You are on page 1of 8

The Stratigraphic Method: A Structured

Approach to History-Matching Complex


Simulation Models
M.A. Williams and J.F. Keating, SPE, Chevron Petroleum Technology Co., and M.F. Barghouty, SPE,
Saudi Aramco

Summary a shaly section, but not in all locations. The SM assumes that these
This paper offers a structured approach to performing a history correlative flow units can be treated independently while history
match on a complex, multilayered reservoir model. We recommend matching. The method also recognizes that the flow units may be
tools for interpreting the data and simulation results, and discuss in vertical communication in parts of the field where the shaly
procedures for improving the quality and efficiency of the history- sections are thin or eroded.
matching process. This approach, the stratigraphic method (SM), The type log in Fig. 1 shows that the five zones are subdivided
was developed over 10 years on various reservoir studies. It has into smaller layers. Currently fine-scale geologic models use hun-
been used successfully on highly complex reservoirs, some con- dreds or thousands of layers, averaging 1 to 3 ft thick. The engineer
taining more than 1,700 wells and over 50 years of production still has to apply some level of scale-up to keep the simulation
history. We also report on which reservoir parameters most affect model within a reasonable size. In some cases, the simulation layers
matching observed-reservoir performance. This may be helpful to may be stratigraphic, but often they are either structural (equal
those with less experience working on a history-match project. thickness) layers to improve vertical resolution, or layers with
similar flow characteristics that have been combined during the
Introduction scale-up procedure. The SM will ultimately focus history-matching
on these individual layers.
Advances in simulation hardware and software have created op- The SM is built on the logic that a complex 3D system can be
portunities to model more complex reservoirs. The underlying reduced to a series of simpler two-dimensional (2D) systems. Then,
assumption is that modeling reservoirs with 100,000 or more cells we apply history-match adjustments at four hierarchical levels until
will better capture the heterogeneity that was blurred with earlier each level is validated. The SM, as its name implies, dictates that
models with coarse layering and grids. The industry uses super- the reservoir is addressed one slice or level at a time. Once we have
computers and advanced-solution techniques to reduce model run matched the observed behavior in one level, we move on to the next
times and computational error. Yet, the largest source of error may level.
be unintentionally overlooked—the geologic description with the Another facet of the SM is approaching the history match starting
engineering control to properly validate it. Geostatistics is oriented with the deepest zones. Because water moves into the reservoir
toward reducing the geologic error, but these probabalistic repre- from the bottom, the match procedure should follow the same
sentations of the reservoir still must be validated with dynamic data sequence. We refer to this as the bottoms-up approach. For reser-
such as pressures, production, and saturations. voirs with free-gas movement, we can similarly use a top-down
The validation of a reservoir model is the most challenging sequence because gas movement usually originates from the pri-
phase, one where the integrated application of geologic and engi- mary gas cap. In this paper, we will present the procedures for
neering principles are exercised. Large models require massive matching an oil reservoir with water encroachment using the
data-management efforts for better understanding of reservoir per- bottoms-up sequence.
formance. Although generating dozens of maps and well plots is The SM is usually applied to full-field models. Therefore, the
necessary, there is no structured approach on how these maps and simulation model is adjusted at four levels: global or fieldwide,
plots are to be used or interpreted. Three-dimensional (3D) displays flow units or layer groups, individual layers, and individual wells.
of simulator results are no substitute for understanding reservoir This hierarchy of adjustment is explained in the next section and
behavior and knowing which parameters to adjust for a reasonable illustrated in Fig. 2.
history match. Furthermore, few publications offer the reservoir
engineer guidance on how to approach such a complex task.
Without a history-match plan, validating a reservoir model can be Hierarchy of Adjustment
an inefficient, haphazard procedure. These inefficiencies waste Global or Fieldwide. History matching usually starts with global
time and also impugn the credibility of reservoir simulation as issues, such as overall energy (pressure) and total field-production
an effective tool; moreover, they delay the model results that are rates. Parameters are adjusted across the model until the pressures
a critical component for making sound reservoir-management and fieldwide water cuts are reasonably matched. Global changes
decisions. correct for conditions such as pressures are too high in all the wells,
or water production is below the observed water rates for most of
Overview the wells. So global changes treat the reservoir much like a large
Today’s models primarily use layering based on reservoir stratig- tank, or one layer, in that the differences between the model results
raphy. Studies devote a large effort to the reservoir characterization and observed data are in the same direction.
phase and often include geostatistics. This is to ensure that the
geologic model accurately portrays the reservoir flow units and Flow Units or Layer Groups. The next level of adjustment
their heterogeneity. Flow units are discrete portions of the reservoir focuses on the primary geologic zones or flow units. In the BLDR
that act as separate or nearly separate reservoirs. Fig. 1 shows a type reservoir in Fig. 1, we will gather performance data that we can
log from a deltaic sandstone which has five major flow units, Zones ascribe to each zone (A through E). Starting from the deepest zone
A through E. The zones in this reservoir are usually separated by (E), we will make changes to the model to match the available data
for Zone E. We will occasionally glance at the other zones to see
if there are any indications that their performance is significantly
Copyright 1998 Society of Petroleum Engineers
affecting Zone E. Otherwise, we will analyze Zone E performance,
Original SPE manuscript received for review 1 July 1997. Revised manuscript make changes only to Zone E, and run the model to see how these
received 15 January 1998. Paper peer approved 20 January 1998. Paper (SPE
38014) first presented at the 1997 SPE Reservoir Simulation Symposium, Dallas, changes affect the match. This is the stage of history matching
8–11 June. where many engineers get tangled up, trying to reconcile massive

