Sie sind auf Seite 1von 6

Reservoir Simulation

Calvin C. Mattax, * SPE, and Robert L. Dalton SPE


Exxon Production Research Co. ' ,

u •• • the
precision with Introduction meability and viscosity), and rock and fluid
In the past 30 years, reservoir simulation has compressibilities.
which variations In Fig. 1 illustrates the most common types
evolved from a research area into one of the
reservoir properties most flexible and widely used tools in reser- of models used in simulation. Models range
can be modeled Is voir engineering. Use of reservoir simula- in complexity from a single block, useful
tion has grown because of its ability to only for classical material-balance calcula-
determined by the predict the future performance of oil and gas tions, to fully 3D models capable of model-
number of blocks In reservoirs over a wide range of operating ing all major factors that influence reservoir
conditions. Reservoir simulators use numer- performance.
the model." Each gridblock in a model has only one
ical methods and high-speed computers to
model multidimensional fluid flow in reser- set of properties; there is no variation in any
voir rock. Reliable simulators and adequate property within a block. For example, phase
computing capacity are available to most saturations in a model block will be volu-
reservoir engineers, so simulation is usually metric averages of the saturations in that part
practical for all reservoir sizes and all types of the reservoir represented by the block.
of reservoir performance studies. Although In this respect, a model block can be visual-
the use of simulation frequently is option- ized as a well-stirred tank (Le., its contents
al, it may be the only reliable way to predict are homogeneous) connected to adjacent
the performance of a large, complex reser- tanks with pipes whose flow capacities are
voir, especially if such external considera- determined by reservoir flow properties.
tions as government regulations influence This visualization, although simplistic, dem-
the production schedule. Even for small onstrates that the precision with which var-
reservoirs where simple calculations or ex- iations in reservoir properties can be
trapolations may be adequate, simulation is modeled is determined by the number of
often faster, cheaper, and more reliable than blocks in the model.
alternative methods for predicting per- A simulator also divides the life of a reser-
formance. voir into discrete increments. Changes in a
reservoir (pressure, saturation, etc.) are
Modeling Concepts computed over each of many time incre-
ments, or timesteps. Conditions are defined
A reservoir simulator models a reservoir as only at the beginning and end of each
if it were divided into a number of individual timestep; nothing is defined at any inter-
blocks (gridblocks). Each block corresponds mediate time within a time interval. The ac-
to a designated location in the reservoir and curacy with which reservoir behavior can
is assigned properties-porosity, permeabil- be calculated generally will be influenced
ity, relative permeability, etc.-believed to by the length of the timesteps as well as the
be representative of the reservoir at that number of gridblocks.
location. In the simulator, fluids can flow The preceding discussion implies that for
between neighboring blocks at a rate deter- any size gridblock and any length of
mined by pressure differences between timestep, there will always be abrupt
blocks and flow properties assigned to the changes in reservoir conditions from one
interfaces between blocks. In essence, the block to the next and from one timestep to
mathematical problem is reduced to a cal- the next. Fig, 2 illustrates this point. At the
culation of flow between adjacent blocks. top of Fig. 2 are plan views of a hypothetical
For every block-to-block interface, a set of two-well reservoir being waterflooded and a
equations must be solved to calculate the four-gridblock simulator model of the reser-
flow of all mobile phases. The equations voir. The two plots show water-saturation
generally incorporate Darcy's law and the distribution at a given time in both the reser-
concept of material balance and contain voir and the model. In the reservoir, water
terms describing the permeability "be- saturation is a smooth function of distance
tween" blocks, fluid mobilities (relative per- but in the model, an abrupt drop in wate;
'Now retired. saturation occurs between Blocks 1 and 2
Copyright 1990 Society of Petroleum Engineers and again between Blocks 2 and 3. A plot
692 June 1990 • JPT
sPETechnology
Today SERIES

"The most useful way


to determine whether a
model adequately
describes a reservoir Is
to simulate past
performance and to
a b c compare the simulation
with the reservoir's
actual performance
history."
-
/'/'/'/LL-;;

./" (
/
./" .J.x
/"

~
-r--'---

'- 9

Fig. 1-Types of models used In reservoir simulation: (a) tank (material balance), (b) 1D
linear, (c) 1D radial, (d) 2D cross-sectional, (e) 2D areal, (f) radial cross-sectional, (g) 3D.

