Você está na página 1de 14

SPE-170669-MS

The Use Of Reservoir Simulation In Estimating Reserves


A.D.W. Jones, BP Exploration; F.R. Denelle, Shell Exploration & Production Co; W.J. Lee, University of Houston;
D.G. MacDonald, BP Exploration; B.J. Seiller, Total S.A.

Copyright 2014, Society of Petroleum Engineers


This paper was prepared for presentation at the SPE Annual Technical Conference and Exhibition held in Amsterdam, Netherlands, 2729 October 2014.
This paper was selected for presentation by an SPE program committee following review of information contained in an abstract submitted by the author(s). Contents
of the paper have not been reviewed by the Society of Petroleum Engineers and are subject to correction by the author(s). The material does not necessarily reflect
any position of the Society of Petroleum Engineers, its officers, or members. Electronic reproduction, distribution, or storage of any part of this paper without the written
consent of the Society of Petroleum Engineers is prohibited. Permission to reproduce in print is restricted to an abstract of not more than 300 words; illustrations may
not be copied. The abstract must contain conspicuous acknowledgment of SPE copyright.

Abstract
This paper proposes an approach for assessing a reservoir simulation model for use in estimating reserves.
A simulation model can integrate complex static data, the physical description of displacement processes,
production constraints and schedules. Hence it can provide important input to business decisions and to
reserves estimation. Confidence in simulation predictions depends on the strength of evidence for the
input data, quality control of the model, robustness of the history match, and whether there is independent
evidence supporting predictions. We explain the principles for evaluating a simulation model and propose
requirements for simulation predictions to be considered as proved reserves. This involves evaluation
against different strands of evidence e.g. static and dynamic characterisation, wells and facilities description, reservoir performance and analogues. Simulation models are often built to support business decisions
using best technical estimates for inputs. There can be instances where a simulation model may be
reasonable and reliable but it only represents abest technical outcome. There may not be sufficient
evidence to count the whole predicted recovery as proved reserves. We propose how such a model may
be modified to also provide proved reserves estimates. The approach is illustrated through a case study
which shows how the principles may be applied with different available data and at different stages of
field life.

Introduction
A simulation model can integrate complex static data, the physical description of displacement processes,
production constraints and schedules. Hence it can provide important input to business decisions and to
reserves estimation. For estimates to be reported as proved reserves sufficient confidence must be
established in the predicted production. Companies and institutions commonly assess reserves using
definitions either from Petroleum Resource Management System (PRMS) (2007) or from United States
Securities Exchange Commission (SEC) (Modernisation 2008). Under these definitions there must be
Reasonable Certainty in estimated production from a commercial project for this to be classified as proved
reserves. If deterministic methods are used then Reasonable Certainty means for PRMS a high degree of
confidence that estimated quantities will be recovered and for SEC much more likely to increase or
remain constant than to decrease. The challenge is to assess whether model predictions from a particular

SPE-170669-MS

simulation meet the standard of Reasonable Certainty and how to adapt less certain predictions to meet
the standard.
Several authors have proposed approaches to assessing or modifying simulation models for use in
reserves estimation. Rietz and Usmani(2005) focus on the situation in which there is sufficient performance data to test a model against history. They highlight assessment of:

