Development and Application of the

Multiwell Productivity Index (MPI)

P.P. Valkó, SPE, L.E. Doublet, SPE, and T.A. Blasingame, SPE, Texas A&M U.

Summary
In this article we generalize the concept of the pseudosteady-state
productivity index for the case of multiple wells producing from
or injecting into a closed rectangular reservoir of constant thick-
ness. The work complements the analytical study by Rodrı́guez
and Cinco-Ley1 for systems produced at constant flowing pres-
sures. Wells are represented by fully penetrating vertical line
sources located arbitrarily in a homogeneous and isotropic reser-
voir. The multiwell productivity index 共MPI兲 is a square matrix of
dimension n, where n is the number of wells. The MPI provides a
simple, reasonably accurate and fast analytical tool to evaluate
well performance without dividing the cluster into single-well
drainage areas. The MPI approach is used to obtain approximate
analytical solutions for constant 共but possibly different兲 wellbore
flowing pressures, and to visualize the resulting pressure field. In
The four corner points of the rectangle are located at 共0, 0兲,
(x e ,0), (x e , y e ) and (0, y e ). In any given time interval, the num-
ber of wells, n, their locations (x w j , y w j ), the wellbore radii, r w j ,
and the skin factors, s j , are considered constant. Fig. 1 shows a
schematic of the reservoir.
For the case where only one well is on production 共say, the jth
one兲, Ozkan5 共see also Ref. 6, page 107兲 gives the pressure distri-
bution in the reservoir during the pseudosteady state as
␣ 1␮ B
p̄⫺p 共 x,y 兲 ⫽ a 关 x D ,y D ,x wD j ,y eD 兴 q j , 共1兲
2 ␲ kh
where the influence function a 关 兴 共a dimensionless drop in pres-
sure兲 is given by

冉 冊
addition, the skin factor trace technique is introduced as a tool to
monitor a cluster of wells. The MPI technique is illustrated using 1 yD yD 2
⫹y wD
2
a 关 x D ,y D ,x wD ,y wD ,y eD 兴 ⫽2 ␲ y eD ⫺ ⫹ 2
a synthetic example taken from Ref. 2, as well as two field cases. 3 y eD 2y eD
⬁
tm
Introduction
⫹2 ␲ 兺
m⫽1 m
cos共 m ␲ x D 兲

Although it has been known from the inception of reservoir engi-

⫻cos共 m ␲ x wD 兲 共2兲
neering studies that any change in the condition of a well 共dam-
age, stimulation, choke change兲 modifies the production charac-
and
teristics of the well-reservoir system, only the pioneering work of
Rodrı́guez and Cinco-Ley1 approached this problem using analyti- cosh关m␲共yeD⫺兩yD⫺ywD兩兲兴⫹cosh兵m␲关yeD⫺共yD⫹ywD兲兴其
tm⫽ , 共3兲
cal techniques. The constant flowing wellbore pressure solution of sinh共m␲yeD兲
Ref. 1 has been improved in several ways by Camacho-V. et al.,2
with x D and y D defined as x/x e and y/x e , respectively. In the
including the possibility of allowing wells starting to produce at
Appendix we introduce several computational simplifications to
different times.
ensure a fast and reliable approximation of the infinite sum in Eq.
Similar to the work in Refs. 1 and 2, we consider liquid flow
2. Also shown in the Appendix is the relation of the influence
within a homogeneous reservoir of constant thickness. This work
function to the Dietz7 shape factor.
is, however, restricted to pseudosteady-state flow conditions. The
By superpositing, for the n-well system with n production/
results can be used for any type of wellbore condition so long as
injection wells, we have
the pseudosteady-state approximation is acceptable.
Pseudosteady-state flow is the 共idealized兲 finite-acting portion n
of the constant-rate solution for a bounded reservoir,3 and this ␣ 1␮ B
lends itself to a simpler description. The physical meaning is that
p̄⫺p 共 x,y 兲 ⫽ ⫻
2 ␲ kh j⫽1 兺
a 关 x D ,y D ,x wD j ,y wD j ,y eD 兴 q j , 共4兲
all elementary portions of the reservoir contribute to the overall
production rate by the same amount. Depletion is a parallel shift where the q j production rates are ‘‘constant’’ for the given time
of the pressure distribution with time. interval. All prior information is contained in the average pres-
While, formally, the pseudosteady state is a limiting case of the sure, p̄, and, as such, we do not need to specifically account for
constant-rate solution, solution-gas drive reservoirs spend most of the initial pressure distribution or the production history.
their life in a series of states, and closely resemble this flow The two basic vector quantities we would like to relate are the
condition.3 pressure drawdown vector,

冋册冋 册
The basic concept in this article is the multiwell productivity
index 共MPI兲 matrix, which relates the production rate vector to the d1 p̄⫺p w f ,1
pressure drawdown vector. The pseudosteady-state flow condition
d2 p̄⫺p w f ,2
ensures the uniqueness of the MPI matrix for a given system. dជ ⫽ ⫽ , 共5兲
共Without the assumption of the pseudosteady-state flow condition, ] ]
this uniqueness would be lost, even for a single-well system.4兲 dn p̄⫺p w f ,n

