You are on page 1of 8

Production Forecasting

for Gas Fields With


Multiple Reservoirs
Lelf M. Mellng, SPE, Per O. M.rkeseth, and
Thore Langeland, Statoil

Summary. A new method to predict field performance for gas fields


consisting of small reservoirs of pointbar origin was developed, implemented, and tested. Conventional
geological and reservoir engineering
modeling of this type of field is difficult
and usually gives erroneous results.
This paper describes the evaluation
and use of exploration- and appraisal-well data as input for a stochastic
geological model and the use of output from that model as input for a
semianalytical reservoir performance
model. The semianalytical model includes production constraints for
well, platform, and field conditions.
Example applications of the stochastic geological model and the semianalytical model are also presented.

Introduction
Oil and gas fields that consist of sand
deposits of fluvial origin may have extremely complicated or even chaotic structures.
Sand lenses are discontinuous and cannot be
correlated from well to well. Fields can consist of thousands of separate reservoirs. The
thickness of the pay intervals can be very
large (e.g., up to 1000 m). Between pay intervals, reservoirs may be water saturated
and fluid properties may change rapidly
from reservoir to reservoir. Fig. 1 illustrates
the complexity of this type of field.
A geological evaluation should estimate
not only in-place reserves, but also reserves
penetrated by wells. The evaluation must
also consider the variations of reserves at
all levels because averaging of volumes and
properties gives overly optimistic field forecasts. Variations usually lead to bottlenecks
and surprises.
Stochastic geological models have been
presented in the literature. Haldorsen and
Lake 1 developed a 2D model that distributes shale intervals stochastically. Augedal
et az.2 developed a 3D model that distributes sand bodies that are parallelpipeds. Both
models assume a constant net/gross ratio
over the field.
The model presented here is a 3D model
designed to describe point-bar deposits in a
mud-rich environment. The analytical model
is designed to handle data from a large number of reservoirs (> 1,0(0) with complex
production controls. The models presented
in the literature 3 ,4 do pot take into account
the complex production procedures and
reservoir management program.

Geology of Meandering
River Systems
Meandering river systems are normally
formed in areas of relatively low slope.
Heavy vegetation and cohesive flood-plain
deposits make rivers more stable and favor
development of meandering rather than
braided river systems. The flow pattern of
meandering systems causes erosion at the
outer bank and deposition at the inner bank.
Thus, the position of the river changes and
point bars form (Fig. 2).
Copyright 1990 SOCiety of Petroleum Engineers

1580

Each individual point bar is a fining-up


sequence, with high-energy deposits (gravel
and sand) at the bottom and low-energy
deposits (such as silt and shale) at the top.
Fig. 3 shows a typical gamma ray response
from such a sequence. The vertical shift
from sand to shale intervals in the wells
reflects the shifting nature of rivers owing
to avulsion of the meander-belt. In sand-rich
meandrous environments with low subsidence rates, meander-belt deposits commonly develop extensive sand bodies that are
parallel to the overall transport direction of
the river, where the continuity is very good.
In a more mud-rich environment with a
high subsidence rate, isolated point bars
form. Some amalgamated sequences may
occur, but the continuity is very poor. Fig.
I shows single and amalg8.!llilted sequences.
The water-saturated intervals in Well B
clearly demonstrate the lack of continuity for
both sequences.
The dimensions of point bars are related
to the size of the river system. Several
papers related to this subject have been published. In this work, we used Leeder's5
correlation between channel depth and channel width and assumed that the channel
deposit thickness was equivalent to channel
depth:
be =3.6h c 1.54. . . . . . . . . . . . . . . . . . (1)

We used Lorenz et at. 's6 correlation between channel width and the meander-belt
width to estimate the dimensions of the point
bars:

bm =7.44b c 1.01 . . . . . . . . . . . . . . . . . (2)


Data From Exploration Wells
Channel thickness can be estimated by inspecting raw logs, especially the gamma ray
log. Intervals can be defined as single or
amalgamated sequences of channels. Singlechannel sequences are used to estimate channel thickness as indicated in Fig. 3. If several
channels are found, an average thickness for
a given formation and a given well can be
estimated. The interpreted logs are used to
estimate the probability of hydrocarbons in
a specific sand interval. A net sand is an interval with a porosity higher than and a clay
content lower than predefined values.
December 1990 JPT

~
...,71'"

2000 m

Well A - - - - - - - - - Well B

GR

OR

OR

200

I
net

net ray

pay

l
1,64

1.01

b< 3.6 h.

b m 7.44 b<

(Leeder 1973)

Single channel

(Lorentz et al. 1985)

Amalgamated channel
sequence

Fig. 1-Sand continuity in a mud-rich


meandering channel system.