intended use of the model; a model not built for reserves may require extra scrutiny
reasonableness in model construction, especially key input parameters
quality of the history match, the reasonableness of changes involved and their impact on
predictions
transition to prediction and assessment of prediction quality through matching early history and
testing predictions of later history (hind-casting)
In a subsequent paper Rietz and Usmani(2009) give examples of how a model built as a best estimate
may be modified to provide a prediction of proved reserves. Where Hydrocarbons Initially In Place (HIIP)
were larger than could be supported for proved reserves either simulation input or output were modified
so that the revised predictions were consistent with a reduced HIIP. For an immature reservoir, with little
performance data to validate the planned secondary recovery process, they propose running a simulation
with depletion drive for the proved reserves case. Secondary recovery prediction is included in the 2P
(proved plus probable) case.
Lee, Sidle and McVay (2011) apply the concept of Reliable Technology which Modernisation (2008)
defines as a grouping of one or more technologies (including computational methods) that has been field
tested and has demonstrated consistency and repeatability in the formation being evaluated or in an
analogous formation. They support the areas for assessment identified by Rietz and Usmani(2005), and
conclude that a simulation model can be validated as a Reliable Technology and used to estimate reserves
of any category for a specific reservoir and a specific model build.
The current paper aims to build on the work by these previous authors and to provide a set of principles
for using simulation as part of a proved reserves assessment. The focus is on deterministic, rather than
probabilistic approaches, and we deal with the technical assessment of gross reserves i.e. whether the
estimate of gross recoverable quantities from simulation meets the standard of Reasonable Certainty
required for proved reserves.
The section on evidenced-based reserves classification reviews the approaches for estimating recoverable quantities and identifies the components of a technical estimate that should be assessed and the
types of evidence which support Reasonable Certainty. This understanding is applied in the following
section (assessing evidence for simulation predictions) to develop the criteria for assessing whether
estimates from a simulation model meet the requirement of Reasonable Certainty. These are consistent
with previous work but provide more structure to the assessment of the simulation inputs and outputs, and
introduce the use of analogues or of independent analysis of performance data to test model output. The
approach is illustrated with a case study.

Evidenced-based reserves classification


PRMS (2007) describes three analytical approaches: volumetric estimates, analogues and production
performance analysis. Volumetric estimates use reservoir rock and fluid properties to calculate hydrocarbons in place and then estimate how much will be recovered by specific projects. The analogue
approach uses fields with comparable reservoir and fluid properties and comparable development plans to
estimate production. Production performance analysis predicts production either from the analysis of
trends e.g. decline curve analysis, or from a model of the physics e.g. material balance.
The confidence in an estimation approach depends both on the quality and quantity of data, and on the
robustness of interpretation and predictions. It is common for a combination of methods to be used in a

SPE-170669-MS

Figure 1Approaches to estimating recoverable quantities. Shading indicates increasing confidence when two or more approaches are used

reserves estimate and confidence is increased where an estimate is supported by more than one approach.
For example a volumetric estimate may be supported by data from analogues for rock properties or
recovery factor and from performance analysis for in place volumes or for dynamic performance. This is
summarised in Fig. 1 which illustrates increasing confidence when two or more methods are used.
In order to classify a deterministic estimate as proved reserves there should be sufficient evidence for
the case to provide Reasonable Certainty. Table 1 provides a framework for assessment of the evidence.
The first five elements, in the table above, describe the reservoir and fluids. The description of
Container, Hydrocarbon Column, Reservoir Properties and Fluid Properties combine together to give
HIIP. The next three elements depend on performance of the planned development. The Defined
Reservoir Area is the area that is in communication with the planned wells and so depends both on the
planned well locations and on reservoir connectivity. Economic Producibility describes the ability of wells
to produce at economic rates, whilst Wells Facilities and Export describes the non-subsurface components
of the development and how they will perform. The last two elements, Recovery Factor and Production
Profile, describe the production of the reservoir resulting from the planned development.
Reasonable Certainty required for proved reserves is determined by assessing the strength of evidence
for each of these elements. For example, reliable interpretation of oil and water pressure gradients in the
same Reservoir Unit would support an oil-water contact used in a proved reserves estimate. (A Reservoir
Unit is a set of connected porous rocks containing both hydrocarbons and water that are in communication
within the hydrocarbon column.) If water pressure data were not from the same unit, and the only other
evidence was from a log in the oil column, then only the oil volume above the lowest known oil (LKO)
would be included in the proved reserves estimate. Predicted recovery from deeper oil would be in a lower
category.
Volumetric analysis, analogues and field performance may each provide supporting evidence for the
technical elements listed in Table 1. As discussed above, the case is stronger when there is support from
more than one method. Recovery factor predicted from analytical or numerical modeling would meet
Reasonable Certainty if there were also clear supporting evidence from field performance or analogues.

