Home > Documents > SPE 110066 MS P Simulacion

SPE 110066 MS P Simulacion

Date post: 05-Nov-2015
Category:
Author: mitsumilagrostorosayas
View: 240 times
Download: 0 times
Share this document with a friend
Description:
SPE 110066 MS P SimulacionSPE 110066 MS P SimulacionSPE 110066 MS P SimulacionSPE 110066 MS P SimulacionSPE 110066 MS P Simulacion
Embed Size (px)
of 10 /10
 Copyright 2007, Society of Petroleum Engineers This paper was prepared for presentation at the 2007 SPE Annual Technical Conference and Exhibition held in Anaheim, California, U.S.A., 11–14 November 2007. 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, as presented, have not been reviewed by the Society of Petroleum Engineers and are subject to correction by the author(s). The material, as presented, does not necessarily reflect any position of the Society of Petroleum Engineers, its officers, or members. Papers presented at SPE meetings are subject to publication review by Editorial Committees of the Society of Petroleum Engineers. Electronic reproduction, distribution, or storage of any part of this paper for commercial purposes 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 where and by whom the paper was presented. Write Librarian, SPE, P.O. Box 833836, Richardson, Texas 75083-3836 U.S.A., fax 01-972-952-9435. Abstract Continuous improvements in reservoir simulation software and the availability of high performance computing equipment are making the use of simulation models commonplace for field development and planning purposes. Naturally, this trend has also increased interest in the use of reservoir simulation model results in the oil and gas reserves estimation  process. As simulation specialists who work in a primarily reserves- evaluation company, the authors are routinely asked to evaluate, and in many cases incorporate, simulation results in the reserves es timation process. In addition, the authors are required to opine on the approach and tactics used by clients while they incorporate numerical models in their reserves  bookings. Since there exists limited published discussion on this topic, the purpose of this paper is to provide some examples of the approach used by the authors. We believe this approach to be appropriate and within the spirit of reserves interpretation as used by typical reserves regulatory  bodies such as t he U.S. Securities and Exchange Commission (SEC). Papers previously published have discussed the use of models in the reserves process, including the evaluation of the models' themselves 1,2 . In contrast, this paper provides three case studies that illustrate how results from various models have been used to assist in quan tifying reserves. Two of the examples are based on history matched models while the third focuses on a pre-production reservoir where no adequate history is available and probabilistic methods were incorporated to help understand the uncertainty in the forecasts. While there is no “cookbook” or step-by-step procedure for using simulation results to estimate reserves, the case studies  presented in this paper are intended to both show some examples and also spark some debate and discussion. Undoubtedly there will be some disagreement with our techniques, but an open discussion should prove to be  beneficial for both reserves evaluators and simulation specialists. Introduction The volumes of reserves and the estimated future income attributable to those reserves remains a fundamental tool used for the valuation of an oil and gas project or an entire company. Scrutiny by regulatory bodies such as the U.S. Securities and Exchange Commission (SEC), in the area of reserves compliance has increased during the past few years along with their powers of enforcement through the implementation of the Sarbanes -Oxley act. As a c onsequence, oil and gas companies affiliated with the U.S. and other major stock exchanges have devoted a large amount of time and effort in reviewing their existing reserves and also the process utilized in the booking of reserves. The authors of this paper are employed in a company which is primarily involved with reserves evaluation work. The examples presented in this paper are intended to demonstrate the applications of model adjustments from the perspective of simulation experts who are routinely asked to modify such third party models for reserves compliance. One additional but perhaps very significant item is worth noting. The process for estimating reserves is inheren tly uncertain. Unless we have a special (magical) machine that can take a snapshot and decisively indicate how much oil is in a particular reservoir, volumetric estimates will always be only estimates. Even if we were to k now exactly h ow much is in the ground, the determination of how much is recoverable is also only an estimate. These s tatements are not made to  belittle the importance of reserves estimates. We are all intimately familiar with their importance; rather, we are trying to place the results of our collective work into perspective. This is why, in the authors opinion, the SEC continues to enforce some controversial requirements, such as year end  pricing. It would appear that the SEC is more concerned with the ability for the investing community (individuals or SPE 110066 Case Studies Illustrating the Use of Reservoir Simulation Results in the Reserves Estimation Process Dean Rietz, SPE, and Adnan Usmani, SPE, Ryder Scott Company
Transcript
  • Copyright 2007, Society of Petroleum Engineers This paper was prepared for presentation at the 2007 SPE Annual Technical Conference and Exhibition held in Anaheim, California, U.S.A., 1114 November 2007. 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, as presented, have not been reviewed by the Society of Petroleum Engineers and are subject to correction by the author(s). The material, as presented, does not necessarily reflect any position of the Society of Petroleum Engineers, its officers, or members. Papers presented at SPE meetings are subject to publication review by Editorial Committees of the Society of Petroleum Engineers. Electronic reproduction, distribution, or storage of any part of this paper for commercial purposes 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 where and by whom the paper was presented. Write Librarian, SPE, P.O. Box 833836, Richardson, Texas 75083-3836 U.S.A., fax 01-972-952-9435.