of saturation vs. time in any block in the lations. Increasing the number of blocks and best. Although study results may be more
model would also be a "stair-step" curve, timesteps will increase the time required to credible to decision makers if the model is
whereas a time/saturation plot for the cor- prepare input data and to interpret results more complex than needed to solve the prob-
responding location in the reservoir would and the cost of calculations. lem, added intricacy almost always increases
be a continuous, smooth curve. The stair- Other important model design steps are the cost of a study.
step approximation approaches actual reser- selecting model dimensions (10, 2D, or 3D) Usually, models are 2D or 3D and con-
voir saturation distribution as the sizes of and determining whether conventional tain from 100 (for a simple 2D model) to
gridblocks and timesteps decrease. black-oil, compositional, miscible, chemi- 100,000 gridblocks (for a very complex 3D
An important step in simulation clearly is cal, or thermal treatment is needed. Design- model) . Small models can be run on a per-
the selection of the number of blocks in a ing the simplest model that will simulate the sonal computer or a standard mainframe
model and the timestep size to use in calcu- displacement process adequately is usually computer, while the most complex ones re-

JPT • June 1990 693


"For conventional
primary and secondary
recovery processes, G_'_V Reservoir S i mulator Model
o

the ability of reservoir 75 75F-_--,

simulators to predict
performance Is 50 50

determined largely by
the quality of the 25 25

reservoir and aquifer


o~----~----~--~~--~ o~----~----~--~~--~
description and the Distance 4

a b Block Number

model design."
Fig. 2-A four-gridblock waterflood model: (a) hypothetical reservoir and Its water-
saturation distribution at a given time In the waterflood and (b) four-grldblock model
and simulated water-saturation distribution.

quire a supercomputer. Computer time and major weaknesses in reservoir description cance of major factors that influence per-
costs for one simulation of reservoir per- or operating problems, such as casing leaks. formance. For example, results of a 3D,
formance can vary from a few minutes, cost- Experience has shown that simulation can full-field reservoir simulation study of the
ing less than $100, to many hours, costing be a powerful reservoir-description tool. Lower Brent reservoir of the Dunlin field
thousands of dollars. in the U.K. North Sea by Exxon Produc-
Simulation Applications tion Research Co. demonstrated that faults
Testing Model Validity Reservoir simulators are most frequently dominate areal sweep and that vertical per-
The most useful way to determine whether used to predict future production from an meability distribution controls vertical
a model adequately describes a reservoir is entire reservoir or a major segment of a sweep. As Fig. 3 shows, injected water has
to simulate 'past performance and to com- reservoir. Such full-field simulators normal- a tendency to "tongue" through the high-
pare the simulation with the reservoir's ac- ly predict oil, gas, and water production quality Etive sands located approximately
tual performance history. If the comparison rates as a function of time from individual midreservoir. The study identified major tar-
is favorable, the model can be used with wells and from the total area modeled. In gets for additional recovery (the Rannoch
some confidence to predict the future. The addition, the simulator may integrate well- sands in the lower half of the section and
simulation should match typical well histo- bore constraints into predicted future per- the Lower Ness sands at the top of the sec-
ries as well as regional and overall reservoir formance by automatically implementing a tion). On the basis of a similar 3D simulation
performance. Pressure, saturation, GOR, logical sequence of well workovers, recom- and development study, 2 the field operator
and WOR should all match to within the pre- pletions, replacements, and additions (infill has already undertaken a program of dedi-
cision needed to attain the objectives of the drilling) to optimize economics. Several pre- cated Rannoch completions to improve
reservoir study. dictions will frequently be made to reflect recovery from this unit.
Usually, not enough is known about a different possible operating conditions or A second interesting application of a full-
reservoir to construct a model and to use it two or more equally probable reservoir field simulator is the modeling 3 of the Troll
without some modification of permeability , descriptions . SPE Monograph Vol. 13, field offshore Norway. Numerical simula-
porosity, continuity, and stratification. An Reservoir Simulation, 1 discusses examples tion is the only practical way to study a
unexpected benefit of simulation sometimes of this type of simulation. large, complex field such as Troll. Note the
comes from discrepancies between actual Simulation can be especially useful when major faulting illustrated in the cross section
and simulated performance that identify uncertainty exists about the relative signifi- of the field (Fig. 4). The simulator used

Fig. 3-Water-saturatlon distribution In


the Dunlin, Lower Brent model
after 9 years of water injection. WATER-INVADED RE&IION~

••
WATER - -- - •
OIL

694 June 1990 • JPT


31/2-13 31/2-2 31/3-1
Authors
31/2-10 W ••----,t--+> E

':J 1400
(j)
::;
S 1600

;:: 1800
Q.