SPE-170669-MS

Table 1A framework for assessing evidence for reserves estimates

However, in the absence of such evidence, recovery factor used in the proved reserves assessment would
be the lower of natural depletion or assisted depletion/aquifer support.
All the available evidence should be considered, and field production data, if available, used as part of
an integrated analysis to support reserves estimates. It is normally expected that there would be strong
supporting evidence for each of the ten elements in Table 1 in order to consider predicted recovery as
proved reserves. Sometimes there may not be sufficient evidence of some of the elements for the full
estimate. In this case it may be possible to define a smaller volume for which there is clear evidence as
in the examples of oil column and recovery factor described above.
In some cases there is sufficient evidence for confidence in the estimate even when the evidence for
some elements is less compelling. For example performance analysis, such as material balance, may
provide confidence in the estimate of in place volumes, even when there is uncertainty in the shape of the
container.

Assessing evidence for simulation predictions


The previous section identified the technical elements to be assessed and the types of evidence that support
Reasonable Certainty. It also highlighted the importance of volumetric analysis, analogues and performance analysis in providing complimentary supporting evidence. This section applies this understanding
to the assessment of simulation prediction and defines eight criteria for assessing whether a deterministic
estimate from simulation meets Reasonable Certainty required for proved reserves. The approach is to
assess the evidence for the recovery and production mechanisms, for simulation inputs and for validation
of simulation outputs.
1. Subsurface and non-subsurface mechanisms impacting production rate and recovery are identified
and appropriately included in the model
An appropriate model is a prerequisite to confidence in predicting recoverable volumes from
simulation. This requires:

SPE-170669-MS

evidence for what recovery processes are operating in the reservoir


confidence that the model can represent the recovery processes
an understanding of the key subsurface and non-subsurface parameters required to model
performance
For example, for an improved recovery technique to be included in the proved reserves estimation
one would expect supporting evidence for its effectiveness. Modernisation (2008) states that this
can be from Successful testing by a pilot project in an area of the reservoir with properties no
more favorable than in the reservoir as a whole, the operation of an installed program in the
reservoir or an analogous reservoir, or other evidence using reliable technology establishes the
reasonable certainty of the engineering analysis on which the project or program was based. Once
there is clear evidence for the recovery process, successful modelling of the performance of
producing wells, an appropriate pilot or an analogue reservoir can provide evidence of the ability
of a model to represent the recovery mechanisms correctly. If a waterflood is to be modelled then
this implies the requirement for evidence of relative permeability properties, injection well
performance and facility water handling capacities.
2. The static and dynamic description and development plan in the simulation model are clearly
supported by evidence, including

Description i.e. Reservoir Container, Hydrocarbon Column, Reservoir Properties, Fluid


Properties and HIIP (see Table 1). Performance data may provide sufficient evidence for
HIIP to supplement evidence for some of the components of the static description
Development i.e. Defined Reservoir Area, Economic Producibility, Wells Facilities and
Export (see Table 1). Properties such as permeability and relative permeability are normally
assessed as part of the reservoir description
This step is an assessment of the evidence for the inputs to the simulation model. Given the
complexity of many simulation models, it is important to demonstrate not only that the defined
inputs are supported by evidence, but also that the inputs are correctly represented in the model.
For example a check would be made that hydrocarbon contacts are reproduced in the model, and
that HIIP calculated by the model is consistent with independent calculations. Checks should also
be made that connectivity in the model is consistent with the static and dynamic description and
that the intended facility capacities and efficiencies are honoured in predictions. Similar checks
would be made for each of the inputs.
3. The model is matched to history at field and local level. Changes are justified, physically and
geologically realistic, and are consistent with evidence. There is a smooth transition to prediction.
This follows the principle, discussed in the previous section, that performance data is used as part
of the integrated analysis. The history match is a calibration of the model and a check that it is
consistent with performance data. In adjusting the model, changes should be consistent with the
evidence supporting the static and dynamic description.
4. Predictions are tested against evidence from performance data or analogues.
a. With sufficient production data, simulation predictions may be supported by performance data
from the reservoir either through successful hind-casting predictions from a robust history
match and prediction or through comparison of predictions with independent analytical
methods such as decline curve analysis
b. Where there is insufficient performance data, simulation predictions may be supported by
demonstrating the ability to model analogues with sufficient performance data
c. Where neither performance data, nor analogues are available, recovery factor used in the
proved reserves assessment should be the lower of natural depletion or assisted depletion/
aquifer support