SPE Reservoir Evaluation & Engineering, April 1998 169


suffice because we can only match what we have observed. Once
those data are matched, we proceed to the next layer above. Often
the individual-layer match moves quickly because those layers
within a major zone are of similar quality. In some cases, however,
a simulation layer may be separated from the other layers within
that flow unit and local adjustments are needed.

Individual Wells. The final level of adjustment addresses indi-


vidual wells. It is at this point that localized changes are made to
refine the pressures or water-cut performance of each well. This
may include changes to the layer allocations or revisions to well-
relative permeability functions. The changes are minor and re-
stricted to the well cell or surrounding cells. These changes should
not significantly disturb the individual layer matches.

History-Match Plan
We recommend that the validation of the model be approached in
two phases: pressure match and saturation match. The pressure- and
saturation-match phases are very different operations with different
Fig. 1—BLDR0027 Type log. objectives. The SM follows the same basic steps for both phases
(Fig. 3): (1) gather data; (2) prepare analysis tools; (3) identify key
wells; (4) interpret reservoir behavior from observed data; (5) run
model; (6) compare model results to observed data; and (7) adjust
model parameters.

Pressure Matching
The objectives of the pressure match are to correctly distribute
fluids areally, vertically, and temporally. Pressure matching can be
viewed as an in-situ validation of pore volume (PV) and perme-
ability. The PV and compressibility define the pressure decline with
time. The permeability defines the spatial-pressure distribution at
a given time. This section describes how we apply the SM to
pressure matching using the seven-step procedure.

Step 1: Gather Data. The data we need to collect for each well
include processed true vertical depth (TVD) openhole well log
(showing layer tops, perforations, etc.); production logs (flowme-
ters, high-resolution temperature); well-completion history (work-
over dates, shut-ins, etc.); shut-in bottomhole pressures; build-up
tests; repeat formation test or selective formation test (SFT) pres-
sures; and production rates.

Fig. 2—Hierarchy of adjustments. Step 2: Prepare Analysis Tools. The tools we recommend for the
pressure match include production vs. time plot; pressure vs. time
plot; SFT pressure plot; completion vs. time diagram; model grid
amounts of information from several zones and becoming over- showing basal barrier distributions; model grid showing fault
whelmed in data analysis. The SM focuses our efforts on one zone orientations; stamp plots; and error-analysis maps.
at a time until it has been matched. The composite pressure vs. time plot (Fig. 4) and the well
pressure/production vs. time plot (Fig. 5) are common displays used
Individual Layers. The next level of adjustment focuses on in- to assess the pressure match. Typically, we plot the observed
dividual layers, and it is where the bottoms-up analysis becomes pressures with the model-calculated pressures and note where
critical. Here we look at each layer and determine if water is moving differences occur.
up too quickly or not enough. We adjust parameters to match The SFT plot (Fig. 6) is becoming more common in stratified
whatever control data we have in that layer alone. There may be reservoirs where differential depletion occurs. Often the wells are
only a few observed water breakthroughs in that layer, but it will completed in many zones and shut-in bottomhole pressures are less

Fig. 3—Stratigraphic method (SM) steps.