~ 2000

o SOGNEFJORO
Calvin C. MaUax, retired division
• HEATHER UNIT B

o FENSFJORO
manager from Exxon Production Re-
search, is a consultant in Houston. He
holds a BS degree from the U. of Tulsa
and a PhD degree from Louisiana State
Fig. 4-Westleast cross section, Troll field.
U., both In chemistry. Mattax is editor
and coauthor of the newest SPE Mono-
several special techniques to model all sig- an excessive number of gridblocks would be graph, Reservoir Simulation, and is a
nificant aspects of reservoir fluid dynamics, needed. The influence of such small-sized member of a Technical Program Com-
including flow through the faults and water heterogeneities should be included in a mittee for the 1990 Annual Meeting.
and gas coning. Fig. 5 shows potential pro- model through combined geologic and reser- Robert L. Dalton, division manager at
duction profiles for two of several possible voir engineering studies that define the ef- Exxon Production Research Co. in
development plans. Engineers indicated that fective reservoir properties needed when a Houston, is editor and coauthor of
the model proved to be an efficient tool for Reservoir Simulation. He holds a BS
practical number of blocks is used. For ex- degree in chemical engineering from
comparing alternative development scenar- ample, Richardson et al. 5 derived effective Rice U. Dalton served on Technical Pro-
ios and for general reservoir management. vertical permeabilities .that would allow gram committees for the 1974 and 1976
Single-well models can be used to study small, discontinuous shales to be modeled Annual Meetings and was a 1984-85
flow at the sandface and in the region near accurately in simulations of oil drainage with Distinguished Lecturer. He has been a
a well. These models are usually radial, and gridblocks larger than the shales. member of the Editorial Review Commit-
gridblocks adjacent to the wellbore may be tee since 1987 and is a 1989-90 Con-
For conventional primary and secondary tinuing Education Committee member.
as small as 1 or 2 in. in radius. A radial recovery processes, the ability of reservoir
cross-sectional model similar to that shown simulators to predict performance is deter-
in Fig. I was used by Addington 4 to pre- mined largely by the quality of the reservoir voir study and to control personnel and com-
dict gas-coning behavior of Prudhoe Bay and aquifer description and the model de- puter costs.
wells for a range of reservoir properties and sign. Our understanding of EOR processes 3. History matching, which is a vital part
perforation thicknesses and locations in the
is less advanced. These processes pose of a simulation study, can be a powerful
producing interval. With the model results,
modeling difficulties, and industry has less reservoir-description tool.
gas-coning correlations were developed for
experience in modeling them, as discussed
use in a 3D Prudhoe Bay field model to
in more detail in Reservoir Simulat{on. 1 References
predict critical coning rates and GOR's of
wells after cone arrival. 1. Mattax, C.C. and Dalton, R.L.: Reservoir
Major heterogeneities usually must be Summary Simulation, Monograph Series, SPE, Richard-
represented in a reservoir model. Detailed 1. Reservoir simulation is an effective son, TX (1990) 13.
technique that is now widely available for 2. Braithwaite, C.I.M. et a!.: "Improving Recov-
representation of small-sized (a few inches to
ery From the Dunlin Field, U.K. Northern
a few feet) reservoir heterogeneities seen in use in reservoir engineering.
North Sea, " paper SPE 19878 presented at the
logs, cores, and outcrops, however, usually 2. Careful design of a simulation model 1989 SPE Annual Technical Conference and
is not practical in simulation models because is required to meet the objectives of a reser- Exhibition, San Antonio, Oct. 8-11.
3. Kydland, T. et a!.: "Application of Uncon-
ventional Techniques in Constructing an In-
tegrated Reservoir Simulation of the Troll
8.000
Field," SPERE (Aug. 1988) 967-76.
- - 01.. START-UP TROll EAST
4. Addington, D.V.: "An Approach to Gas-Con-
1.000 - 01.., STAAT-UP TROlL WEST
ing Correlations for a Large Grid Cell Reser-
- - WATER. START-UP. TROLL EAST
voir Simulator," JPT (Nov. 1981) 2267-74.
! 6,000 - - - WATER. STAAT-UP TROLl WEST
5. Richardson, G.J. eta!.: "The Effect of Small,
E
~ 5,000
Discontinuous Shales on Oil Recovery," JPT
(Nov. 1978) 1531-37.
I
w
....
4.000
51 Metric Conversion Factors
: 3.000 r:::.----------- bbl x 1.589 873 E-Ol = m'

2.000
;f """-------------------------------- ft x. 3.048"
in. x 2.54"
E-Ol = m
E+OO = em
1.000
I • Conversion factor is exact.