Fig. 2-Size estimate of a meandering


river system.

Fig. 3-Channel thickness and net pay estimated from gamma ray log.

An additional criterion for net pay is a


water saturation less than a given value. The
probability of a hydrocarbon-filled interval
is equal to the sum of hydrocarbon-filled intervals divided by the sum of net-sand intervals.
For each formation and well, the thicknesses of the sand and shale intervals are
estimated. On the basis of these data, distribution functions are made (see Figs. 4 and
5). Distribution functions for porosity and
water saturation are estimated in the same
manner.
A correlation between porosity and permeability is estimated by correlating welltest permeability to log porosity and water
saturation:

can also be made for reservoir pressure and


temperature. These correlations can be used
to estimate fluid properties in the model.

as ,gas- or water-saturated according to the


probability of hydrocarbons in that area.
Water-bearing intervals are then deleted, as
well as sand intervals thinner than a predefined value. For the remaining sands,
porosity and water saturation are estimated
with a random-number generator and the
distribution functions.
The radii for the sand bodies are estimated
by using the interpolated channel depth
values as input for Eqs. I and 2. The additional HCPV included by this event is then
estimated and compared with the expected
value for the whole formation. New locations are picked, and the same procedure is
used until the generated HCPV is equivalent to the value calculated by conventional
geological interpretation.
A predefined well pattern and well paths
are included in the model. For each well,
directly and indirectly penetrated reservoirs
are registered and the HCPV is calculated.
Net pay, porosity, and water saturation were
registered for the directly penetrated sand
intervals as illustrated in Fig. 7.

k=exp(alcP+a2Sw+a3)' ........ (3)

Fig. 6 shows the predicted vs. the estimated


permeabilities.
Fluid data are evaluated, and depth correlations for gas gravity, critical pressure and
temperature, and heating value and condensate content are made. Depth correlations

Stochastic Geological Mode.


It is impossible to duplicate the complicated
field geometry with numerical models. With
statistical methods, however, it is possible
to design a comparable model. Our model
distributes point-bar reservoirs stochastically. Therefore, the results must be considered to be possible geological realizations.
In the model, each formation is handled
separately. A random location within the
field limits is picked and distribution functions for shale and sand interval thicknesses,
porosity, and water saturation are generated
by interpolation. Distribution functions are
also estimated for channel thickness and
probability of hydrocarbons. The interpolation is performed with data from the three
nearest wells.
The distribution functions are used to
generate probable shale and sand thicknesses
by depth. Each sand interval is registered
40 1

30

;,
~
~

:;;
~

C>

20

;':
~

I
~

10

00

0.1

0.2

0.30.4
CUM

0505

PROBABle i TY

0.70.8

0.9

1.0

(FRAC.)

00

0 1

0.2

0.3
CUM.

Fig. 4-Cumulatlve probability, individual sand Interval


thickness.
JPT December 1990

0.4

0.5

PROBABI L I TY

0.5

0.7

0.8

0.9

1.0

(FRAC.)

Fig. 5-Cumulatlve probability, Individual shale Interval


thickness.
1581


10000 Permea~ility from cor!e~~!~ __________ _

1000

D
D

Permeability in md

Directly penetrated
Indirectly penetrated
Un penetrated

100

10

10

100

1000

10000

Test permeability

Fig. 6-Permeabllity from well tests compared with correlation.