170 SPE Reservoir Evaluation & Engineering, April 1998


Fig. 4 —Composite pressure plot for all wells. Fig. 7—Completion vs. time diagram.

Fig. 5—Pressure and production vs. time plot for well BLDR0051.

Fig. 8 —Basal barrier map showing vertical transmissibility.

from a model. The basal-barrier map was converted into an array


of Tz multipliers, 0 or 1, which indicates if the layer is open or
closed, respectively, to the layer below.
The fault orientations (Fig. 9) are more informative when dis-
played on the model grid. If the faults are being treated as non-

Fig. 6 —SFT plot for well KK-A7.

descriptive. The model-calculated pressures are posted next to the


measured pressures to assess the differences.
The completion vs. time diagram (Fig. 7) is used to check which
layers are exposed to the wellbore at various times. This is a
necessary tool for wells that have been recompleted or when the
completion layers vary among wells. It is difficult to analyze the
well plots and assess zonal performance when the wells are
completed in different zones. The completion layers should be
reconciled with the TVD logs and well histories during model
construction.
In some reservoirs, the geologic units are separated by shale
barriers or impermeable facies. Often the extent and character of
these strata have more impact on the fluid movement than the
variations of rock quality have within the flow unit. The shale map
shown in Fig. 8 is an example of a vertical-transmissibility (Tz) map Fig. 9 —Model grid showing faults.

SPE Reservoir Evaluation & Engineering, April 1998 171


Fig. 10 —Stamp plot of observed and simulated pressures. Fig. 11—Error-analysis bubble map of pressure error.

vertical, then the grid maps should be prepared for each major
geologic zone.
Stamp maps are model grid maps (Fig. 10) with small stamp-
sized pressure plots* superimposed on the well locations. This tool
is intended to provide a 2D representation of the distribution of
pressure differences.
The error-analysis map is a bubble map (Fig. 11) of the differ-
ence between the simulated and observed data (pressures or pro-
duction). Differences between two simulation runs can also be used.
Although these differences are not errors, for simplicity, these maps
will also be called error maps. These differences are taken at the
well locations only. Before the differences can be calculated, the
pressure data must be made continuous in time by the engineer (i.e.,
a line must be drawn through the discrete data). The animation of
this bubble map shows the areal and temporal distribution of
pressure errors. These bubbles can then be placed on other maps
such as permeability or Tz to see if there is any relation between
geologic properties and errors in pressure. Fig. 12—Key wells schematic.