    Abstract Continuous improvements in reservoir simulation software and the availability of high performance computing equipment are making the use of simulation models commonplace for field development and planning purposes. Naturally, this trend has also increased interest in the use of reservoir simulation model results in the oil and gas reserves estimation process.

    As simulation specialists who work in a primarily reserves-

    evaluation company, the authors are routinely asked to evaluate, and in many cases incorporate, simulation results in the reserves estimation process. In addition, the authors are required to opine on the approach and tactics used by clients while they incorporate numerical models in their reserves bookings. Since there exists limited published discussion on this topic, the purpose of this paper is to provide some examples of the approach used by the authors. We believe this approach to be appropriate and within the spirit of reserves interpretation as used by typical reserves regulatory bodies such as the U.S. Securities and Exchange Commission (SEC).

    Papers previously published have discussed the use of models in the reserves process, including the evaluation of the models' themselves1,2. In contrast, this paper provides three case studies that illustrate how results from various models have been used to assist in quantifying reserves. Two of the examples are based on history matched models while the third focuses on a pre-production reservoir where no adequate history is available and probabilistic methods were incorporated to help understand the uncertainty in the forecasts.

    While there is no cookbook or step-by-step procedure for using simulation results to estimate reserves, the case studies presented in this paper are intended to both show some examples and also spark some debate and discussion. Undoubtedly there will be some disagreement with our techniques, but an open discussion should prove to be beneficial for both reserves evaluators and simulation specialists.

    Introduction The volumes of reserves and the estimated future income attributable to those reserves remains a fundamental tool used for the valuation of an oil and gas project or an entire company. Scrutiny by regulatory bodies such as the U.S. Securities and Exchange Commission (SEC), in the area of reserves compliance has increased during the past few years along with their powers of enforcement through the implementation of the Sarbanes-Oxley act. As a consequence, oil and gas companies affiliated with the U.S. and other major stock exchanges have devoted a large amount of time and effort in reviewing their existing reserves and also the process utilized in the booking of reserves.

    The authors of this paper are employed in a company which is primarily involved with reserves evaluation work. The examples presented in this paper are intended to demonstrate the applications of model adjustments from the perspective of simulation experts who are routinely asked to modify such third party models for reserves compliance.