Analytical Reservoir Model


This model predicts potential production by
combining material balance, bottornhole
deliverability, and wellbore hydraulics.
The material balance for dry gas can be
written as
p/z =(p;/z; )(1-ER) ............. (4)
and rearranged to become
p2 = [z(p;!z; )(1- E R)]2. . ........ (5)
The bottornhole deliverability is best described by the laminar/inertial/turbulent
flow equation:

The laminar flow coefficient is defined as

x(

7.056re )
2

+O.869s, . ..... (7)

Fsrw

B F =3.14 x lO-6-yZT{3/h 2 rw '

......

(8)

Several correlations for the internal flow


coefficient have been published. This model
uses Geertsma's 7 correlation:

4.85 X lO4
{3=

[4>(1-Sw)p5k O.5

............ (9)

............. (lO)

This equation can be used as a regression


equation for predictions of wellbore hydraulics for more complicated models. For the
example runs shown later, the depths of the
different reservoirs varied considerably.
Therefore, both AT and BT were correlated
with reservoir depth (an average well deviation was used):
A T =a4 +a5D+a6D2, .......... (11)

In field studies, it is often impractical to


use tables to describe pressure drop in tubing. Several runs on wellbore-hydraulics
models are often performed to predict pres-

25

sure drop at different rates, wellhead pressures, tubing lengths, and/or gaslliquid ratios (GLR's). The results from these runs
can often be summarized with an appropriate regression equation. The single-phase
Smith S equation for vertical gas flow can
be written as

Pwl =AT+ BTq/

and the turbulent flow coefficient as

p2 -pw/ =AF Qg+BFQg 2 . ........ (6)

The pressure squared is valid only for systems where 2p/Jl.g z is linear with pressure.
This is normally true for low pressures, but
for practical purposes the pressure squared
can be used for high pressures if the drawdown is not too large.

Fig. 7-Directly, indirectly, and unpenetrated reservoirs.

and BT=a7+asD+a9D2 . ........ (12)


The correlations can also easily include variations in wellhead pressure and GLR.

Gas rate potential (MMscf/d)

-----------

20

"Our model distributes


point-bar reservoirs
stochastically.
Therefore, the results
must be considered to
be possible geological
realizations. "

15
10

-~

5
0
0

0.2

0.4

0.6

0.8

Recovery factor

--

Very good zone

-j-

Good zone

Poor zone

--8-

Very poor zone

Fig. a-Potential gas rate vs. recovery.


1582

December 1990 JPT

~ERCf~jT~G~

READ DATA FROM

MAKE WELL
POTENTIAL
CORRELATIONS

INITIALIZE

STOCHASTIC

MODEL

GEOLOGICAL

,,0

MODEL

40

COMPLETE
OR

20

RECOMPLETE

WELL, PLAT
FORM AND
FIELD

CHOOSE WELLS

POTENTIAL

TO BE

t-

98

1 00

II
1 OJ

04

t+DElTAt

I RECOMPLETED

IN PLACE. RESlRVlS (PERCENT OF AVERAGfl

YES

Fig. 10-Variations in penetrated gas


reserves.

REDUCE
RATE FOR

UPDATE

WELLS THAT

MODEL

DO NOT MEET
SPECIFICATIONS

PERCENTAGE
4Q
NO

YES

PRODUCE

3C

FIELD,
PLATFORMS
AND WELLS

NO

AT MAXIMUM

20

RATE

NO

SALES
SPECIFICATIONS

YES
I

OK?

PRODUCE AT
PLATEAU

I> I MAX?

10

ACCORDING TO
POTENTIAL

YES

27
VOLUME

PRINT

81

4.3

(BSCF)

RESULTS

Fig. 11-Variations in reserves on a well


basis.

Fig. 9-Reservoir model flow chart.


Wellhead deliverability can be estimated
by combining Eqs. 6 and 10 as follows:

(Br+BF)qgZ +AFqg +Ar-Pz =0.