Step 3: Identify Key Wells. Key wells are those completed in only
one flow unit. Because the bottomhole pressures represent only this quickly, particularly if the regular model is large and requires a lot
flow unit, these wells are strong indicators of the discrete zonal of CPU time. Also, ineffective parameter adjustments can be
pressures. Fig. 12 shows the concept of key wells, applied for the identified and avoided during the detailed pressure match with the
pressure and saturation match. The completion diagrams are helpful fine-grid model.
in this step. Also, wells with SFT data are key wells because these
plots indicate the pressures in individual flow units and layers. Step 6: Compare Model Results to Observed Data. The tools
listed in Step 2 are also used to compare the model results to the
Step 4: Identify Reservoir Behavior. As we collect the pressure observed data. It is not important which tools are used the most or
data, we must make initial interpretations on what is occurring in in which order. We recommend automatically generating well
the reservoir. Contour plots of reservoir pressures (isobaric maps) plots, SFT plots, stamp plots, and error maps because they will be
can be created to illustrate the spatial-pressure gradient. SFT plots used extensively.
are useful in assessing the degree of stratification as well as
showing individual zonal pressures.
Step 7: Adjust Model Parameters. As mentioned earlier, the key
The well plots are useful in assessing the temporal-pressure
pressure-matching parameters will be PV, rock compressibility,
behavior. Identifying pressure trends and relating them to aquifer
permeability, and vertical communication. Occasionally, changes
strength, bubblepoint pressure, or response to water injection are
will be required to fluid properties and fault sealing, depending
important for understanding the pressure process. As we run the
upon their initial values. These all assume that the production
model and make changes, our initial interpretations of the pressure
figures are reasonably accurate, and no major modifications will be
behavior may change.
made to the historic rates.
Global-pressure adjustments usually include permeability mul-
Step 5: Run the Model. During the pressure match, we will tipliers, aquifer PV factors, rock compressibility (not adjusted if
specify the historic oil, gas, and water production for each well. The free gas is present), K v /K h ratio, and aquifer transmissibility. At this
model converts these volumes into reservoir volumes, which are stage, we are trying to bring the overall reservoir energy into line
then withdrawn. Specifying the total reservoir voidage eliminates and are not concerned with variations between zones or wells.
the need for having the three phases moving perfectly during the Flow units or layer-group changes usually include only lateral
pressure match. It is more important that the water cuts be correct permeability and vertical transmissibility. Changes to the zonal
during the saturation match than at this point of the process. permeability will create changes to the production allocation to
Sometimes we recommend using coarser-grid models during this each flow unit. Changes to the vertical transmissibility either allow
step. Simpler models can arrive at the same conclusions more or prevents communication with other zones. Individual layer
changes are handled similarly; however, we usually do not have
Personal communication with T.J. Keating (retired), Exxon Production Research Co., much pressure data on individual layers (unless the zone only has
Houston (1994). one layer).