SPE-170669-MS

This step is a test of the last two elements of Table 1, which are outputs of the simulation model.
As discussed in the Introduction, previous papers on the use of simulation for reserves estimation
have mostly focused on using hind-casting predictions from a history matched model. If this
approach is used then it is necessary to demonstrate that the history match and predictions are
robust. If a range of predictions can be achieved with alternative matches, then for proved reserves
there must be Reasonable Certainty in the estimated recovery.
With sufficient performance data, a model may also be validated by comparing predictions with
independent analytical methods such as decline curve analysis. This is preferred to the hind-casting
approach as it is a more independent test. In this case the comparison would normally be of the
case with no further development. The validated model could then be used to predict recovery
including some additional wells if this involved the same recovery processes and the current and/or
historical wells were analogues for the performance of new wells.
Where there is insufficient performance data to validate predictions, the recovery factor from
analogue reservoirs may be used. In this case there must also be evidence that the recovery process
is effective in the planned development, such as demonstrating response to water injection in a
waterflood recovery process. Recovery factor for the proved reserves case should be consistent
with that achieved in analogues. If simulation predicts higher than the consistent recovery factor
then the difference would be classified in a lower category.
Without validation from performance data or analogues, recovery factor for proved reserves
should be limited to the lower of natural depletion or assisted/aquifer support. Natural depletion
would typically be lower for oil reservoirs, but aquifer influx may give lower recovery for gas
reservoirs.
5. The model is kept current through regular updates and is consistent with the most recent
development plan and with technical evidence. The most recent performance does not have to be
included in the history match as long as the model is still valid
This step follows the principle that all the available evidence should be considered. The reservoir,
well and facility performance in simulation should be consistent with historical performance and
should model the planned development.
6. There is an understanding of the model sensitivities and all the key inputs impacting predictions
are supported by evidence
This sensitivity is to support points 1 and 2 above on appropriate model and model inputs, not to
provide a probabilistic view to define categorization. In a deterministic approach, confidence in the
proved case is established by evidence, not by the difference between proved, 2P and 3P reserves
estimates. The quality and completeness of the evidence will determine which elements of the
estimate are clearly defined and which have an uncertainty range, leading to a range in recovery
estimates. Simulation sensitivity provides confirmation that there is an understanding of what
impacts the predictions and enables a check that there is evidence for each of the key inputs for
the proved reserves case.
7. A primary model, which represents a best technical case, can be used to calculate the proved
reserves volumes if the differences between the primary model and proved case are understood
and an approach has been established for developing predictions which meet standards of
evidence for proved reserves
Simulation models are often built to support business decisions using best technical estimates for
inputs. There can be instances where a simulation model may be reasonable and reliable but it only
represents a best technical view. Instances in which this applies are where the evidence or certainly
level required for proved status has not been reached in either the inputs or mechanism the
simulator is representing. Each of the elements of the case is tested against evidence. Where this

SPE-170669-MS

is insufficient a revised case should be defined which is clearly supported by evidence. This is
illustrated with two examples:

Evidence supports a proved case with a Hydrocarbon Column only down to LKO, but the
business case model has a deeper OWC. The model may be modified by raising the contact to
LKO so as to provide an estimate of proved
Evidence supports a proved case with a limited Defined Reservoir Area, but the business case
has a larger area in communication with the development wells. The model may be modified
by preventing communication from beyond the proved case Defined Reservoir Area
In these examples the model was changed to develop the proved reserves case. Alternatively it may
be appropriate to modify the model output by scaling the predicted production by the fraction of
the HIIP remaining once the contact has been raised to LKO, or the Defined Reservoir Area has
been limited. The appropriate approach should be selected after considering whether there will be
a material change in recovery factor.
8. The model is documented, including inputs, history match adjustments, any adjustments for the
proved reserves case, outputs and supporting evidence
Documentation of the model and supporting evidence should demonstrate that the proved reserves
case is consistent, supported by evidence and compelling.

Case Study
This section presents a case study of estimating reserves from an oil reservoir producing under waterflood.
It is based on an actual field, but some aspects have been changed to help illustration.
The reservoir is a series of stacked basin floor submarine fan and base of slope channel sands with three
major intervals and two minor intervals, separated by thick mudstones which act as vertical barriers. It
extends over an area of approximately 55km2. Depositional sand bodies range from 10m to 60m thick and
up to 3km wide. Reservoir quality is good with average porosity 26%, net-to-gross 55%, and permeability
1000mD. Typical reservoir oil properties are 26 API, 4cP viscosity and 2900psi bubble point. Several
appraisal wells drilled through primary gas caps. The field is areally divided into five segments due to a
combination of faulting and stratigraphic boundaries. The field has been developed with 26 producers and
15 water injectors and has been on production for over 15 years. Two further producers and one injector
are being drilled in the next year.
There is an extensive dataset with 51 logged wells, over 1100m of core from 8 wells, fluid samples
from 10 wells, high quality 3D and time-lapse seismic, pressures in most wells from a formation testing
tool (FTT), bottom hole temperature and pressure gauges, frequent well tests and a series of tracer tests.
A detailed geological model has been built based on seismic and well data and a history-matched
simulation model predicts remaining production to end of license in 10 years.
Fig. 2 illustrates how components of the total recovery predicted from the business case model are
classified. It will be shown that there is clear evidence for most of the predicted recovery from existing
wells, and from new wells, which is classified as proved reserves. The evidence is less conclusive for
incremental recovery below LKO, or due to property improvements or improvements in facilities on-time.
These incremental volumes are classified as probable reserves. The following shows how the classification is determined by applying each of the eight criteria, defined in the previous section.
1. Subsurface and non-subsurface mechanisms impacting production rate and recovery are identified
and appropriately included in the model
The field is covered by multiple 3D seismic surveys and these have been re-processed several
times, as technology has evolved, to increase the resolution. Moderate reservoir depth, the nature
of the over-burden formations, and the reservoir rocks all combine to give favourable conditions
for seismic. As a result the high quality seismic combined with more than 60 penetrations allows

SPE-170669-MS

Figure 2Classification of recovery predicted from simulation

Figure 3Well oil rate and GOR response to water injection

tightly controlled mapping of surfaces, distinction of sand bodies across the field and imaging of
small-scale faulting. A few faults completely offset the reservoir but the majority do not and the
modelling of the shale gauge ratio suggests that where there is sand-on-sand juxtaposition the
faults will be non-sealing.
The dynamic data provides additional understanding of connectivity beyond the static description,
and some adjustment of communication is an important factor in achieving a good history match,
discussed below. The wealth of FTT, injection response, tracer data and time-lapse seismic,
together with material balance analyses, demonstrate that there is pressure continuity across most
of the field within the main stratigraphic units.
The presence of a free primary gas cap indicates much of the reservoir is at or close to bubble
point, so the development scheme has focused on maximising sweep and pressure support through
water injection. A range of surveillance data has demonstrated the effectiveness of the waterflood.
Fig. 3 shows an example of oil production rate and gas oil ratio (GOR) from a well. Injection
started at a neighbouring injector after approximately one year and, in response to the pressure
support, oil rate increased and GOR continued to decline. Waterflood sweep and displacement
efficiency were measured through saturation logging of a well drilled behind the flood front. The
log showed good vertical conformance and low residual oil saturation. Multiple 3D seismic
surveys have been shot over the field. Seismic amplitude responds to pressure and saturation and
difference maps show water advance and pressure support in most sectors.
Early in the development, the extent of aquifer support was uncertain. Material balance and