................ (13)

Fetkovich 9 suggested a similar procedure


that replaced Eq. 13 with the backpressure
equation. This is not necessary. Solving for
qg and replacing pZ with Eq. 5 gives

qg=[ -AF+(AFZ -4(Br+BFHAr

qg =A(ER)Z + B(ER) + C ......... (15)


describes the development of maximum
potential gas rate vs. recovery with good accuracy. Exponential decline implies that the
gas rate vs. recovery is linear (A =0).
Recovery at a given economic rate can be
estimated by solving the decline-curve
equation:

-B-..JBZ -4A[C-(qg)min]
2A
. ............... (16)

-[z(p;!z;)(1-ER)]2}) 'h ]+2(Br +BF).


................ (14)

Eq. 14 gives the well potential at a given


wellhead pressure as a function of recovery .
In theory, an iteration procedure must be
used to solve the equation. In practice, however, it is possible to speed up computation
time by generating decline curves.
For some selected pressures (and thus
recoveries), the potential gas rates are calculated by solving Eq. 14 numerically. Fig.
8 shows that the maximum potential gas rate
as a function of recovery exhibits a nonlinear
trend. Regression analysis indicates that the
cubic equation
JPT December 1990

By definition,

ER=GpIG; . ................... (17)


Substitution into Eq. 15 gives

t=JG [A(:Y+B(:)+Cr1dGp.
................ (20)
Eq. 20 is solved analytically with the following substitutions:

a=AIG;Z, .................... (21)


b=BIG;, ..................... (22)
c=C, ........................ (23)
u=Gp +b/2a, ................. (24)
and D= -(4ac-b z )/4a Z. ......... (25)
The solution of the integral then becomes

1[ 2.JD1 [( luo+.JDI)
uo-.JD

t=-;;

In

-(lnl:~11)J],
Gas rate is defined as the change of cumulative production per time unit:

qg=dGpldt, .................. (19)


which leads to

...........

(26)

where Uo =b/2a. . ............... (27)


The minimum time needed to produce a
given reservoir to its economic limit can be
estimated by specifying (Gp ) max calculated
from Eqs. 16 and 17.
1583

PERCENTAGE

40

35

---------1

Gas rate potential (MMscf/d)

30

30

25
20

20
15
10

10
5

o
o

o
VOLUME

15

20

25

30

35

(BSCF)

Fig. 12-Varlatlons In reserves for individual reservoirs.


The solution of the integral can be rearranged to become
-2-JD(I +e E)

2(1-e E)
-([2-JD (I +e E)]2 - {4(1-e E )
x [-JD (I-eE)]}) 'h !2(I-e E )
-bI2a, ...................... (28)

where the superscript .

E=2(ln\:::1\-2a-JD~t) ..

(29)

and Uo =(Gp)old + (bI2a) . ......... (30)


The average potential production for a
given period (~t) can then be estimated
because

(<1> )new -(Gp )old

........ (31)

~t

Production Control
Production control is designed to be autonomous. The principles are based on the production and reservoir management most
likely to optimize plateau length, well and
platform startup, recovery, and sales-gas
specifications. Fig. 9 is a flow chart of the
model.
Both single and dual-completed wells can
be handled by the model. The upper part of
the well is assumed to be produced by a
short string and the lower part by a long
string.
Production startup for a well can be given
as a fixed or floating startup time. Wells
specified with a floating startup time are put
on production when needed to maintain the
plateau production. A time delay can be used
to take into account the drilling schedule on
a platform. As an initial guess, each reservoir is assumed to be produced to the minimum economic rate, which is given as input.
The maximum recoverable gas volume for
1584

10

Cumulative production (Bscf)

Fig. 13-Potentlal gas rate vs. cumulative production.

each sand and the minimum time needed to


produce these reserves can then be calculated. If the minimum production time for
a producing reservoir is less than the
timestep length, then the reservoir is produced at maximum rate in that timestep and
the string is recompleted in the next
reservoir.
For each string, the potential full wellstream gas production is calculated. The
potential string sales-gas production is calculated with Eq. 32:
(qgps )string = (qgpw)string(R ws ). . ...