    One additional but perhaps very significant item is worth noting. The process for estimating reserves is inherently uncertain. Unless we have a special (magical) machine that can take a snapshot and decisively indicate how much oil is in a particular reservoir, volumetric estimates will always be only estimates. Even if we were to know exactly how much is in the ground, the determination of how much is recoverable is also only an estimate. These statements are not made to belittle the importance of reserves estimates. We are all intimately familiar with their importance; rather, we are trying to place the results of our collective work into perspective. This is why, in the authors opinion, the SEC continues to enforce some controversial requirements, such as year end pricing. It would appear that the SEC is more concerned with the ability for the investing community (individuals or

    SPE 110066

    Case Studies Illustrating the Use of Reservoir Simulation Results in the Reserves Estimation Process Dean Rietz, SPE, and Adnan Usmani, SPE, Ryder Scott Company

  • 2 SPE 110066

    institutional) to be able to quantitatively compare two different companies as opposed to trying to add precision to a reserves number that will inherently contain uncertainty.

    The authors believe that the above statement is important

    because we are illustrating some methodologies that compromise the rigor of the complex simulation model itself. While we understand this, we also accept it, due to the fact that uncertainties are pervasive throughout the entire reserves estimating process. It is when we step back and look at the use of reservoir modeling results to help arrive at an estimate of reserves, as opposed to a detailed production forecast (still an estimate), that we see the true benefit of reservoir models in the reserves process.

    When estimating reserves using conventional analog data,

    professionals use judgement and experience while incorporating this information3. One should apply the same judgement and experience when making modifications to complex simulation models in order to estimate reserves. This is the basic premise of this paper.

    Reservoirs, Reserves and Models Certain assumptions regarding model input and construction are assumed as a premise for using a models results for the purpose of estimating reserves. These assumptions include the following:

    The seismic, geological and petrophysical input leading to the construction of the simulation model have all been reviewed independently by experts in each respective discipline, and any assumptions made during this process are appropriate and reasonable.

    The level of detail used to construct the model is consistent with the end use of the models results. Generally, the more detail a model contains, such as reservoir description and production history, the higher the level of comfort or confidence associated with the results. Models used to forecast reserves should contain as much detail as is necessary to adequately represent the physics of the reservoir during the anticipated life of the oil or gas field.

    The simulation model history match looks reasonable. This is a multifaceted statement and reference is made to a directly related paper which discusses in detail nine steps used to evaluate a models history match2.

    There is no broad-brush approach used in the modification of simulation models to provide an estimation of reserves. Each case study highlighting the use of a simulation model for reserves is considered individually on a case-by-case basis. The basic principle here is that each reservoir is unique.

    With the recent approval of new SPE/WPC et al guidelines4,

    some dialogue has been published regarding reserves classifications5. For the purpose of this paper, we merely repeat that proved reserves are more likely or have greater certainty to be recovered than probable, and likewise, probable

    reserves are more likely to be recovered than possible reserves. Furthermore, the SEC emphasizes that revisions to proved reserves bookings should be upward as time goes on, rather than downward. It is noted that for SEC reporting purposes only proved reserves can be reported and are therefore relevant. Compliance with Regulatory Guidelines Financial regulatory bodies, such as the SEC, do not provide detailed criteria for the use and application of reservoir simulation models in the reserves estimation process other than their stated position that the model must have a good history match. The use of the term a good history match is often challenging for both the simulation engineer employed to develop the match and the reviewer evaluating the match for compliance.

    Different engineers or evaluators will have varying opinions and methods for determining if a models history match is considered to be reasonable. For example, a given well may actually be producing slightly more water than predicted by the simulation model during the history match period. However, during the prediction forecast, the model results are going to be much more optimistic and the water cut will rise slower than in reality and theoretically be allowed to continue to produce for a longer period of time.

    In this example, the reviewer decided to amend the oil projections from the simulation model by adjusting for the cumulative oil production over time. Generally, the simulation models are history matched based on production volumes from each well associated for the primary phase which is either oil or gas. The issue in this case was not the cumulative oil production to date (during history match) but rather the expected ultimate recovery dictated by the water cut progression.

    Additional work may be necessary to compensate for deficiencies in the history match rather than simply say the model cant be used since it has a poor match. This is where experience and good engineering judgment, aligned with the spirit and intent of the regulatory guidelines, are important.