SPE-170669-MS

Table 2Summary of evidence for static and dynamic description and development plan

time-lapse seismic response have now shown that aquifer support is generally moderate to poor.
The effect of rock compressibility is limited because the rock is reasonably competent (compressibility 12e-6/psi), so natural depletion is not effective. Gas cap drive is also largely ineffective due
to the low reservoir dip of 3 4 degrees and moderate oil viscosity.
A detailed geological model was built based on the extensive well and seismic dataset. This
modelled thesedimentary and structural heterogeneities and the understanding of communication.
Core data was used to calibrate the porosity and permeability modelling. The static model was
imported into simulation with fluid and dynamic properties assigned from laboratory measurements. The dynamic behaviour of the model was calibrated against the detailed surveillance data
described above.
The performance of the development depends not only on reservoir description but also on well
and facility performance. Modelled well injectivity and productivity is calibrated to field measurements, with typical initial rates of 20mbd production and 40mbd injection. Lift curves and
pressure drops through flowlines are modelled as are facility capacities and efficiencies. Predicted
production is below the production facility limits for all three phases but is close to water handling
capacity and is limited by injection capacity and also by facilities on-time.
2. The static and dynamic description and development plan in the simulation model are clearly
supported by evidence
Table 2 summarises the supporting evidence. In the assessment of the simulation model this
evidence is examined in detail to check where it provides Reasonable Certainty or where evidence
is weaker. For example the table briefly summarises the evidence for Hydrocarbon Column, but
evidence for contacts in each Reservoir Unit should be examined. Where seismic is used to provide
evidence for Reservoir Container, Reservoir Properties or Defined Reservoir Area, it should be
established that in this field the seismic is a Reliable Technology. The model is also reviewed to
check that the inputs are correctly represented.
There are three aspects, highlighted in red in the table, where the evidence supporting the inputs
to the business case simulation model does not meet the standards of Reasonable Certainty:

10

SPE-170669-MS

Figure 4 Examples of a) well-level history match and b) well-level transition to prediction

In one Reservoir Unit existing wells have only penetrated the oil column. There is clear
evidence for lowest known oil from logs, but evidence for the deeper contact is weaker. It
depends on using water pressures measured from a local aquifer but it is unclear whether the
aquifer in the Reservoir Unit is at the same pressure. This will be resolved when the planned
injector is drilled
Comparison of mapped net-to-gross and porosity with well averages shows improvement
away from well control in segments with few well penetrations. This improvement is inferred
from geological understanding and analogues but is not yet supported by well data. This can
potentially be resolved when additional wells are drilled or later when there is sufficient
production data to establish HIIP from material balance
The business plan is based on an on-time efficiency predicted from studies, but this is higher
than supported by track record
In each of these cases there is some supporting evidence but not sufficient for Reasonable
Certainty. This leads to categorizing some of the predicted recovery as probable reserves rather
than proved, as discussed below under criterion 7.
3. The model is matched to history at field and local level. Changes are justified, physically and
geologically realistic, and are consistent with evidence. There is a smooth transition to prediction
The 15 years of performance history and the range of surveillance provide a good data set for
calibrating the simulation model. In history the wells are controlled on oil rate and the model is
matched at a well-by-well level to FTT pressure data, bottom-hole pressure build ups, gas-oil ratio,
watercut and flowing bottom hole pressures, illustrated in Fig. 4. Saturation and pressure distributions are compared with time-lapse seismic results.
The changes made to achieve the history match are consistent with the supporting evidence
summarized in Table 2. For example some changes were made to connectivity between sand
bodies and across faults to reflect communication, baffles and barriers inferred from pressure data,
well response, tracers and time-lapse seismic. Some modifications were made to local permeability
and well skin factors to improve modelling of well performance. However, contacts were not
moved, nor were porosity, net-to-gross and HIIP changed. The model is well matched to history