(32)

If the calculated potential string production


is higher than a maximum predefmed value,
the potential production is reduced to this
maximum value.
If necessary, the potential string values are
corrected for startup andlor shut-in within
the timestep.
The potential platform-well sales-gas production is calculated by adding the potential values for all wells connected to that
platform:
(qgps ) plat = (Feff) platE (qgps )string'

................ (33)

The efficiency factor takes into account that


all wells will not be available for production at all times and reflects downtime resUlting from wireline work, workovers,
maintenance, etc.
If the calculated potential platform-well
sales-gas production is higher than a maximum predefined value, the potential
platform-well sales-gas production is reduced to this maximum value.
The potential field sales-gas production is
calculated in the same manner as the potential platform-well sales-gas production. The
target is the sum of the sales-gas volume and
the fuel gas consumed on the platforms.
Both inputs are given as time-specified
figures:
(qgst ) field = (qgs )field + (qfuel) field . " (34)

The potential field sales-gas production target is defined as


(qgpst ) field = (qgs ) field (Fswing) + (q fuel) field'

............... (35)

The swing factor takes into account variations in market demand and allows for production at higher rates for shorter periods
of time if necessary.
For each timestep, sales-gas production
and potential production targets are compared with the potential field sales-gas production. If the potential field production
exceeds the target, then no optimization is
necessary. If the potential production is less
than the target, then optimization is performed if possible. Two optimization
methods may be used: (I) recompletion (the
first choice) and (2) drilling of new wells.
The additional field production needed to
meet the targets is calculated for each
timestep. For each platform, the potential
spare sales-gas production is calculated with
Eq.36:
(qgsS)plat = (qgsmax)plat -(qgps )plat

............... (36)
This potential production is the difference
between the maximum platform production
and the potential platform sales-gas production at that timestep.
A maximum recompletion rate is entered
into the model. Reservoirs with potential
rates lower than this rate are assumed to be
available for recompletion. For strings that
satisfy this criterion, the additional production that can be obtained by recompletion
is then calculated with Eq. 37:
(~qgps ) string = (qgps ) newres - (qgps ) oldres .

............... (37)
Additional potential platform and field
production is then calculated. The additional
potential platform production is set equal to
the potential spare production if the sum of
the additional string production exceeds the
December 1990 JPT

PERCENTAGE

PERCENTAGE

30

40

30
20

20

10
10

o
0.750

0.753

0.756

0.759

0.575

MAX. RECOVERY FACTOR (FRAC.)

JPT December 1990

0.625

0.650

ACT. RECOVERY FACTOR (FRAC.)


Fig. 15-Actual field recovery.

Fig. 14-Maxlmum field recovery.


potential spare production at a given platform. If the additional potential field production from recompletion plus the potential
field production without recompletion exceeds the production target, then recompletions are preferred.
Strings are recompleted according to the
following guidelines.
1. Strings producing from reservoirs with
the smallest remaining reserves are recompleted first.
2. No recompletion is performed on a
platform with a potential sales-gas rate equal
to or higher than the maximum rate specified for that platform.
.
If possible, recompletions are performed
until the potential field sales-gas target is
reached. If obtaining the field sales-gas production target by recompletion is not possible, then new wells are put on production
(if such wells are defmed) and recompletions
are not performed.
Minimum platform and field rates are
entered into the model. If the minimum platform rate cannot be obtained by recompletion, then the platform is shut in. If the
minimum field rate cannot be obtained by
recompletions or new wells, then all platforms are shut in.
If the potential sales-gas production is
lower tlian the target sales-gas production,
then the platforms and strings are produced
at their potential rate. If the potential salesgas production is higher than the target salesgas production, then rates are allocated according to the following guidelines.
1. Strings that can produce their reservoirs to the minimum economic rate will do
so and are recompleted within the timestep.

0.600

2. A field sales-gas target is set for each


platform according to its potential sales-gas
production.
3. A platform production target is then set
for each individual string according to its
sales-gas production target.
Platform rates are set according to Eq. 38:
(qgs)plat = (qgts ) field [

(qgps ) plat ]
.

E(qgps ) plat

........... , .... (38)


Well rates are set with 'Eq. 39:
(qgs) string = [(qgs)plat - (dqgs)plat]

x[ (qgSp):~ps)string
(dqgs)plat

] ..... (39)
'.