    Simulation model matches are seldom perfect from all perspectives (pressure, rates, ratios, and cumulatives at both field and at well level), and there are often instances where over production from one well is compensated by under production from another. Production forecast comparisons between traditional methods such as decline curve analysis and numerical simulation remains an integral part of the simulation review process.

    The term reasonable relates to both the input parameters used to construct the model and also the output generated, such as recovery factors. It is highly recommended that the recovery factors generated from the simulation model be compared to analogous fields to keep the results in perspective. From experience, simulation models for immature reservoirs with little or no production history tend to

  • SPE 110066 3

    be optimistic as they often lack the degree of heterogeneity and/or some other intrinsic field dependent parameter which would lower the modeled cumulative volumes and recovery factors.

    Following, we present three examples that incorporate models and their results in the reserves process. Each of these cases highlights a different aspect of this process. Case 1, Modification of Simulator Output This scenario addresses the concept of obtaining an appropriate model and adjusting the model output to account for the reserves-compliant original oil-in-place (OOIP). The following reasons may require the reviewer to follow this path to obtain reserves from the supplied simulation model:

    Only simulation results are provided; that is the actual model (and the ability to re-run the model) is not provided.

    Typically, simulation models are constructed for the 2P or most likely case, as this is thought to be the most likely outcome and the most useful and realistic for field development purposes.

    Time limitations often do not permit the reviewer from making further simulation runs.

    When input parameter adjustments are made to coincide with reserves compliant OOIP values, but these changes render the history match (along with confidence in the models ability to forecast) to be no longer acceptable.

    In Case 1, a simulation model was provided by the client for

    a mature field with several years of production history. The independent reviewer was asked to comment on using projections from the model (prior to economics cashflow) for the purpose of estimating reserves.

    The model construction, history match, and projections were reviewed and were found to be reasonable. As part of this review process, all significant aspects of the model, ranging from input data assumptions to the modeling approach were studied and deemed appropriate. This type of investigation typically involves input from a multidisciplinary team which includes geophysicists, geologists, and petrophysicists6.

    The unadjusted model projection for the existing wells case or Proved Developed Producing (PDP) case is shown in Table 1.

    Well Cum. Oil Production at Existing Wells Case Estimated Remaining Oil

    End of HM at Nov 2005 Oil Recovery at Jan 2031 as of Nov 2005

    MMstb MMstb MMstb

    Well-1 5.679 5.679 0Well-2 6.628 6.628 0Well-3 1.044 1.044 0Well-4 2.969 2.969 0Well-5 0.466 0.466 0Well-6 8.191 8.191 0Well-7 0.007 0.007 0Well-8 3.162 3.162 0Well-9 5.475 8.653 3.179Well-10 3.361 5.835 2.474Well-11 0.751 0.751 0Well-12 0.697 4.325 3.628Well-13 0.002 0.002 0Well-14 0.204 2.077 1.873Well-15 0.263 7.591 7.328Well-16 0 4.652 4.652

    Total 38.897 62.031 23.134

    As a quick check, note that the projected recoveries from each of the wells are not unreasonable. The projections from all the model wells do not exceed the highest cumulative already observed. Although this, in itself, does not make the model reasonable, it does however, provide some comfort that the model forecasts are within a sensible range of realistic recoveries.

    The only major issue was that the simulation model OOIP (323.8 MMstb) did not conform to either the independently estimated volumetric 1P (290.8 MMstb) or the volumetric 2P (401.4 MMstb) OOIPs. As the model OOIP was slightly above the independently verified 1P OOIP, the reviewers concluded that in this instance, there was indeed a strong case for treating the simulation model as a representative analogy to the field and adopting the model recovery factors and applying them to the lower certified 1P OOIP.

    Fig. 1 shows the procedure for calculating the expected recovery for the 1P OOIP, assuming the same recovery factor is applied as in the reviewed models case. Note that the recovery factor used in this case (19.16%) was based upon an appropriate OOIP. That is, the model was checked to ensure that no undrained volumes (fault blocks or horizons) were included in the calculation of the simulation forecasted recovery factor.