172 SPE Reservoir Evaluation & Engineering, April 1998


Changes to the wells are usually intended to fine-tune production The water-occurrence map (Fig. 14) for each zone or layer must
allocations among the zones in a well. This can be accomplished be prepared concurrently with the water-occurrence table. This
by adjusting the completion productivities (well conductivity). shows a picture of how well the model is matching the frontal
Generally, the pressure match is complete after matching the zonal movement seen in the field. Any well that has been or still is
pressures. With the exception of SFT data, the pressure data are completed in this layer is annotated. If they have produced free
usually not available at levels below the flow unit. water in this layer, a lateral or vertical entry is listed. In some cases,
the wells were coning so the occurrence was a wellbore phenom-
Saturation Matching enon and not a true water entry into the layer. The symbols indicate
The objectives of the saturation match in an oil reservoir are to what type of occurrence we interpreted for that well. This will help
mimic the movement of water and free gas in the reservoir. The when you decide what adjustments to make to your basal-barrier
normal procedure is to match the breakthrough timings and loca- descriptions.
tions by adjusting reservoir parameters, then match water-cut The basal-barrier maps (Fig. 8) indicate whether or not a vertical
performance by adjusting well parameters. As with the pressure flow boundary exists at a well. Generally, some layers exhibit a
match, we will follow the general procedure for the SM. particular fluid-flow pattern, such as water fingering. These layers
usually have a barrier at their base. In the example shown in Fig.
Step 1: Gather Data. The data to collect for each well includes 8, this extensive shale (there were some scattered holes) served as
processed TVD openhole-well logs (showing layer tops, perfora- a platform for water fingering from the flanks and nearby shale
tions); pulsed-neutron cased-hole logs (thermal delay time); pro- “windows.” In this study, the adjustment of the Tz map was the key
duction logs (flowmeters, high-resolution temperature); well-com- to matching water movement for each unit.
pletion history (workover changes, shut-ins); dates of first-water or For highly stratified reservoirs, the barrier descriptions are more
free-gas production; water/oil ratio (WOR) and gas/oil ratio (GOR) easily generated. They may even be unnecessary if it is certain that
(separator tests or shake-out samples); and water-salinity tests every major flow unit is separated from the others. Therefore, the
(injected water vs. formation water). Tz assignments will be set to zero in most locations. This means that
few changes are needed unless there are data to suggest that some
flow units are communicating vertically.
Step 2: Prepare Analysis Tools. The tools we recommend for the
saturation match include completion vs. time diagram (for each
well), WOR or GOR vs. time plot (for each well), water-occurrence Step 3: Identify Key Wells. We have already identified the key
table (for each layer), water-occurrence map on model grid (for wells for the pressure match. Similarly, we need to identify key
each layer), basal-barrier map on model grid (for each layer), saturation-match wells. Single-layer completions that have pro-
model-calculated water-saturation map, and water-encroachment map. duced water are obvious hard-data sources. Wells with pulsed
The completion diagram (Fig. 7) is again used to check which neutron logs that have detected water fingers are also hard-data
layers are exposed to the wellbore during a water occurrence. sources.
Because wells are often plugged-back following a water occur- If there are several layers completed in the well that have
rence, the completion intervals will change with time. New infor- produced water (Fig. 12), we must speculate on which is the
mation becomes available that represents different zones. The offending layer. We cannot be sure without production-log mea-
completion diagram will show which wells represent which layers, surements in this or neighboring wells. Therefore, data from mul-
which will be helpful in Step 3. tiple-completion wells will be considered “soft.” Because water
The WOR vs. time plots (Fig. 13) are self-explanatory. A moves from the bottom upward, we often assume that it enters the
software routine is needed that creates batch plots of model results deepest completion first. This is not always true, especially if there
vs. observed data. These plots, ultimately, are the tool by which the are strong contrasts in layer quality. In most fields, the engineer
quality of the match is assessed. should have some idea which geologic zones cycle the most fluid
The water-occurrence table for each layer (Table 1) is the and which are dormant. This is where the field experience becomes
primary tool to compare the model frontal movement to the ob- critical. The model may also suggest that other layers are the
served breakthrough data. The example in Table 1 was taken from offending layers.
a study where the sources of data were TDT logs and dates of wet Other key wells will be newer wells whose openhole logs detect
production. The left half of the table shows observed data, such as some water encroachment. These wells provide the hardest satu-
the dates and type of each occurrence, and an interpretation on ration data for multiple-reservoir zones. In older fields, the modern-
whether the water intrusion was vertical or lateral. The right half well logs often confirm or modify the original oil/water contact
of the table shows model results for a particular run. The coding on because the old wells have no logs or records of contact depths.
the water location indicates where the waterfront is located at that
time. For instance, 1SE means the waterfront is one cell to the Step 4: Identify Reservoir Behavior. We must make initial
southeast; 2W means two cells to the west. interpretations on what is occurring in the reservoir based on
observed data. Each well’s water occurrence needs an interpretation
of entry type (vertical rise, lateral fingering, coning, etc.). Then, we
can expand that interpretation across the layer. Correlating the well
data, we infer whence the water is coming and how far it has
encroached into the layer. The result is an interpreted water-
encroachment map such as shown in Fig. 15.
Fig. 15 shows a water-encroachment map for a layer. This map
is an interpretation of the current waterfronts based on well logs and
wet production. This map is critical to understanding and matching
water movement. They are usually prepared for each geologic zone.
We will ultimately generate a model-calculated water-saturation
map and compare it with this observed-encroachment map. Pre-
paring the encroachment maps is not a linear process. After we have
made our initial runs, the model may indicate that some initial
interpretations appear improbable. If other scenarios emerge, we
should amend the maps or, at least, note where the waterfronts may
be different. It is important to be flexible when interpreting the
water movement if the model suggests some scenarios to be
Fig. 13—Water/oil ratio vs. time wellplot. improbable. The model rigorously accounts for PV’s, reservoir and

SPE Reservoir Evaluation & Engineering, April 1998 173


TABLE 1—WATER-OCCURRENCE TABLE

Observed Occurrence Model Occurrence

Well (I,J) Type Date Source Date Direction Matched? Remarks

BLDR0037 (54, 57) Vertical 06/82 TDT Yes Lowest perfs in Layer 10
09/86 TDT 1979 Yes
09/89 TDT Yes