Multiwell Productivity Index and the surface production-rate vector,

冋册
We consider a rectangular homogeneous reservoir of uniform
thickness, h, porosity, ␾, permeability, k, and no-flow outer q1
boundaries. The single-phase fluid viscosity, ␮, and the total com- q2
qជ ⫽ . 共6兲
pressibility, c t , are considered constant. The wells are represented ]
by line sources. qn

Copyright © 2000 Society of Petroleum Engineers Applying Eq. 4 for a point located at the circumference of well-
Original SPE manuscript received for review 20 July 1998. Revised manuscript received 22
bore i and taking into account the drop in pressure due to the skin
July 1999. Paper (SPE 51793) peer approved 14 December 1999. factor, we obtain

SPE Journal 5 共1兲, March 2000 1086-055X/2000/5共1兲/21/11/$5.00⫹0.50 21

 The dimensionless MPI matrix 关 JD兴 plays the same role as the
dimensionless productivity index,

J D ⫽1/ ln 冉 rw 3
⫺ ⫹s ,
re 4 冊 共15兲

for a single well located in the center of a circular reservoir.

Using vector notation, the overall production rate, q T , can be
written as
q T ⫽ជ To dជ , 共16兲
ជ To
where the overall productivity index, , is the vector obtained as
the sum of the rows of the MPI matrix. The overall production
rate is the scalar product with the drawdown vector. 共The super-
script T denotes ‘‘transposed,’’ and it is used here because vectors
without the superscript T are treated as column vectors.兲

Applications
Constant Flowing Wellbore Pressure Solution. The overall ma-
terial balance for a slightly compressible liquid system is
dp̄ ␣2
Fig. 1–Well-reservoir system. ⫽ q . 共17兲
dt c t V p T
Substituting Eq. 16 into the material balance relation we obtain

冉兺
dp̄
␣ 1␮ B
n ⫽c 2 ⫺c 1 p̄, 共18兲
dt
p̄⫺p w f ,i ⫽ ⫻ a 关 x wDi ,y wDi
2 ␲ kh j⫽1 where

冊
⫹r wDi ,x wD j ,y wD j ,y eD 兴 q j ⫹s i q i . 共7兲 c 1⫽
␣2
c tV p
共关 1 1 ¯ 1 兴 ជ o 兲

Eq. 7 can be written for all wells, i⫽1,...,n, in matrix form as and

␣ 1␮ B ␣2
dជ ⫽ 共关 A兴 ⫹ 关 Ds 兴 兲 qជ , 共8兲 c 2⫽ 共 Jជ T pជ 兲
2 ␲ kh c tV p o w f

冋 册
where the matrix 关 A兴 is the influence matrix, do not depend on time. The physical meaning of c 1 is that it is a
decline exponent 共time⫺1 兲, and c 2 /c 1 is a limiting pressure, which
a 11 a 12 ¯ a 1n is the weighted average of the constant flowing wellbore pres-
a 21 a 22 ¯ a 2n sures, with the weights being the elements of the overall produc-
关 A兴 ⫽ , 共9兲 tivity index. If a specified flowing wellbore pressure is less than
] ]  ] c 2 /c 1 , the particular well will turn into an injector at late times,
a n1 a n2 ¯ a nn as explained in Ref. 1. By solving Eq. 18, we find that the average
reservoir pressure follows an exponential decline,

冉 冊
with elements a i j representing the influence of well j on the pres-
sure at the circumference of well i. The diagonal matrix 关 Ds兴 is c2 c2
p̄⫽ ⫹ p̄ i ⫺ exp共 ⫺c 1 t 兲 . 共19兲

冋 册
constructed from the vector of skin factors as c1 c1
s1 0 ¯ 0 In Eq. 19 we use the average value of the initial pressure,
where ‘‘initial’’ means the start of the pseudosteady-state flow
0 s2 ¯ 0 condition, and time is measured from that point. Hence, there is
关 Ds 兴 ⫽ . 共10兲
] ]  ] no need to know the initial pressure distribution, and obviously it
needs not be constant. This is an important distinction, because we
0 0 ¯ sn
may wish to use Eq. 19 for a series of pseudosteady-state flow
Eq. 8 can be solved for the unknown production rates. In ma- conditions, where it is not reasonable to assume that the pressure
trix notation this solution is given as distribution is constant at the beginning of each subsequent pseu-
dosteady state.
2 ␲ kh
共关 A兴 ⫹ 关 D兴 兲 ⫺1 dជ .
qជ ⫽ 共11兲 By substituting Eq. 19 into Eqs. 5 and 11, the production rates
␣ 1␮ B are easily obtained. All time functions will be of the following
form: a constant plus an exponential term with the same decline
From Eq. 11 we note that the matrix 关 J兴 ,
exponent. The constant will be zero for the overall production
2 ␲ kh rate. This is in accordance with the results derived in Ref. 1 using
关 J兴 ⫽ 关J 兴, 共12兲 a more rigorous approach.
␣ 1␮ B D
Our solution procedure uses the concept of pseudosteady state
with 关 JD兴 defined as for the constant flowing wellbore pressure case. There is an ap-
parent contradiction in this approach, because the pseudosteady
关 JD兴 ⫽ 共关 A兴 ⫹ 关 Ds兴 兲 ⫺1 , 共13兲 state is not possible with constant wellbore pressure 共the wellbore
can be considered as the MPI matrix, because it describes the pressures must decline at the same rate as the average reservoir
proportionality relationship between production rate and draw- pressure兲. Nevertheless, one can consider a series of time inter-
down: vals: in each interval the wellbore pressures decline by a small
value, then the production rates are reduced, and the wellbore
qជ ⫽ 关 J兴 dជ . 共14兲 pressures jump by a small value. Considering the whole process,