    TABLE 1UNADJUSTED PROJECTIONS, (PDP CASE)

  • 4 SPE 110066

    Fig. 1Example of recovery factor application.

    The annual modified projections were estimated using a spreadsheet. An approximate additional 7.1% annual field oil decline rate was applied to the model projections (actual field oil decline rate of approximately 24.0%). Both the adjusted and the unadjusted case field oil rates and cumulatives are shown in Table 2.

    Figs. 2 and 3 show the results graphically and compare both the unadjusted and adjusted forecasts from the model. This is shown for both oil rate and cumulative oil.

    Fig. 2Graphical representation of adjusted and unadjusted oil rate projection. Fig. 3Graphical representation of adjusted and unadjusted cumulative oil projection.

    Note that to preserve the reasonable forecast at the start of the prediction period, the additional decline was applied a few years (five in this case) after the end of the history match. This makes sense since you would expect the forecast near the end of the history match (having properly calibrated the model) to have greater certainty than projections in the more distant future.

    We need to be careful not to over extend the application of the recovery factor to an OOIP significantly different or dissimilar in geologic or stratigraphic distribution. That is, if the simulation model OOIP is 100 MMstb, then we cannot simply apply the model derived recovery factor to a reservoir containing 500 MMstb as the field performance and reservoir dynamics are likely to be significantly different. In this case the model unadjusted OOIP (323.8 MMstb) was quite close to the independently verified volumetric 1P OOIP (290.8 MMstb).

    Model Results} OOIP = 323.760 MMstb} EUR Cum Oil = 62.031 MMstb} Recovery Factor = 19.16 %

    Volumetric from Independent 3rd Party (SEC Proved)} OOIP = 290.760 MMstb} Recovery Factor applied = 19.16 %} Recovery = 55.708 MMstb

    Model EUR Revision at 1/1/2031} For PDP Case 62.031 MMstb 55.708 MMstb

    YEAR Simulation Simulationof Production Avg Yearly Rate Projected Cum

    stb/day MMstbPDP PDP

    2005 39.7022006 8,458 42.7922007 5,791 44.9072008 4,635 46.6002009 4,069 48.0862010 3,196 49.2532011 2,626 50.2122012 2,422 51.0972013 2,268 51.9252014 2,152 52.7112015 2,053 53.4612016 1,971 54.1812017 1,885 54.8702018 1,812 55.5312019 1,746 56.1692020 1,689 56.7862021 1,629 57.3812022 1,577 57.9572023 1,529 58.5152024 1,488 59.0592025 1,443 59.5862026 1,405 60.0992027 1,370 60.5992028 1,341 61.0892029 1,305 61.5662030 1,273 62.031

    Unadjusted Simulation

    Results

    Oil RateYear

    r Projected Rate Projected CumSTB/DAY MMstb

    PDP PDP

    39.7028,458 42.7925,791 44.9074,635 46.6004,069 48.0862,970 49.1702,267 49.9981,943 50.7081,691 51.3251,490 51.8701,321 52.3521,178 52.7821,047 53.165935 53.507837 53.812753 54.087674 54.334607 54.555546 54.755494 54.935445 55.098403 55.245365 55.378332 55.499300 55.609272 55.708

    Adjusted (PDP) Simulation

    Results

    Oil Rate Oil CumOil Cum

    YEAR Simulation Simulationof Production Avg Yearly Rate Projected Cum

    stb/day MMstbPDP PDP

    2005 39.7022006 8,458 42.7922007 5,791 44.9072008 4,635 46.6002009 4,069 48.0862010 3,196 49.2532011 2,626 50.2122012 2,422 51.0972013 2,268 51.9252014 2,152 52.7112015 2,053 53.4612016 1,971 54.1812017 1,885 54.8702018 1,812 55.5312019 1,746 56.1692020 1,689 56.7862021 1,629 57.3812022 1,577 57.9572023 1,529 58.5152024 1,488 59.0592025 1,443 59.5862026 1,405 60.0992027 1,370 60.5992028 1,341 61.0892029 1,305 61.5662030 1,273 62.031