(Feff)plat

A sales-gas specification can be entered


into the model. If the sales-gas production
does not meet this specification, then the
model reduces the potential sales-gas production for the strings that do not meet specification. Platform and string rates are reset.
This optimization is performed until the
specification is met or until the adjusted
potential field production is equal to the target field sales gas production.

Applications
The models were tested in field studies. Ten
simulation runs were performed. The number of cases was not adequate for a complete
Monte Carlo simulation, but the cases give
a good indication of possible production
scenarios.

The in-place HCPV was identical for all


runs, but the wells penetrated various
reserves. Roughly 79% of the in-place
reserves were penetrated by wells. The variations between the different cases were
rather small, ,.. 5 %, as indicated in Fig. 10.
On a well-by-well basis, however, the
variations in reserves were more significant,
as indicated in Fig. 11. The model predicted
that about 10% of the wells were dry holes.
This information is important for future field
development. In an actual production situation, this can be accounted for by allowing sidetracking or new wells.
Fig. 12 shows the volume distribution for
the individual reservoirs. Note the similarity between Figs. 11 and 12. On average,
each well penetrates 10 reservoirs. The
volume distribution for the wells is therefore scaled with a factor of 10 compared
with the volume distribution for the individual reservoirs.
A well may have large fluctuations in well
productivity vs. time and cumulative production. Some reservoirs deplete quickly
and the well must be recompleted (see Fig.
13). The large variations cause difficulties
in maintaining a production plateau while
optimizing recovery.
Production is optimized to maintain the
plateau level as long as possible. Wells are
recompleted, and new wells and platforms
are put, on production. Recovery is controlled by the well, platform, and field production constraints. The recovery factor
obtained is defined as the actual recovery
factor. If the field is perfect, all reservoirs,
wells, and platforms are produced to their
economic limits and the recovery and recovery factor are maximum. In our study, a
1585

PLA TFNO= 7

Most likely

Gas rate
PERCENTAGE
30

Gas
potential
20

Optimistic

Pessimistic
10

o
7.8

8.4

9.0

9.6

PLATFORM START UP (YEARS)

Fig. 16-Predlcted startup times for new


platforms.

Fig. 17-Slmulated productiontlme scenarios.

large difference ("" 14%) was found between the actual and maximum recovery factors. This difference can be seen by
comparing Figs. 14 and 15. The difference
was caused by the production constraint that
optimizes plateau length before recovery.
Variations in reserves at the reservoir, well,
and platform levels enhanced the difference.
For the fields discussed in this paper, it
is difficult to predict an actual time when
new wells and/or platforms are needed to
maintain plateau level. Fig. 16 shows predicted times for new platforms.
The simulations give 10 different production scenarios. Fig. 17 shows three of these.
Sales-gas production and potential sales-gas
production are included in the figure. At the
plateau, the struggle to maintain the production level is indicated by the erratic development of the potential production. In the
most optimistic case, the potential production is obtained with fewer interventions and
is less erratic.

Nomenclature
a,b,c,

Conclusions
1. The stochastic geological model makes
possible geological realizations for fields that
consist of point-bar reservoirs.
2. A complex reservoir management program can be tested with the semianalytical
performance model. Critical bottlenecks can
be identified and resolved. The autonomous
production control makes the simulation
runs easy to perform.
3. Simulations show the importance of
describing the variations in reserves at the
reservoir, well, and platform levels.
4. The models give possible production
scenarios resulting from different geological realizations. Differences can be large,
depending on the type of field studied.
1586

A ,B, C = correlation coefficients


AF = laminar flow coefficient, LIT
equation, psi 2 /(scf/D)
[kPa 2/(std m 3 /d)]
AT = static flow coefficient, Smith

be =
bm =
BF =

BT =

D =

ER =
F =

Fs

Gi =

Gp =
h =
he =
k =

p=
Pwf =
qfuel =

qg =
Te

Tw

Rws

equation, psi 2 [kPa 2]