22 Valkó, Doublet, and Blasingame: Multiwell Productivity Index SPE Journal, Vol. 5, No. 1, March 2000
 TABLE 1– WELL-RESERVOIR PARAMETERS FOR THE
SYNTHETIC EXAMPLE IN REF. 2

Reservoir dimensions x e ⫽14,000 ft

y e ⫽7,000 ft
h ⫽100 ft

Porosity ␾ ⫽0.15

Permeability k ⫽100 md

Viscosity ␮ ⫽0.5364 cp

Initial reservoir pressure (and its average) p i ⫽3,000 psi

Formation volume factor B ⫽1,558 bbl/STB

Total compressibility c t ⫽2⫻10⫺5 psi⫺1

Well 1 starting at t ⫽0 x w 1 ⫽3,266.67 ft

x w 1 ⫽5,133.33 ft
Fig. 2–Approximation of the constant flowing wellbore pres- r w 1 ⫽0.25 ft
sure condition by a series of pseudosteady states, conceptual s 1 ⫽10
explanation. p wf ,1⫽1,000 psi

x w 2 ⫽10,266.67 ft
Well 2 starting at t ⫽0 y w 2 ⫽5,133.33 ft
the wellbore pressures ‘‘scatter’’ a little around the constant val- r w 2 ⫽0.25 ft
ues in a saw-like manner as shown schematically in Fig. 2. The s 2 ⫽0
power of the pseudosteady-state concept lies in the fact that such p wf ,2⫽1,000 psi
an approximation causes little error in the calculated production
x w 3 ⫽7,466.67 ft
rates. 共A less rigorous but more popular formulation of the same
Well 3 starting at t ⫽100 day y w 3 ⫽2,333.33 ft
fact for a single-well configuration is that ‘‘constant rate’’ and
r w 3 ⫽0.25 ft
‘‘constant pressure’’ productivity indices are essentially the
s 3 ⫽⫺3
same.兲
p wf ,3⫽1,000 psi
Visualization of the Pressure Field. Eq. 4 can be used to calcu-
late pressures at any point in the reservoir once the wellbore pro-
duction rates are known. Rates can be obtained as observations in
the case of field data, but we can also use the solution procedure
described above to generate rate profiles. In either case, Eq. 4 is retical PI, and the production rate. Reservoir heterogeneity and
used to compute the pressure distribution. The effectiveness of our nonuniform reservoir thickness will affect the results, and once
new approach to create pressure maps may find several applica- the pressure drops below the bubblepoint, gas evolution and mul-
tions. tiphase flow effects will distort the calculated values. Therefore,
for a single-well, pseudosteady-state data do not allow us to sepa-
Solving for Skin Factor. Given long-term production data for a rate phenomena related to the bulk of the reservoir from phenom-
cluster of wells, 共i.e., a time, pressure and rate stream for each ena restricted to the vicinity of the well. The idea presented here
well兲, we can analyze the performance history of each well with applies only to the multiwell scenario. In that case, it can be
respect to the other wells. By rearranging Eq. 8, we express the assumed that phenomena related to the bulk of the reservoir affect
skin factor vector as all wells similarly. Therefore, if Eq. 20 is used to create a skin

ជs ⫽ 关 Dq兴 ⫺1 冋 2 ␲ kh
␣ 1␮ B 册
ជd ⫺ 关 A兴 qជ , 共20兲
factor trace plot, comparison of the curves may reveal a potential
problem for a particular well, where one or more wells behave
differently from the other members of the cluster. One may argue

冋 册
where
q ⫺1
1 0 ¯ 0
0 q ⫺1
2 ¯ 0 originate from the interaction of neighboring wells.
关 Dq兴 ⫺1 ⫽ . 共21兲
] ]  ]
0 0 ¯ q ⫺1
n third well 共an infill兲 begins to produce. Reservoir, fluid properties,
When using Eq. 20, the average reservoir pressure required for
calculating the pressure drawdown vector can be obtained using
the cumulative production, pore volume, and total compressibility
coupled with an overall material balance. Wellbore flowing pres-
sures and production rates are assumed to be available, as well as
the permeability, thickness, viscosity, and formation volume fac-
tor. An element of the skin vector calculated from Eq. 20 is not
necessarily the mechanical skin factor of the individual well, but
rather a suitable measure of the deviation of the performance of
the well from an ideal one. Nevertheless, if plotted against time,
the ‘‘skin factor trace’’ obtained may reveal differences in the
behavior of the individual wells. At first, the rationale for such a
suggestion is not obvious. For a single well, this would mean the
calculation of the skin factor from pressure drawdown, the theo-
that production decline curves of the individual wells can serve
the same purpose. The skin factor trace technique is, however,
more appropriate because it correctly accounts for the effects that
Synthetic Example. In Ref. 2, a synthetic example is considered
for a reservoir initially drained by two wells. After 100 days a
as well as other information are presented in Table 1.
We first wish to calculate the MPI matrix and the overall pro-
ductivity index vector for the first 100 days. The influence of the
first well on its perimeter, that is, the a 11 element of the influence
matrix, is calculated using the procedure described in the Appen-
dix. The dimensionless coordinates are x wD1 ⫽0.233 333 33,
y wD1 ⫽0.366 666 67, and r wD ⫽0.000 017 857, therefore
a 11⫽a 关 0.233 351 184, 0.366 666 67, 0.233 333 33,
0.366 666 67, 0.5 兴
⫽10.6867.
Repeating the calculations for the remaining three elements we
obtain

Valkó, Doublet, and Blasingame: Multiwell Productivity Index SPE Journal, Vol. 5, No. 1, March 2000 23
 Fig. 4–Pressure distribution near the end of the first period
„100 days…, synthetic example.