SPE-170669-MS

11

Table 3Summary of evidence for recovery factor and production profile

Figure 5Comparison of well-level analytical predictions from decline curves and water-oil ratio trends

and there is a smooth transition to prediction as illustrated in Fig. 4. Fig. 4b also shows analytical
predictions which will be discussed next.
4. Predictions are tested against evidence from performance data or analogues
Evidence supporting the predicted recovery and profile is summarized in Table 3
As a check of simulation predictions, several alternative analytical approaches were used to predict
the performance of existing wells directly from performance data:

linear oil cut decline curves against cumulative production for each well
semi log plots of water-oil-ratio increase with cumulative oil production both at a well and
field level
field-level fractional flow analysis to predict recovery as a function of pore volumes of water
injected
Quantities were plotted against cumulative production (injection) to remove the impact of well and
facility on-time. Comparison of the two methods used at a well level demonstrates very consistent
predictions (Fig. 5). Comparison of all four methods at the field level, together with simulation of
existing wells also shows consistency (Fig. 6) implying that the analytical predictions are robust
and that these support the simulation model.
Predicted recovery from new wells is much less than from existing wells in this mature field (see
Fig. 2). These predictions are supported by using the existing wells as analogues to compare initial
rates, decline, watercut buildup, and total recovery. In addition the predicted recovery from all
wells is supported by comparing the recovery factor with recovery factor from analogue fields
(Fig. 7). This figure shows waterflood recovery from fields with similar oil mobility, similar
net-to-gross and producer well spacing, classified according to reservoir complexity, see Dromgoole and Speers (1997). Fields with similar complexity to this case study generally achieve higher
recovery factors. It is concluded that both reservoir performance data and analogue data support
the simulation predictions.
5. The model is kept current through regular updates and is consistent with the most recent
development plan and with technical evidence
At the time of the assessment, the model had been history matched to the end of the previous year

12

SPE-170669-MS

Figure 6 Comparison of analytical and simulation predictions of recovery from existing wells

Figure 7Comparison of predicted recovery factor with performance of analogue fields

i.e. a further 12 months of performance data was available. The planned activity set for new well
schedule, well work activity and surveillance shut-ins was updated in the model. Predictions were
consistent with history for the last 12 months. The analytical predictions from performance data,
described above were up to date and supported the model.
6. There is an understanding of the model sensitivities and all the key inputs impacting predictions
are supported by evidence
A sensitivity study was conducted on the business case simulation model in which the impact of
description, waterflood performance, facilities capacity, efficiency and well performance were
evaluated. Key impacts were:

static description (net-to-gross, porosity, OWC) in the less-developed segments


water-cut trend
water production and injection capacities
facilities on-time
As identified in the discussion of simulation inputs, there are some uncertainties in facilities
on-time and in the static description of the less developed segment. The proved case uses a more
conservative description than the business case: no improvement in on-time above track record, no
improvements in rock properties, OWC at the logged LKO.
Some sensitivities were run to help understand performance. Late in life the predicted water