This paper is SPE 20399. Technology Today Series arti·


cles provide useful summary information on both classic and
emerging concepts in pelroleum engineering. Purpose: To
TIME (years> provide the general reader with a basic understanding of
a significant concept, technique, or developmenl within a
specific area of technology.

Fig. 5-Productlon profiles, southern oil province, Troll field. JPT


JPT • June 1990 695
Discussion of Reservoir Simulation
Leendert Schenk, SPE, independent

I read with interest the Technology Today "For example, results of a 3D, full-field a. For example, how does it jibe with
Series article "Reservoir Simulation" by reservoir study of the Lower Brent reser- Point 2 of the Summary: "Careful
Mattax (retired) and Dalton of Exxon Pro- voir of the Dunlin field in the U .K . North design of a simulation mod~
duction Research Co ., published in the June Sea by Exxon Production Research Co . quired to meet the objectives of a
1990 JPT (Pages 692-95). I have the fol- demonstrates that faults dominate areal re'servoir study and to control per-
lowing comments to make. sweep and that vertical permeability distri- sonnel and computer costs."
1. Am I really to accept the statement at bution controls vertical sweep. As Fig. 3 b . Does the statement hold only for the
the end of the Introduction: "Even for small shows, injected water has the tendency to Exxon corporate culture or should it
reservoirs where simple calculations or ex- tongue through the high-quality Etive also extend to outside-Exxon deci-
trapolations may be adequate, simulation is sands. . . ." Does one really need simula- sion makers-e.g., those in compa-
often faster, cheaper, and more reliable [my tors to draw the underlined conclusions? Or nies involved with joint ventures or
underlining] . .. "? Too bad that the article is it sometimes a relief to find that simula- in government agencies?
does not offer any proof. tors can predict the obvious? Regarding the c. Considering that " . .. added intri-
2. In red at the top of Page 694 is the Troll field offshore Norway, the paper says: cacy almost always increases the
statement, "For conventional primary and "Engineers indicated that the model proved cost. . . ," what increase is consid-
secondary recovery processes, the ability of to be an efficient tool for comparing alter- ered advisable or permissible? May-
reservoir simulation to predict the perform- native development scenarios and for gener- be 25 or 50% ?
ance is determined largely by the quality of al reservoir management." This sounds 5. Finally, I found it a great relief to read
the reservoir and aquifer description. .. ." more like an article of faith* than a techni- in the Introduction that" . .. the use of simu-
The last sentence of Testing Model Validi- cal argument. lation frequently is optional . . . ." In con-
ty says: "Experience has shown that simu- 4. I was flabbergasted by a statement in clusion, I am sad to have to say that I was
lation can be a powerful reservoir-descrip- the Modeling Concepts section at the bot- disappointed to encounter the article in a
tion tool. " What a beautiful vicious circle: tom of Page 693: "Although study results journal of the caliber of the JPT, particu-
the simulator itself can "predict" what it may be more credible to decision ma"kefsif larly in the Technology Today Series.
needs to make predictions! the model is more complex than needed to
3. The examples given in the Simulation solve the problem, added intricacy almost (SPE 21606) JPT
Applications section do not look very con- always increases the cost of the study."
vincing for the needs of reservoir simula- What a concept this is! It gives rise to many • The Devi/'s Dictionary by Ambrose Bierce gives this defi·
tors . The third column on Page 694 says: questions . nition: " Faith-belief without evidence . "

JPT • November 1990 1447


Authors' Reply to Discussion of
Reservoir Simulation
Calvin C. Mattax,* SPE, and'Robert L. Dalton, SPE,
Exxon Production Research Co.