Fig. 3–Semilog graph of production rates, synthetic example.

冋 4.1343
关 J兴 ⫽ 0.5011
0.3126
0.5011
8.1812
0.1163
0.3125
0.1165 .
12.846
册
关 A兴 ⫽ 冋 10.6876
0.005 88
⫺1.263 21
10.4619
册 .
The overall productivity index vector 共in field units兲 is ជ T0
⫽ 关 4.9480 8.7786 13.275兴 .
In this time interval c 1 ⫽0.008 034 day⫺1 and c 2 /c 1
Adding the skin factor matrix, ⫽1,000 psi. At 500 days the average reservoir pressure is 1,054

冋 册
psi. Our Fig. 3 shows similar production behavior to Fig. 5 of Ref.
10 0 2, except for the few days after the infill well was turned on to
关 Ds 兴 ⫽
0 0 production. This small discrepancy is expected, because our solu-
tion assumes the pseudosteady state, whereas in Ref. 2 all flow
to matrix 关A兴, yields regimes are accounted for.

冋 册
Fig. 4 is the three-dimensional visualization of the pressure
20.6876 ⫺1.263 21 field for the synthetic example at the end of the first 100 day
关 A 兴 ⫹ 关 Ds 兴 ⫽ .
⫺1.263 33 10.4619 period. The pressure field is shown as both a surface and as a
contour plot. The pressure distribution at the end of the second
After matrix inversion 共LU decomposition and back substitu- period is shown in Fig. 5. A comparison of Figs. 4 and 5 reveals
tion兲 we obtain the effect of the infill well on the pressure distribution.

冋 册
We note that, whenever a negative skin factor is involved, the
0.0487 0.005 881
关 JD兴 ⫽ 共关 A兴 ⫹ 关 Ds兴 兲 ⫺1 ⫽ . calculated near-wellbore pressure is not realistic. While this is the
0.005 881 0.096 23 known limitation of the concept of negative skin factor, rather
than the limitation of the proposed approach, it must be kept in
The MPI in STB/D/psi is mind when the calculated pressure field is interpreted.

关 J兴 ⫽ 冋 4.1266
0.4983
0.4983
8.1601
册 ,
Since this example is synthetic, the skin factor trace calculated
from our solution, shown in Fig. 6, is almost trivial. As can be
seen, the skin factor trace consists of three horizontal lines. It is
and the overall productivity index vector 共in the same units兲 is
ជ T0 ⫽ 关 4.6250 8.6584兴 .
For this time interval, c 1 ⫽0.003 95 day⫺1 and c 2 /c 1
⫽1,000 psi. At 100 days the average reservoir pressure is 2,347
psi.
The results for the period 100⭐t⭐500 days, during which
three wells are producing, are

关 A兴 ⫽ 冋 10.6876
⫺1.263 33
⫺0.491 922
⫺1.263 21
10.4619
⫺0.064 001
⫺0.491 79
⫺0.064 142
9.609 13
册
and
关 JD兴 ⫽ 共关 A兴 ⫹ 关 Ds兴 兲 ⫺1

冋
⫽ 0.005 9139
0.048 7870

0.003 6586
0.005 9133
0.096 3076
0.001 3729
0.003 6877
0.001 3747 .
0.151 594
册 Fig. 5–Pressure distribution at the end of the second period
Therefore, the MPI in STB/D/psi is given by „500 days…, synthetic example.

24 Valkó, Doublet, and Blasingame: Multiwell Productivity Index SPE Journal, Vol. 5, No. 1, March 2000
 TABLE 2– NRU „CLEAR FORK… ROCK
AND FLUID PROPERTY DATA

Reservoir Properties
Estimated net pay interval h ⫽83 ft (case 1)
h ⫽123 ft (case 2)
Average formation porosity (excluding ␾ ⫽0.061
irreducible water)
Average effective oil permeability k o ⫽0.1 md

Fluid Properties
Initial oil formation volume factor B oi ⫽1.25 RB/STB
Oil viscosity ␮ o ⫽1.5 cp
Oil compressibility c o ⫽25⫻10⫺6 psi⫺1
Total compressibility c t ⫽30⫻10⫺6 psi⫺1

Production Parameters
Initial reservoir pressure p i ⫽2,800 psia
Flowing bottomhole pressure p wf ⫽100 psia
Nominal well spacing (primary) 40 acres
Wellbore radius r w ⫽0.33 ft

Fig. 6–Skin factor trace, synthetic example.