    Unadjusted Simulation

    Results

    Oil RateYear

    r Projected Rate Projected CumSTB/DAY MMstb

    PDP PDP

    39.7028,458 42.7925,791 44.9074,635 46.6004,069 48.0862,970 49.1702,267 49.9981,943 50.7081,691 51.3251,490 51.8701,321 52.3521,178 52.7821,047 53.165935 53.507837 53.812753 54.087674 54.334607 54.555546 54.755494 54.935445 55.098403 55.245365 55.378332 55.499300 55.609272 55.708

    Adjusted (PDP) Simulation

    Results

    Oil Rate Oil CumOil Cum

    TABLE 2ADJUSTED AND UNADJUSTED PROJECTIONS

    stb/day

    Case Study 1: Projected Oil Rate Forecast

    100

    1,000

    10,000

    100,000

    2000 2005 2010 2015 2020 2025 2030

    Year

    Oil

    Rat

    e st

    b/da

    y

    PDPHistorical

    Model Output

    Case Study 1: Projected Cumulative Oil Forecast

    20.0

    25.0

    30.0

    35.0

    40.0

    45.0

    50.0

    55.0

    60.0

    65.0

    2000 2005 2010 2015 2020 2025 2030

    Year

    Cum

    ulat

    ive

    Oil

    MM

    stb

    PDP

    Historical

    Model Output

  • SPE 110066 5

    The following are some of the points that need to be considered when comparing models and actual fields with significantly different OOIP (Case 1):

    Geology and field oil distribution Well count and placement for both producers and

    injectors Field production and injection rates (facilities

    constraints) Field concession duration Operational philosophy/scheme Drive mechanism

    The reader is encouraged to refer to recent publications on

    the use of analogs for use in the estimation of reserves3. Case 2, Modification of Model Input As mentioned earlier in the discussion for Case Study 1, most simulation models are constructed for the 2P or the most likely reservoir case; conversely this case (Case 2), although not common, allowed the evaluators to make minor revisions to the model such that the model would be compliant for the 1P scenario.

    In this scenario, the model input is modified such that the OOIP is compliant with reserves guidelines as stipulated by the regulatory body.

    For Case Study 2, the consultant received a simulation model for a mature oil field constructed by the client (Model 1) for a Gulf of Mexico asset which did not comply with SEC regulations for volumetrically determined OOIP. This volatile oil field was evenly divided into two distinct and non-communicating reservoirs. This field contained a total of six producing wells and the oil-water contacts (OWCs) were defined by other non-producing, down-dip wells. The fact that the OWC was known and defined by a log facilitated the correction of this model to be a compliant 1P model.

    In order to satisfy the regulatory conditions, the simulation model input geology was amended. The following structural and geological changes were made which were expected to reduce the model OOIP:

    No reservoir thickening between the wells OOIP of Model 2 independently verified by 3rd party

    to comply with SEC proved volumetric criteria

    Having made the model (Model 2) input geology SEC compliant, an attempt was then made to history match the simulation model. The following changes were necessary in order to reach a reasonable and defendable history match:

    Further reductions to the OOIP were needed to match the average reservoir pressure (volumetric match).

    Minor adjustments were needed to match well productivity.

    The fact that the history matched model OOIP is somewhat

    lower than the independently verified OOIP provides:

    Added value of the model to arrive at a more accurate OOIP (as it was based on more data from the field).

    Higher level of confidence in the reserves for SEC compliance (absence of model may have resulted in future reserves reduction).