BLDR0043 (41, 58) 07/80 SI wet None S1 No Layer 10 swept out


Lateral
02/81 TDT 1985 Late

BLDR0054 (60, 52) Vertical 10/91 TDT 1970 Yes Lowest perfs in Layer 10

BLDR0058 (44, 63) Lateral 1973 SI wet 1972 Yes Bypassed oil in Layer 10

BLDR0085 (52, 62) Lateral 05/91 PDK 1978 Yes Questionable—No OWC detected

Step 5: Run the Model. During the saturation match, we will be


specifying only the historic oil production for each well and will let
the model calculate the water and gas rates. The water and gas rates
will not be constrained by historic voidage rates. However, it is a
good idea to place reasonable producing constraints on the model
wells so that wildly improbable rates do not occur. If the model
calculates extremely high water or gas rates, model instabilities and
other ill effects may result, which further complicates the match.
For instance, we usually assign each well the following constraints:
minimum-flowing bottomhole pressure, maximum fluid rate, and
maximum gas rate (the latter two take into account tubing capacity).

Step 6: Compare Model Results to Observed Data. The inter-


pretation of model results will become more streamlined after the
first runs. We recommend the following steps in this task: generate
batch plots of oil rate and water cut; identify wells not making oil
rate; identify wells with incorrect breakthroughs; and review frontal
movement on layer or flow unit of interest using observed- and
Fig. 14 —Water-occurrence map. model-saturation maps.
Check which model wells are not meeting their historic oil rates.
The waterfront may be arriving too soon, or the flowing bottom-
hole-pressure constraints may be restricting the well.
Next, note which wells are significantly mismatched on their
water-cut performance. If wells are wet in the field and have not
produced water in the model, or vice versa, these are the biggest
offenders. Other wells may simply be experiencing slightly delayed
or early breakthroughs. Indicate on a model-grid map whether the
water is late, early, or not present in each well.
Next, review the calculated water movement for the zone of
interest using a graphics package. Animating the water saturation
with time, note the calculated water movement patterns and when
arrivals are occurring for the model wells. See where the fronts are
moving quickly and where they are moving slowly. The calculated
fronts should resemble those depicted on the water-encroachment
maps.
Finally, generate a hardcopy-model water-saturation map (Fig.
16). We recommend a black-and-white display which can be easily
copied onto a sheet of clear film using the same scale as the
water-encroachment maps. By overlaying this clear film onto the
encroachment map and the Tz map, you can quickly identify
problem areas. The key to success is to confine your review to one
layer at a time so you avoid getting inundated with data.
During the review of the model results, keep in mind the
hierarchy of parameter changes we mentioned earlier. Are the
water-production discrepancies affecting all of the wells (global
Fig. 15—Water-encroachment map, Layer 8, 1992.
level), or are the problems different among the major flow units
(layer group level)? Knowing the answers to these questions is
preparation for the next step.
well geometry, and production volumes. Therefore, it is often a
better estimator of fluid movement and energy distribution. The SM Step 7: Adjust Model Parameters. This is the biggest challenge
supports the concept of learning from the model during history of history-matching water movement. We mentioned earlier that
matching. this paper will not discuss how to adjust each parameter because