extensive fluid level measurements兲. Because we are using 6
month average production data, we can safely assume that for all
worth comparing the skin factor behavior of Well 1 with the pro- cases p w f is approximately 100 psia. The rock and fluid property
duction decline of the same well shown in Fig. 3. There is a data for the NRU cases are shown in Table 2.
change in the rate of decline of production at 100 days due to the The average reservoir pressure trend for each of the study areas
appearance of the infill well. However, this behavior is not the was estimated based on the initial reservoir pressure and numer-
result of the deterioration of the near-well conditions and indeed, ous shut-in bottomhole pressure data recorded at the end of the
it is ‘‘filtered out’’ from the skin factor trace by Eq. 20. primary producing period 共June 1987兲.
Since these wells produce from a rather large gross reservoir
Analysis of Field Data interval 共u1,000 ft兲, it is often difficult to determine exactly how
Because of the importance of this potential application, our field much of the gross interval actually contributes to production. By
examples will focus on the analysis and interpretation of well using the average reservoir pressure as an input variable, we de-
performance. We consider two study areas from the North Rob- termined the net thickness of the study area from material balance.
ertson unit 共NRU兲 in Gaines County, Texas. The North Robertson We also note that these resulting net thickness values correlated
共Clear Fork兲 Field was developed on a nominal 40-acre well spac- well with the ‘‘reservoir quality’’ defined by previous geologic
ing beginning in 1956. The dominant reservoir producing mecha- studies.
nism for the original 141 wells was solution-gas drive. The initial Field Example 1. The first field example consists of five wells in
reservoir pressure in the Lower Clear Fork was estimated to be the south-central portion of the NRU shown in Fig. 7. This area is
2,800 psia. The original oil in place was estimated to be approxi- considered to be of moderate to poor reservoir quality. Using the
mately 215 MMSTB from decline curve analysis, with primary average reservoir pressure at the end of primary production as an
production prior to the implementation of a field-wide waterflood input parameter, we obtained an effective net pay thickness of 83
共1987兲 of 17.5 MMSTB. Individual well primary recovery factors ft. The calculated original oil in place from the multiwell tech-
are low, ranging between 5 and 10%. nique was 10.5 MMSTB, while the sum of the oil-in-place values
The Clear Fork formation is a shallow-shelf carbonate com- for the individual wells from decline curve analyses yielded 6.7
posed primarily of a massive dolomite section with varying de-
grees of anhydrite cement. The geologic setting at the time of
deposition and subsequent diagenesis contributed to the heteroge-
neous nature of the formation. The wells were initially completed
in Lower, Middle, and Upper Clear Fork, at measured depths of
between 6,200 and 7,200 ft. The majority of the original comple-
tion intervals were in the Lower Clear Fork, which is considered
the main pay interval. All the wells in the North Robertson unit
were eventually hydraulically fractured, usually in two stages, a
Lower/Middle Clear Fork fracture stimulation and a subsequent
Upper Clear Fork fracture stimulation.
The oil flow rate data were allocated to individual wells on a
tract basis. The wells were tested 共for allocation purposes兲 on a
semiannual basis. For that reason, we choose to average the pro-
duction data over 6 month intervals, although the analysis could
easily be done on a monthly or even hourly basis, if such data
were available.
Although there are no flowing bottomhole pressure data avail-
able for the North Robertson wells, almost all the wells were
placed on rod pump very soon after their initial completion. These
wells typically ‘‘pump off’’ in anywhere from 2 to 12 weeks,
depending on the flow capacity of the reservoir surrounding the
individual wellbores, at which time the bottomhole producing
pressure can be assumed to be approximately 100 psia 共based on Fig. 7–Well configuration, field example No. 1.

Valkó, Doublet, and Blasingame: Multiwell Productivity Index SPE Journal, Vol. 5, No. 1, March 2000 25
 Fig. 8–Individual well oil production rates and average reservoir pressure trend, field example No. 1.
MMSTB. The discrepancy between the two values is likely a
function of the reservoir area we used in our analysis and the
result of the reservoir heterogeneity that exists within the Clear
Fork interval. That is, each individual well does not completely
contact all the oil within its drainage area. We would expect that
in a less heterogeneous system these values would be much closer
to one another and, in fact, the difference in the results between
the two analysis techniques could be used to define the degree of
reservoir heterogeneity.
The oil rates for individual wells and the calculated average
reservoir pressure trend for the study area are shown in Fig. 8.
The calculated skin factor traces for all wells are shown in Fig. 9.
As expected, the multiwell skin factors tend to increase with time.
The explanation for sharp decreases could be always found from
well records and were the result of individual well remediations.
The results for each individual well are shown in Figs. 10
through 14, where we have annotated these figures to indicate
workovers and recompletions. The transient 共early-time兲 skin fac-
tors from individual well decline curve analyses are shown for
each well for comparison purposes 共these are given as horizontal
lines兲. We note good agreement between the multiwell skin factor
trace and the decline curve skin factor, e.g., for early-time data in
Fig. 9–Multiwell skin factor trace, field example No. 1.
26 Valkó, Doublet, and Blasingame: Multiwell Productivity Index
Fig. 10, for all times in Fig. 12, for late time data in Fig. 13, and
for intermediate times in Fig. 14. For Fig. 11 the agreement is
poor, indicating that the pseudosteady-state and transient analyses
do not always agree. The ability of the skin factor trace to signal
well problems is obvious for two wells, NRU 3003 and NRU
3006. These wells showed a more intense performance deteriora-
tion than the other wells in the cluster, and it is not by chance that
those two wells were abandoned before the end of the primary
production period. Looking only at the oil production rates 共Fig.
8兲, the poor performance of those two wells becomes obvious
much later than from Fig. 9. The pressure distribution at 6,000
days 共Fig. 15兲 shows that, because of the high skin factors of the
two problematic wells, the northern part of the study area was not
being effectively produced at that time.
Field Example 2. Our second field example consists of five wells
in the southeastern portion of the NRU 共Fig. 16兲 in an area of
moderate to high reservoir quality. Using the average reservoir
pressure at the end of primary production as an input parameter,
we obtained an effective net pay thickness of 123 ft. The calcu-
lated original oil in place from the multiwell technique was 16.8
MMSTB, while the sum of the oil-in-place values for the indi-
SPE Journal, Vol. 5, No. 1, March 2000
 Fig. 10–Oil rate and skin factor trace, Well NRU 3003, field example No. 1.