    A good history match was obtained from the model at both

    the field and well level for the following: Average field pressure for both reservoirs Field production rates, ratios and cumulatives Well production rates, ratios and cumulatives

    Figs. 4, 5 and 6 graphically show the quality of the model

    history match (red solid line) obtained from the model compared to actual reservoir observed data (blue marker points). The model was then calibrated and run in prediction mode. (There are no marker points for comparison in prediction mode). These three figures illustrate the matches for the average reservoir pressure (Fig. 4), field production rates and cumulative volumes (Fig. 5), and productions rates, cumulatives and ratios for one of the producing wells (Fig. 6).

    Fig. 4Average reservoir pressure match.

    Fig. 5Field production rates and cumulative volumes.

  • 6 SPE 110066

    Fig. 6Example well production rates, ratios and cumulatives.

    The simulation projections from this model (Model 2) were then compared to the decline curve analysis (DCA) projections from each of the wells as part of the model review process for reasonableness. The model and DCA projections were found to be in close agreement, providing an additional degree of comfort with the modeling results. Based on the above, the model forecasts were deemed appropriate and used to provide Proved projections. Note that these projections were subject to economic screening in order to be classified as reserves. Further validation of the model was supported by the production performance of a newly drilled infill well that was accurately predicted by the simulation model.

    The premise of this approach is that a good history matched model that is built to comply with 1P (SEC or other regulatory body) OOIP (or OGIP) is appropriate for estimating reserves (subject to economic constraints).

    In summary, for mature reservoirs which are similar to the scenario describing Case Study 2:

    The use of simulation models should be considered as a mathematical analog to the field in question, with the appropriate usage of all reservoir properties. Hence, in order for a field to be considered as an analog field for water-flood purposes under SEC guidelines, the following six reservoir properties need to be better or at least the same in the field under scrutiny compared to the analog (in addition to some other parameters such as fluid properties):

    1. Porosity 2. Permeability 3. Permeability distribution 4. Sand thickness 5. Reservoir continuity 6. Hydrocarbon saturation

    Although not explicitly defined, it is common practice to apply these criteria when considering all analogs (not just waterflood scenarios).

    For a model to be used for reserves, it is suggested that the model input be either known from actual field observations or toward the higher confidence-side of the range of estimated parameters.

    Specifically for SEC proved, the model should reflect performance that is consistently on target or slightly lower than the actual field.

    Simulation models constrained by history match are likely to produce reasonable field/well recovery factors.

    Simulation models with adjusted input (as in Case 2) which still show a good history match increase the level of comfort associated with the model.

    Greater confidence in results at the field level rather than at the individual well level.

    Greater uncertainty associated with undeveloped locations and un-drained areas.

    May want to consider running a sensitivity analysis similar to Case 3 for undrained fault blocks.

    Once again, models that are both reliable and compliant

    with regulatory proved volumes are rare. Usually the greatest (but not the only) uncertainty and divergence between a reliable model with a good history match and 1P compliance is the OWC or GWC, which in this case was determined from actual log measurements.

    In situations where the model history match for a mature reservoir required an adjustment above the compliant OOIP and where there is virtually no other way to match observed reservoir behavior, the model still could be a reliable tool for estimating proved reserves. In these cases there should exist compelling evidence that the model is representative (reasonable predictions, and a high quality history match) and that the model results are being applied in a manner consistent with the use of traditional production trend analysis, even though the original hydrocarbon in-place cannot be fully reconciled with the reserves definitions1. Case 3, Limited or No Production History Field This scenario is very important as it addresses the use of simulation models for reserves from immature fields with little or no sustained production history. Simulation models are commonly used as a powerful field development tool containing not only the subsurface information but are also increasingly performed in combination with a surface production/injection network. These models employ the use of real-time nodal analysis at each time-step, therefore providing an integrated estimate of reservoir performance and production network deliverability.

    A 3D geological model based upon the data provided by the client and seismic interpretation was constructed which was subsequently used in the simulation model.

