping. The estimated correlation, 4 wells using Hagedorn and
incremental production due to this modification Brown’s correlation, and only one welI using
using the model was 800 STBOPD. An estimated Mukh’jee and Br!!l’s Ccme!aticm.
increase in production of 600 STBOPD was real-
ized after completing this modification. 4- The pipeline performance connecting “Bw
and “C” platforms to the process platform
The introduction of a newly proposed well “RB-PP” is best described using the Muk-
to the production system was investigated using the herjee and Brill’s correlation, whereas the
model. The results indicate that the well would pipeline from “A” platform to “PP-B” is best
produce more than 4000 STBOPD with a minimum described using Oliemans correlation.
impact on the production rate of other producers
either in the amount of gas required to lift that 5- The removal of excessive pressure losses
well or through induced back pressure effects on taking place through the very short pipeline
other producers within the system. from “B” to “PP-B” platforms (as indicated
by a g!~ba! ~der estirnatica Cf ~Z~SSiii$2
129
10 Dcvetopment of a Network and Gas LiftAffocation Model for Production Optimization
2. El- Maaary, Y. :“Construction of a Network Model ence, Bahrain, 3-6 April 1993.
for an Integrated Production System and Appli- 16. Marshall, D.L., Edwards, R.J.E., Wade, K.C.
cation to the Ziet Bay Field” MS thesis, Cairo :*A Gas Lift ‘Optimization and Allocation Model
University, Giza, (1994). For Manifolded Subsea wells’” SPE 20979 pre-
3.”Ras Budran Reservoir Engineering Study”, sented at the European Petroleum Conference,
August 1987, BP Petroleum Development LTD. The Hague, Netheriantis, Gctober 22-241990.
4. Standing, M. B. :“A General Pressure Volume-
Temperature Correlation for Mixtures of California
011 and Greases,”Drill. and Prod. Prac.,API (1947),
275.
5. Lasater, J. A. :“Bubble Point Pressure correlati-
on,”Trans. ,MME (19S8),379 .
6. Glaao, O. :“Generalized Pressure Volume Tern-
perature Correlation,”JPT @lay 1980),p. 785.
9I. X?. -m..a” ..u. .. arid Beggs, H. D.:”cQrre!lM@ts
v usquG&, for
fluid Physical Property Prediction,” Paper SPE
6719, presented at the 52 nd Annual Technical
Conference and Exhibition of the Society of Petro-
leum Engineers,of AIME,Denver, Colorado (1977).
8. Brown, K. E. :“The Technology of Artificial Lift
i$%’tlmds,” PeiifiWe!! l%biishkg Co., Tdsa,(1977-
),Volume (l).
9. Brill, J. P. :“Multiphase Flow in Wells,’ JPT
(Jm.1987) 15-21.
10. Brown, K. E. and Lea, J. F. :“Nodal Systems
Analysis of 011 and Gas WeIls,” JPT (October 1985-
130
lTdie2SunnwydValiiMicalfhnvCmkiinCunbktiofj