Fig. 11–Oil rate and skin factor trace, Well NRU 3006, field example No. 1.

Fig. 12–Oil rate and skin factor trace, Well NRU 3007, field example No. 1.

Valkó, Doublet, and Blasingame: Multiwell Productivity Index SPE Journal, Vol. 5, No. 1, March 2000 27
 Fig. 13–Oil rate and skin factor trace, Well NRU 3008, field example No. 1.

Fig. 14–Oil rate and skin factor trace, Well NRU 1702, field example No. 1.

Fig. 15–Pressure distribution at 6,000 days, field example

No. 1. Fig. 16–Well configuration, field example No. 2.

28 Valkó, Doublet, and Blasingame: Multiwell Productivity Index SPE Journal, Vol. 5, No. 1, March 2000
 Fig. 17–Individual well oil production rates and average reservoir pressure trend, field example No. 2.

Fig. 18–Multiwell skin factor traces, field example No. 2.

vidual wells from decline curve analyses was 11.6 MMSTB. The
oil rates for the individual wells and the calculated average reser-
voir pressure for the study area are shown in Fig. 17. The calcu-
lated multiwell skin factor trace, shown in Fig. 18, indicates high
completion efficiencies and only moderate or no damage evolu-
tion. Any significant improvement in the skin factor trace could be
identified as a direct result of individual well workovers. The
pressure distribution at 6,000 days 共Fig. 19兲 shows that the wells
depleted the cluster in an essentially balanced manner.

Conclusions
The multiwell productivity index provides a useful analytical tool
which expands the capabilities of the reservoir/production engi-
neer. Approximate analytical solutions can be easily generated, as
can a pressure map for a cluster of wells. Based on the concept of
the multiwell productivity index, we introduced a technique to
analyze the performance behavior of a cluster of wells. The ap-
proach is best suited to analyze a moderate number of wells in
Fig. 19–Pressure distribution at 6,000 days, field example areas that can be described by fairly homogeneous reservoir prop-
No. 2. erties. Therefore, the scope of the proposed technique lies be-