channel width, ft [m]
meander-belt width, ft [m]
turbulent flow coefficient,
LIT equation,
psi2/(sef/D)2
[kPa 2/(std m 3/d)2]
turbulent flow coefficient,
Smith equation,
psi2/(sef/D)2
[kPa 2/(std m 3/d)2]
depth, ft [m]
recovery, fraction
factor
shape factor
initial gas in place, sef
[std m 3 ]
produced gas, scf [std m 3 ]
formation thickness, ft [m]
channel thickness, ft [m]
permeability, md [ltm 2 ]
reservoir pressure, psi [kPa]
flowing bottomhole pressure,
psi [kPa]
fuel gas consumption, sefID
[std m 3 /d]
gas production rate, sefID
[std m 3 /d]
drainage radius, ft [m]
well radius, ft [m]
gas well-stream ratio, scf/bbl
[std m 3 /m 3 ]

Sw
t
Ilt
T

=
=
=
=

z=
ex =
{3 =
l' =

Itg =
tf> =

skin factor
water saturation, fraction
time, days
timestep, days
reservoir temperature, OR
[K]
gas deviation factor
correlation coefficient
interual resistance coefficient,
ft- 1 [m-I]
full well-stream gas gravity
(air = 1)
gas viscosity, cp [mPa s]
porosity, fraction

Subscripts
eff =
i =
max =
min =
P =
plat =
res =
s =
S =
t =

efficiency
initial
maximum
minimum
potential
platform
reservoir
sales
spare
target
w = wellstream

Acknowledgments
We thank Statoil for permission to publish
this paper and Olav Vikane for his helpful
review.

References
I. Haldorsen, H.H. and Lake, L.W.: "A New
Approach to Shale Management in Field-Scale
Models," SPEJ (Aug. 1984) 447-57.
2. AugedaI, H.O., Stanley, K.O., and Omre, H.:
"SISABOSA, A Program for Stochastic Modelling and Evaluation of Reservoir Geology,"

December 1990 JPT

3.

4.

5.

6.

7.

Report SAND 18/86 presented at the 1986


Conference on Reservoir Description and
Simulation With Emphasis on EOR, Oslo,
Sept. 5-7.
Saif, M.A., Kumar, R., and Shanyoor, M.:
"Mixed Integer Linear Programming Model
for MuItireservoir Strategic Planning, " paper
SPE 15759 presented at the 1987 SPE Middle
East Oil Show, Bahrain, March 7-10.
EI-Feky, S.A.: "A Combination Gasfield Development Model Evaluated With Field Data,"
paper SPE 16937 presented at the 1987 SPE
Annual Technical Conference and Exhibition,
Dallas, Sept. 27-30.
Leeder, M.R.: "Fluviatile Fining-Upward
Cycles and the Magnitude of Paleochannels,"
Geological Magazine (May 1975) 110, No.3,
265-76.
Lorenz, J .C. et al.: "Determination of Widths
of Meander-Belt Sandstone Reservoirs From
Vertical Downhole Data, Mesaverde Group,
Piceance Creek Basin, Colorado," AAPG
Bulletin (May 1985) 69, No.5, 710-21.
Geertsma, J.: "Estimating the Coefficient of
Inertial Resistance in Fluid Flow Through
Porous Media," SPFJ (Oct. 1974) 445-50.

JPT December 1990

8. Smith, R.V. : " Determining Friction Factors


for Measuring Productivity of Gas Wells,"
Trans., AIME (1950) 189, 73-82.
9. Fetkovich, MJ.: "Multipoint Testing of Gas
Wells," SPE Mid-Continent Section, Continuing Education Course on Well-Test Analysis
(March 17, 1975).

Authors

51 Metric Conversion Factors


ft x 3.048*
E-Ol = m
ft' x 2.831 685
md x 9.869233

E-02 = m'
E-04 = p.m'

'Conversion factor Is exact.

Provenance
Original SPE manuscript, Production Forecasting for Gas Fields With Multiple
Reservoirs, received for review Oct. 2,
1988. Paper accepted for publication Sept.
18, 1990. Revised manuscript received Dec.
21,1989. Paper (SPE 18287) first presented
at the 1988 SPE Annual Technical Conferc
ence and Exhibition in Houston, Oct. 2-5.'

JPT

1587

You might also like