lTwqoIWMMMti-sj
k
Table 1 Measured Versus Calculated PVT for Ras Budran Fieldl
For B#nd GOR
I
ill
Pressure Loss HOLDUP Flow Me@ Pmswe Lose NOLDLIP F&M-
I I I
Cp (Iow.r
!M
Pr22mur9 Bo Ro9B/STB G~. KF12TB I Uo, Unit) OR OR DWTD OR OR DR/TO
p-h Mou. tic. .ss.1 Cak. M*SS ,.- 1
S.000 l.let l.tee S*2 I 392 I 2.94 I 3.5* I BBR I BBR I BB/TD I
l=H--
4.000 t.leo 1.%79 9*2 3?2 2.se I 2.me I ORK I ORK I ORK I
e ,000 *.*79 1.*75 312 I, ---
-*- ,
I ----
9.*nl ----
c.sn
—
2,000 *lee l.*es ---
*42 1 342 I Z.*O I 2.*S 1
H=]
BJA BJA1/BJA2 TD
MB MB MB
F
-PA
A6 3.076 139.7s LSO1 2.73S 66
A’1 1.910 1Ss.7s ‘um S.l U .142 G w ‘r 4,,
A9 1.U9 Mam SAW S.W .249 WELLIIEAD w P
Bl 2.720 1s9.42 1.* 2.119 294
PLATFORM T
n 2,s1 1S9.U 4.2S7 4AM .247
M.* 4.890 . l=
RR-C
m 2.Ss6 4.4Ss
St 1.W4 LW.42 MM 2>74 -2.S8
w
m 1.%0 L..* X4% I 3.9U .437
M SW lm.u S,44’J km! .1.6 LEGEND:
et Lcn M9.u ZSa 1s19 NJ
w Mm I.w.u lam 1.617 4S0 P= 1Z“ Prc.duc lion Line PROCESS
es 4.s20 1s.42 m 890 4 PLATFORM
C4 s.m9 141-W 4.1s 4.6t9 .S.4 T =S- Test Llnc RB. PP-B
TOTAL -- IU.1’J I 4s,0n 41>m -1.-
I
14C=S”
I#lah Prrss. Lltt Gas
Total Field Network Model Verification : WELLREAI)
LP =20” Low Press.Prod. Gas
PLATFORM
Cumulative Error % (CE%)= 4.4 W= 12” Waler InJccllon
-4000 =
=
g -s000
g -5600-
s .-s -6060-
-s000 - ~
=
: - -70@o-
=
= -lobo -
-M60 -
-Boon-
-s000 -
-9000-
-Imoo -
- I 6boo-
-I IQOO -
-11000 -
500 1000 1500 Xmo Moo
100 400 600 ma 1000 1?00 1400 1600 1800 2000 2100
P,e,s,Ir (PSIA)
r,, (2-0) [.~le
P,,,,,, * (PSIA)
ot V*I1 Pe,l.,mmte Ilislw, match
f,q (1-b) etmole .1 .rll Pttlwmo.ce Ili.lety Melcb
I M
~~w
if”<-
‘K
\
100
tHACfOORONt DROWN RIVISIO CORRIIAIION
95 XOUK1[R (AWF1A161CAM) COR XOUK1[R (ACA+fLANtCAM) CORRILAIION
0B8R
:Z;’’:::$lq (BR MAP) G OBER (BB MAP)
90-1 AOUSIS& ROSS CORRIIAIION
~wIIHIRJ[[”A R@ltL CORR[lAIION VAIUVIIfRJrta BBILL CORRILAIION
I “-
85 @OllKLIR ANO CA1ON IOUKLfR m cAlotI
4011 CORR[lAIION iOll CORR[lAilOK
m TAcA+f LkNIW (ERIHIIII noto-up) VhCAiFLAn16AN (8RIUI#l noso-up)
OUEASURIOOUILII PIUSSUR[ ON 09/09/93. EtAItASURtOOUILtI PRcssuRcosIo9/09/1993.
15- >
132
-,.
+
CALCULATE FLOW RATES IN
INPUT SRANCHES (WELLS)
+
WRLLS
MIX
ODTAIN
FUJI09
COMPOOITt
PROM
+
ON IN
TO
PIPELINR6
I
ITERATE FOR PRESSURE
CALCULATION AT THE SEPARATOR,
+
[,
——. .-- .. . 1 I
I ~Fig. 4 NETWORK iViODEL CALCUWI iwN A~~~~i~i+ivil
l—
4500
. . .. ... ...
~ ~~
—
r ~Se. ,~to —
erformanc
atch
Evaluation
Previouml y e:st epara or
P.L,w.c.,FGLR,
Selected r resalts,BHP SuWeY
IGLR,Pr,P.O.G.E.
Correhstion
Iv .3 . %’
~–
r wELL/PIPELINE
PERFORWNCE MODEL —
K&-l-
1 Flow Tables I
4
istor
PREDICTION Yes
~~ atch
ii
..-
1
-4
.2
[! 0
27 36 45 54