Valkó, Doublet, and Blasingame: Multiwell Productivity Index SPE Journal, Vol. 5, No. 1, March 2000 29
tween production decline curve analysis of individual wells and Subscripts
large-scale numerical flow simulation of entire fields. In the first e ⫽ boundary value
field case we demonstrated that the multiwell skin factor trace i ⫽ initial value
may allow early detection of well performance problems at a time i,j ⫽ well index
when visual examination of the production decline plot might yet m ⫽ index in the summation
be inconclusive. The analysis of the second field case concluded T ⫽ total
that within the particular study area the wells performed uni-
formly. Based on the field examples, we suggest that the tech-
nique can be incorporated into real time reservoir surveillance Superscript
systems. T ⫽ transposed
Nomenclature
Acknowledgments
a关 兴 ⫽ influence function
We acknowledge the permission to publish field data provided by
B ⫽ formation volume factor, RB/STB
Fina Oil and Chemical, Co. 共Western Division, USA兲, as well as
B0 ⫽ oil formation volume factor, RB/STB
the financial support of the U.S. Department of Energy 共DOE兲 for
CA ⫽ Dietz shape factor 共single well兲
funding provided through the DOE Class II Oil Program. We also
c1 ⫽ decline constant, t⫺1, day⫺1
wish to acknowledge the technical and computing support ser-
c2 ⫽ arbitrary constant, m/共Lt3兲, psi/day
vices provided by the Department of Petroleum Engineering at
co ⫽ oil compressibility, Lt2/m, psi⫺1
Texas A&M U.
ct ⫽ total system compressibility, Lt2/m, psi⫺1
h ⫽ formation thickness, L, ft
J ⫽ productivity index, L4t/m, 共STB/D兲/psi References
JD ⫽ dimensionless productivity index 1. Rodriguez, F. and Cinco-Ley, H.: ‘‘A New Model for Production
k ⫽ effective permeability, L2, md Decline,’’ paper SPE 25480 presented at the 1993 SPE Productions
n ⫽ number of wells Operation Symposium, Oklahoma City, Oklahoma 21–23 March.
N ⫽ number of terms in the summation 2. Camacho-V, R. et al.: ‘‘Optimum Position for Wells Producing at
Constant Wellbore Pressure,’’ SPEJ 共June 1996兲 155.
p ⫽ pressure, m/共Lt2兲, psi
3. Slider, H.C.: Worldwide Practical Petroleum Reservoir Engineering
p̄ ⫽ material balance average reservoir pressure, Methods, PennWell Books, Tulsa, Oklahoma 共1983兲.
m/共Lt2兲, psi 4. Helmy, M.W. and Wattenbarger, R.A.: ‘‘New Shape Factors for
p̄ i ⫽ initial material balance average reservoir pres- Wells Produced at Constant Pressure,’’ paper SPE 39970 presented at
sure, m/共Lt2兲, psi the 1998 SPE Gas Technology Symposium, Calgary, 15–18 March.
Pwf ⫽ flowing bottomhole pressure, m/共Lt2兲, psi 5. Ozkan, E.: ‘‘Performance of Horizontal Wells,’’ PhD dissertation, U.
q ⫽ oil flow rate, L3/t, STB/D of Tulsa, Tulsa, Oklahoma 共1988兲.
re ⫽ drainage radius for a single well, L, ft 6. Raghavan, R.: Well Test Analysis, PTR Prentice-Hall Inc., Englewood
rw ⫽ wellbore radius, L, ft Cliffs, New Jersey 共1993兲.
7. Dietz, D.N.: ‘‘Determination of Average Reservoir Pressure From
s ⫽ skin factor
Buildup Surveys,’’ JPT 共August 1995兲 955; Trans., AIME, 234.
S T ,S 1 ,S 2 ,S 3 ⫽ terms in the summation 8. Larsen, L.: ‘‘A Simple Approach to Pressure Distributions in Geo-
t ⫽ time, t, days metric Shapes,’’ SPEJ 共February 1985兲 113.
xe ⫽ size of study area in the x direction, L, ft
ye ⫽ size of study area in the y direction, L, ft
xw ⫽ individual well x coordinate, L, ft Appendix: Calculation of the Influence Function
yw ⫽ individual well y coordinate, L, ft The procedure required to calculate the influence function is a
xD ⫽ x/x e , dimensionless x-coordinate well location slightly modified version of the procedure proposed in Ref. 5.
yD ⫽ y/x e , dimensionless y-coordinate well location 1. Making use of the symmetry of this problem, we calculate
Vp ⫽ pore volume, L3, ft3
a 关 x D ,y D ,x wD ,y wD ,y eD 兴 ⫽a 1 关 max共 x D ,x wD 兲 , max共 y D ,y wD 兲 ,
Greek Letters min共 x D ,x wD 兲 , min共 y D ,y wD 兲 ,y eD 兴 .
␣ 1 ⫽ conversion factor 共for field units⫽887.22, for con- 共A-1兲
sistent system⫽1兲
␣ 2 ⫽ conversion factor 共for field units⫽5.615, for con- 2. Further use of symmetry allows us to exchange the role of x
sistent system⫽1兲 and y 共if necessary兲:
␾ ⫽ porosity, fraction a 1 关 x D ,y D ,x wD ,y wD ,y eD 兴
␮ ⫽ fluid viscosity, m/共Lt兲, cp
␥ ⫽ Euler’s constant 共0.577 215 66...兲
⑀ ⫽ small step, dimensionless ⫽ 再 a 0 关 x D ,y D ,x wD ,y wD ,y eD 兴
a 0 关 y D ,x D ,y wD ,x wD ,1/y eD 兴
if 兩 x D ⫺x wD 兩 ⬎ 兩 y D ⫺y wD 兩 ,
otherwise.
Vectors 共A-2兲
0
dជ ⫽ drawdown vector, m/共Lt2兲, psi 3. The function a is defined by Eq. 4 as
ជj 0 ⫽ overall productivity index vector, L4t/m, 共STB/D兲/
psi
qជ ⫽ production rate vector, L3/t, STB/D
a 关 x D ,y D ,x wD ,y eD 兴 ⫽2 ␲ y eD 冉 1 yD yD
⫺
3 y eD
⫹
2
⫹y wD
2
2y eD
2

冊
⫹S T ,

pជ w f ⫽ vector of flowing wellbore pressures, m/共Lt2兲, psi 共A-3兲

where
Matrices ⬁
tm
关 A兴 ⫽ influence matrix, dimensionless
关 Ds兴 ⫽ diagonal matrix of skin factors, dimensionless
S T ⫽2 兺
m⫽1 m
cos共 m ␲ x D 兲 cos共 m ␲ x wD 兲 , 共A-4兲

关 Dq兴 ⫽ diagonal matrix of surface production rates, L3/t, and

STB/D
cosh关m␲共yeD⫺兩yD⫺ywD兩兲兴⫹cosh兵m␲关yeD⫺共yD⫹ywD兲兴其
关 J兴 ⫽ productivity index matrix, L4t/m, 共STB/D兲/psi tm⫽ . 共A-5兲
关 JD兴 ⫽ dimensionless productivity index matrix sinh共m␲yeD兲