SPE-170669-MS

13

production is close to facility limits, hence high watercuts or reduced waterhandling capacity could
limit production. There is also the potential for increasing recovery within the license period by
increasing water injection, together with increased waterhandling capacity. The study highlights
the importance of monitoring and managing water production and injection to maximize recovery.
The description of waterflood performance in the model is appropriate since injection and
production limits are based on track record and the performance of the waterflood is supported by
watercut data, logged saturation behind flood front and time-lapse seismic.
7. A primary model, which represents a best technical case, can be used to calculate the proved
reserves volumes if the differences between the primary model and proved case are understood
and an approach has been established for developing predictions which meet standards of
evidence for proved reserves
As identified above, most inputs to the business case simulation model are clearly supported by
evidence, but there are 3 areas where the evidence is less conclusive: improvement in facilities
on-time above the track record, reservoir property improvement and OWC in the less-developed
segments. The simulation model was rerun with no on-time improvement, no rock property
improvement and OWC at LKO. This provides the basis of the proved reserves case shown in Fig.
2, and the difference between this and the business case prediction defines the probable reserves.
An alternative approach to estimating proved reserves would have been by post-processing
predictions from the business case model: calculate a revised HIIP in the less developed segments
and apply the same recovery factor as in the business case; next scale back the profile further to
account for no on-time improvement.
8. The model is documented, including inputs, history match adjustments, any adjustments for the
proved reserves case, outputs and supporting evidence
The documentation provides a full description of the evidence, which has been summarized in this
section. The documentation is a series of montages and written descriptions that are presented for
detailed review, together with the complete input dataset and model files and analysis files which
can be recalled and interrogated.

Summary and conclusions


This paper has described a systematic approach for assessing whether the estimate of gross recoverable
quantities from simulation meets the standard of Reasonable Certainty required for proved reserves. Ten
technical elements of a reserves case were identified, covering the static and dynamic description,
development and production. These elements provide a framework for evaluating the strength of evidence
for a deterministic estimate and can be applied to an assessment in which reservoir simulation is the
integrating tool for predicting recovery. The criteria for assessing a simulation model include:

Production mechanism and appropriate model


Evidence supporting the static and dynamic description
History match and transition to prediction
Test of predictions against performance data or analogues
Model is current
Sensitivities
Development of proved reserves predictions from a best technical case
Documentation

The application of these criteria provide a practical approach for assessing and using simulation in
estimating reserves, as illustrated by the case study.

14

SPE-170669-MS

Acknowledgements
The authors would like to thank Trevor Newley, Dianna Sommer, Roger Skinner and Remco Aalbers for
their advice and help. They would also like to thank BP Exploration, Shell Exploration & Production Co.,
University of Houston and Total S.A. for permission to publish this paper
Nomenclature
2P
proved plus probable reserves
3P
proved plus probable plus possible reserves
FTT formation testing tool
GOR gas oil ratio
HIIP hydrocarbons initially in place
LKO lowest known oil
mbd thousand barrels per day
OWC oil water contact
PRMS Petroleum Resource Management System
SEC United States Securities Exchange Commission

References
Dromgoole, P. and Speers, R. 1997. Geoscore: A method for quantifying uncertainty in field reserves
estimates, Petroleum Geoscience 3 (1):112
Lee, W.J., Sidle, R.E. and McVay, D.A. 2011. Reservoir Simulation: A Reliable Technology? Paper
SPE 146524 presented at the Annual SPE Technical Conference and Exhibition, Denver, Colorado, 30 Oct
2 Nov
Modernisation of Oil and Gas Reporting. Conforming Version. 2008. 17 CFR Parts 210, 211, 229,
and 249 [Release Nos. 33 8995; 34 59192; FR-78; File No. S715 08] US SEC, Washington, DC. (29
December 2008). http://www.sec.gov/rules/final/2008/33 8995.pdf Downloaded 29 May 2014
Petroleum Resource Management System. 2007. Society of Petroleum Engineers, World Petroleum
Council, American Association of Petroleum Geologists, Society of Petroleum Evaluation Engineers.
http://www.spe.org/industry/docs/Petroleum_Resources_Management_System_2007.pdf Downloaded 12
Feb 2012
Rietz, D. and Usmani, A. 2005. Reservoir Simulation and Reserves Classifications Guidelines for
Reviewing Model History Matches to Help Bridge the Gap Between Evaluators and Simulation Specialists, Paper SPE96410 presented at the Annual SPE Technical Conference and Exhibition, Dallas, Texas
9 12 October
Rietz, D. and Usmani, A. 2009. Case Studies Illustrating the Use of Reservoir Simulation Results in
the Reserves Estimation Process, SPE Res Eval & Eng, Feb 2009

Você também pode gostar