We wish to emphasize that simulation is a water influx. The engineer found, however, understanding of the reservoir. What a simu-
reservoir-management tool and only a tool; that she could use a very simple, 1D model lator provides is quantitative results that
it should not be used as a substitute for en- to simulate performance and to match the forecast timing of future events and pro-
gineering knowledge and skill. The quality 5, 100-ft -sand pressure history. duced and injected fluid volumes. These
of the results of any simulation and the ease The model was used to estimate an op- types of forecasts can provide invaluable in-
with which those results can be obtained de- timal depletion rate and to demonstrate that formation for planning depletion and de-
pend entirely on the skill and judgment of reserves would be lost at lower rates. As a veloping operating strategies.
the engineer using the simulator. ** It is with consequence, a special allowable was grant- Our experience and that of others (see the
this philosophy in mind that we offer the fol- ed and the blowdown project was im- Reservoir Simulation monograph for refer-
lowing comments in response to Schenk's plemented immediately. Blowdown was ences) suggest, however, that in many reser-
five specific questions and remarks. completed late last year with a recovery very voirs with complex geology and where het-
1. Whether simulation has an advantage near that predicted by simulation. erogeneity and multi phase fluid flow must
over simple calculations or extrapolations 2 and 3. There may be some misunder- be considered, prior expectations based on
depends on the engineer' s experience with standing about the use of history matching intuitive judgment are likely to be in error.
simulation and with alternative methods. It to improve reservoir description. In histo- 4. The degree of complexity needed for
also depends heavily on the reservoir- ry matching, the engineer first attempts to a reservoir-simulation model must be based
management issue being considered. Most model history by use of his best current first on what model design will provide an
alternative methods do not allow the reser- description of the reservoir. If historical per- answer to the reservoir-management ques-
voir engineer to consider directly the physics formance cannot be matched with current tion being considered. Often, an important
description, then either the description is design factor concerns what simplifying as-
of the reservoir-depletion mechanism . We
wrong or the physics of the reservoir deple- sumptions will be accepted by the decision
are aware of many situations where, in
tion mechanism is being modeled incorrect- maker for whom the study is intended.
retrospect, starting with a simple , cheap,
ly. In either case, a better understanding of When planning a simulation study , it is im-
easy-to-run simulation model would have
the reservoir is necessary if future perform- portant to determine whether the cost of add-
saved time and money because critical fac- ance is to be predicted reliably . Using a ing model complexity to lend credibility to
tors affecting reservoir performance were simulator is a practical way to consider the study results can be justified. Frequently ,
not well understood. influence of acceptable changes in reservoir in dealing with others who may not have an
An example of this is the reservoir study description on past performance. in-depth reservoir engineering background,
of the "Tom O'Connor 5,lOO-Ft Sand in The Dunlin reservoir discussed in the it is cost-effective to construct a complex
South Texas." (A paper describing this Technology Today Series paper is an excel- model rather than to do the work necessary
study is planned for submittal to SPE for lent example of the benefits of history to develop convincing evidence that simpler
presentation at a general meeting.) This match- models are adequate. Of course, there is no
waterdrive gas reservoir began repressuring ing. The prime question answered in that answer to Schenk's question about how
in the mid-1980's as a result of reduced pur- study was which faults were sealing and much intricacy is advisable. It depends on
chaser takes . A reservoir study was needed which were not. We know of no way to an- the objectives of the study .
to confirm and quantify reserve losses from swer that question without developing a 5. In Reservoir Simulation, we repeated-
continued water encroachment to secure reservoir flow model and matching history ly made the point that the reservoir engineer
Texas Railroad Commission approval for with the model. Perhaps we were too super- should use the fastest and cheapest engineer-
blowdown by special allowable . ficial in our comments on the benefits of the ing methods that will give adequate answers
Initial attempts with simple, tank-type, Dunlin study . Certainly, with the reservoir to the reservoir-management questions un-
material-balance calculations were unsuc- description we now have, it is obvious that der consideration. If simulation is not need-
cessful in reproducing the repressuring ob- "faults dominate areal sweep ... ," etc. ed, and if it is more costly than an alternative
served in the 5,100-ft sand. It was concluded More specifically , however, what are the method, then by all means, don't use it. Our
that conventional, material-balance methods quantitative effects of faulting and vertical contention, however, is that in today's envi-
could not properly account for the compres- permeability and how much injected water ronment, computing power is often cheaper
sion of the gas that was trapped with rep res- will flow through the Etive? Only with a and faster than manpower and that simula-
suring and, hence, could not be used to simulator can a reservoir engineer hope to tion is a superb tool that every reservoir en-
evaluate the influence of repressuring on study enough probable reservoir descriptions gineer should at least know how to evaluate.
to develop rational answers to these
'Now retired .
questions . SI Metric Conversion Factor
We point out here that in most reservoir ft x 3.048* E-Ol : m
, 'Many excellent simulators are available for modeling con-
ventional oil/gas/water displacements. For a large frac· simulations , results are qualitatively similar
tion of reservoir studies. the simulators can safely be to prior expectations. There are seldom 'Conversion factor is exact.
assumed to solve the match correctly. The SPE Mono·
graph Reservoir Simulation discusses exceptions to this
major "surprises," especially if geologists
general statement. and engineers have attempted to develop an (SPE 21620) JPT
1448 November 1990 • JPT

Das könnte Ihnen auch gefallen