174 SPE Reservoir Evaluation & Engineering, April 1998


method to alter the interpretation by reconciling the flow-barrier
locations with historic-water breakthroughs.
Relative-permeability curves and saturation endpoints may also
be changed for the flow unit at this stage. This is necessary when
trying to redistribute the water that is entering a layer. Although
some are opposed to modifying laboratory-relative permeability
curves, the field data may strongly suggest that conformance and
displacement efficiencies are poorer than laboratory estimates. This
requires that changes be made. The model is probably the best
estimator of effective residual-oil saturation because it accounts for
the dynamic interaction of reservoir heterogeneity and fluctuating
flow rates. Changes to the flow unit level must be confined until
all flow units are in reasonable shape based on the water-
encroachment maps.
If there are significant differences between the layers within a
flow unit, we may have to adjust the individual layers. This is done
only if there are significant quality variations between the layers
within a flow unit. An example would be a localized interbedded
shale within a flow unit seen in some wells but not in others. Local
separation may be needed for a particular area, but would not make
a significant impact on the whole model. Also, if the geologists
suggest that there are facies variations within the flow units, or
some areal variance is present, i.e., rock-quality degrading in some
direction, then we can address that at the layer level. Often,
Fig. 16 —Model-calculated water-saturation map, Layer 8, 1992. however, there are not significant property variations between the
sublayers for a given flow unit. As with the flow units, layer-
saturation overlays and Tz maps are needed during this step.
each model has its own specific challenges. But we will mention Changes to individual wells are the final stage of the SM. This
which parameters have been most effective in getting the water to may include adjustments to the amount of fluid coming from a
move correctly. Our experience has been that relative-permeability particular layer. This is often the case when we do not have
functions are the most effective parameter changes for saturation flowmeters to confirm how much of the well’s production is
matching. Although changing the permeability and PV will also coming from the layer of interest. In history mode, the layer
have an effect, this will also ruin the pressure match obtained productivity-index assignments determine the allocation of the
earlier. So we confine ourselves to changing these parameters: the well’s production to a given layer. Adjustments to this are common
shape of the oil/water relative-permeability curve, Swir and S orw , to promote or restrict water movement in a certain flow unit or
Krw, and Krow, vertical transmissibility (Tz), and layer-productivity layer. Unless there are only a few wells in the model, an allocation
index assignments. These parameters allow us to redistribute the change will probably not affect the surrounding wells. That is, one
water in the model without significantly affecting the reservoir well’s production is not a significant fraction of the total reservoir
pressure. production.
Initially, the model calculates poor gas and water rates for most Changing flow allocations can affect the pressure match, so this
of the wells. This is common because laboratory-relative perme- move must be made with caution. If flowmeters are available,
ability is measured under different flow regimes than exist in the specify this allocation during the pressure match so that changes
reservoir. The dispersed flow in core floods does not normally will be less severe during the saturation match. In stratified reser-
occur in the reservoirs where gravity is a dominating influence. voirs, the flow allocations are changed during the pressure match
These authors have noted that sandstone reservoirs use an initial and probably will not be further modified.
Krw that is usually too high, causing the model to overpredict the
water rates.
Global changes should be made if all wells or all major flow units Summary
are experiencing similar behavior. Globally increasing or decreas- We propose that the key to history-matching a complex reservoir
ing the Krw curve will usually make a satisfactory overall impact. model is to break it down into component slices and adjust each
If we see some zones with early breakthroughs and others with late slice accordingly. With the projected use of larger and more
breakthroughs, then we probably should move on to the next level complex models, this tactic will become critical in history match-
of change, which is the flow unit or layer-group level. ing. We emphasize the importance of exercising engineering and
Now we can address these differences among the major flow geologic control during the history match using a process that is
units, starting from the deepest. Because the flow unit is comprised structured, controlled, and validated one step at a time. We sum-
of several layers, we will change all the layers within a specific marize with the following:
layer group at the same time. 1. Increasingly complex simulation models will make it more
When the flow units are separated by discontinuous barriers, look difficult to manage and interpret results from the model. The SM
at Tz between the layers. As we move from the deeper flow units can convert complex 3D problems into simpler 2D components.
into the shallower flow units, the key to matching them may rest 2. The SM is based upon reservoir stratigraphy, which is re-
with how well we have matched the underlying unit. Overlaying the flected in the correlated flow units. Understanding the individual
Tz map (Fig. 8) onto the layer water-saturation map (Fig. 16) is a flow units and addressing them individually are key to making an
good way to identify where holes might be opened to the underlying effective history match.
flow unit or where they must be closed off. This is straight-forward 3. Maintaining effective engineering and geologic control is
in the well cells because we have well logs to indicate the shale’s more difficult without an organized history-match plan. The SM
presence. The challenge is refining the shale description between offers an approach that incorporates all observed data, facilitates
and away from the wells. This is the most difficult and time- understanding of the reservoir behavior, and directs changes that
consuming step in the SM because we have to make many runs are within the framework of observed data.
before we achieve a reasonable match on each flow unit’s water 4. The SM prescribes proven analysis tools for understanding
movement. Sometimes holes must be opened several cells away to reservoir behavior and achieving a controlled and structured history
get water to arrive at the proper time. The lateral extent of a shale match. These tools focus on comparing observed data with model
is always subject to interpretation. Simulation provides a rigorous results.