30 Valkó, Doublet, and Blasingame: Multiwell Productivity Index SPE Journal, Vol. 5, No. 1, March 2000
 The infinite sum in Eq. A-4 can be replaced by a finite approxi- For example, for x eD ⫽1, x wD ⫽0.5, and y wD ⫽0.5, our algo-
mation consisting of three parts: rithm yields ln CA⫽3.430 14 and C A ⫽30.8811 which is in excel-
lent agreement with the known value 共30.88兲. While our proce-
S T ⫽S 1 ⫹S 2 ⫹S 3 , 共A-6兲
dure does not provide new results relative to the generally used
where algorithm of Larsen,8 it is simpler in the sense that it does not
N require numerical evaluation of the exponential integral function.
tm
S 1 ⫽2 兺
m⫽1 m
cos关 m ␲ x D 兴 cos关 m ␲ x wD 兴 , 共A-7兲
SI Metric Conversion Factors
tN ⫻ 4.046 873
S 2 ⫽⫺ ln兵 关 1⫺cos共 ␲ 共 x D ⫹x wD 兲兲兴 2 ⫹ 关 sin共 ␲ 共 x D ⫹x wD 兲兲兴 2 其 acre E⫺03 ⫽ m2
2 bbl ⫻ 1.589 873 E⫺01 ⫽ m3
tN cp ⫻ 1.0* E⫺03 ⫽ Pa•s
⫺ ln兵 关 1⫺cos共 ␲ 共 x D ⫺x wD 兲兲兴 2 ⫹ 关 sin共 ␲ 共 x D ⫺x wD 兲兲兴 2 其 , ft ⫻ 3.048* E⫺01 ⫽ m
2
md ⫻ 9.869 233 E⫺16 ⫽ m2
共A-8兲
psi ⫻ 6.894 757 E⫹03 ⫽ Pa
and psi ⫻ 1.450 377 E⫺04 ⫽ Pa⫺1
N
1
兺
*Conversion factors are exact. SPEJ
S 3 ⫽⫺2t N cos关 m ␲ x D 兴 cos关 m ␲ x wD 兴 . 共A-9兲
m⫽1 m
Peter P. Valkó is a visiting associate professor in the Dept. of
The first part is the usual finite approximation stopping after the
Petroleum Engineering at Texas A&M U., College Station,
Nth term in the summation. The second and third parts are ob-
Texas. e-mail: p-vaklo@tamu.edu. His primary research interest
tained from the identity, is in stimulation of hydrocarbon-producing wells. He previously
⬁ worked at Mining U., Leoben, Austria; for the Hungarian Oil
1
兺
m⫽1 m
cos关 m ␲ x D 兴 cos关 m ␲ x wD 兴 Co. (MOL); and at Eötvös U., Budapest, Hungary. Valkó holds
an MS degree in chemical engineering from Veszprém U.,
Hungary, and a PhD degree from Novosibirsk Inst. of Catalysis,
1 Russia. A member of the Editorial Review Committee, he also
⫽⫺ ln兵 关 1⫺cos共 ␲ 共 x D ⫹x wD 兲兲兴 2 ⫹ 关 sin共 ␲ 共 x D ⫹x wD 兲兲兴 2 其 served as a 1998–99 member of the Forum Series in North
4
America Steering Committee and on Annual Meeting techni-
1 cal committees. Louis E. Doublet is a associate research en-
⫺ ln兵 关 1⫺cos共 ␲ 共 x D ⫺x wD 兲兲兴 2 ⫹ 关 sin共 ␲ 共 x D ⫺x wD 兲兲兴 2 其 , gineer and PhD-degree candidate in the Dept. of Petroleum
4
Engineering at Texas A&M U. Previously, he worked for Dresser
共A-10兲 Atlas Oilfield Services, Marathon Oil Co., and Halliburton Log-
ging Services. Doublet holds a BS degree in mechanical engi-
and from the fact that t m alone converges ‘‘fast.’’ The advantage
neering from Georgia Inst. of Technology and an MS degree
of this algorithm is that only a few hyperbolic functions have to be in petroleum engineering from Texas A&M U. Thomas A. Blas-
evaluated since the number of terms, N, is usually less than 100. ingame is an associate professor in the Dept. of Engineering
As a byproduct of this work, we obtain a convenient algorithm at Texas A&M U. His research interests include identification of
to calculate the Dietz7 shape factors. Traditionally, when the Dietz methods for evaluation and prediction of gas reservoir perfor-
shape factors are presented the dimensionless distances are based mance, field-scale programs in reservoir description and man-
on y e , therefore y eD ⫽1 and x eD is greater than or equal to unity. agement, and development and application of methods for
Based on the relation derived by Ozkan,5 the shape factor is ex- analysis and interpretation of well-test and production data.
pressed as Blasingame holds BS, MS, and PhD degrees in petroleum engi-
neering from Texas A&M U. He is a member of the Monograph
4 ⑀ Committee and of the Colloquium on Petroleum Education
ln C A ⫽ln ⫺ ␥ ⫺2 ln
x eD x eD Engineering Committee. He was 1993–94 Texas A&M Student

冋 册
Chapter Faculty Sponsor, a 1991–94 member of the Editorial
x wD ⫹ ⑀ y wD x wD y wD 1 Review Committee, a 1995–96 member of the Forum Series in
⫺2a , , , , , 共A-11兲 North America Steering Committee, and a 1996–99 member
x eD x eD x eD x eD x eD
of the Career Guidance Committee. He has also served as a
where the Euler constant is ␥ ⫽0.577 215 66,..., and ⑀ is a suitable member and as chair of Annual Meeting technical commit-
small positive number 共e.g., 10⫺6 兲. tees.

Valkó, Doublet, and Blasingame: Multiwell Productivity Index SPE Journal, Vol. 5, No. 1, March 2000 31