SPE Reservoir Evaluation & Engineering, April 1998 175


5. The SM provides a step-by-step procedure for reviewing Model Study, 1993,” Chevron Petroleum Technology Co., Houston
model results and comparing them to observed performance data. (December 1993).
This provides guidance and improved efficiency for reservoir- Barghouty, M.F., Keating, J.F., and Williams, M.A.: “Berri Field, Hanifa
management teams. Reservoir, Tight Area Pattern Flood Evaluation, 1995,” Chevron Petro-
6. History matching can be expedited by using temporary models leum Technology Co., Houston, (December 1995).
with a coarser grid over the entire model or in areas where less Harrison, C.J., McKay, D.V., Havrilla, M.A., and Williams, M.A.: “Tshiala
resolution is needed. Field, Lower Pinda Reservoir Study,” Chevron Overseas Petroleum Co.,
7. The SM has been successfully used on many complex simu- San Ramon, California (June 1996).
lation studies, reducing history-match cycle time and improving the
integrity of the match.

Nomenclature SI Metric Conversion Factors


Kh 5 horizontal permeability ft 3 3.048* E201 5 m
Krow 5 oil-relative permeability to water *Conversion factor is exact. SPEREE
Krw 5 water-relative permeability
Kv 5 vertical permeability
Sorw 5 residual oil saturation to water
Swir 5 irreducible water saturation Mark Williams is a Consultant Reservoir Engineer with Chevron
Tz 5 vertical transmissibility Petroleum Technology Co. in Houston, where he specializes in
applied reservoir-simulation projects. His areas of interest are
the integration of finite-difference reservoir simulation and ge-
Acknowledgments
ology, principally through the history-matching process. He also
The authors wish to extend their appreciation to the following conducts courses in applied reservoir simulation at the U. of
colleagues for their advice in helping to develop and deploy this Houston. He holds a BS degree in petroleum engineering from
method: S.G. Johnson, P.S. Kumar, J. Afifi, M.T. Aktan, W.M. the U. of Southwestern Louisiana. James Keating is a Lead
Brummett, M.A. Havrilla, and W.Y. Clark. We also want to thank Reservoir Engineer with Chevron Petroleum Technology Co. in
our Publications and Presentations staff, Sharon Sloan, Larry Houston. He has history matched several large full-field black-oil
Startz, and Betty Lundquist, for their preparation and editing efforts and compositional models. Keating is also a strong proponent
on this paper. of using finite-difference simulation as a practical alternative to
analytical techniques for well testing and material balance. He
General References holds BS and MS degrees in petroleum engineering from Texas
Tech and a PhD from Texas A&M U.
Acharya, U.K., Grant, C.W., Keating, J.F., Stockey, J.L., Williams, M.A.,
Al-Humam, F., and Barghouty, M.F.: “Safaniya Field, Khafji Reservoir,
Full-Field Model, 1994,” Chevron Petroleum Technology Co., Houston
(December 1994).
Afifi, J., Aktan, M.T., Ballengee, W.G., Brenneman, R.J., Johnson, S.G.,
Kumar, P.S., McKay, D.V., Merrell, J.M., Williams, M.A., Zapata, V.J.,
Abdul Rahman, B., and Al-Hamoud, J.A.: “Greater Burgan Field, Sim-
ulation Model History Match, December 1994,” Chevron Overseas
Petroleum Co., San Ramon, California (January 1996).
Al-Amoudi, S.M., Al-Awami, A.A., Jensen, C.L., Kasischke, W.F., and
Williams, M.A.: “Safaniya Field, Safaniya Reservoir, Onshore Sector Williams Keating

176 SPE Reservoir Evaluation & Engineering, April 1998

You might also like