    Because these simulation models contain little or no history

    match (comparison to actual field performance) data, the results from a single deterministic case need to be put in perspective of the entire range of likely outcomes (possible ultimate recoveries and recovery factors).

    In the Case Study 3 example, a simulation model was constructed by the consultant for an oil field situated in a

  • SPE 110066 7

    frontier region. This field was separated into two major fault blocks (Central and Western) with similar geology and fluid properties. The depth of the oil water contact (OWC) was known for the Central fault block but was unknown for the Western fault block. Depth of the OWC was run as part of a sensitivity study. A probabilistic Monte Carlo study was proposed to help estimate the reserves for this field.

    Note that the authors of this paper are not experts in the area of probabilistic techniques, and the example presented is one of many currently available and utilized by others in the industry. The reader is reminded that the goal of this paper is to illustrate reasonable and defendable projections to be used for reserves; but, in this specific case (Case 3), we also incorporate probabilistic methods which result in the generation of ranges of recoveries. This approach will certainly help capture the range of uncertainty much better than a few deterministic models.

    The authors are aware of other more sophisticated probabilistic techniques such as Experimental Design but lack first hand experience in this area. The authors used single parameter changes to the simulation model for each sensitivity run which some emerging methods suggest as being outdated. However, the authors have performed some non-exhaustive test cases to account for sensitivity interdependencies. (In both of these cases, the proxy function was tuned to the respective run results).

    The overall Monte Carlo results from both the single parameter variation and the more complex model using multi-component adjustment were essentially indistinguishable. Overall, no matter which probabilistic tool is used to conduct the study, the results should be repeatable using any reasonable and defendable probabilistic technique.

    The range of outcomes is investigated by creating an initial base case simulation model containing the most likely input parameters. For this portion of the evaluation, only waterflood operations are investigated which would be considered 2P (proved plus probable). This 2P designation is due to both recovery techniques (water injection) and a geological description that significantly exceeds 1P volumes. The next step is to run sensitivity (both higher and lower if possible) for each parameter, e.g., rock compressibility. Each simulation was performed by varying only a single parameter from the base case to see the impact of that parameter on the results. The sensitivity study provides an opportunity to evaluate and specify ranges for all major input parameters, such as geological, petrophysical, rock, PVT, initial reservoir pressure, dynamic simulation, surface facilities and production/injection rates, in order to fully understand the effect. The uncertainty in measurement was used to create a range (high and low) for each parameter. This step is where emerging technologies, such as Experimental Design, can be employed.

    A plot containing the results for ultimate (model cumulative) oil production (Western and Central fault blocks) from all of the sensitivity cases is shown below in Fig. 7.

    Fig. 7Cumulative oil production from all of the sensitivity cases.

    The solid red line was the initially assumed (2P) base case;

    the lines in blue refer to the sensitivity cases which define the upper and lower range for all the possible ultimate recovery volumes and the lines in green were the results from all the other sensitivity cases. The gray line below the lower blue line represents the unlikely case where the operator is unable to provide water for injection purposes. The ability to be able to provide an adequate amount of water to meet voidage replacement requirements is a reasonable assumption based upon experience from neighboring projects.

    The results from Fig. 7 are shown as a tornado plot in Fig. 8 with the x-axis displaying differences in ultimate recovery from the base case. + Fig. 8The results of the sensitivity cases in a tornado plot.

    Monte Carlo Methodology Having run the desired number of sensitivity cases, it was decided to run a probabilistic Monte Carlo study to assess the likely range of outcomes for ultimate recoveries and recovery factors.

    Western & Central Cumulative Oil Recovery Comparison

    0

    10,000

    20,000

    30,000

    40,000

    50,000

    60,000

    70,000

    80,000

    90,000

    100,000

    Jan 2

    009

    Jan 2

    010

    Jan 2

    011

    Jan 2

    012

    Jan 2

    013

    Jan 2

    014

    Jan 2

    015

    Jan 2

    016

    Jan 2

    017

    Jan 2

    018

    Jan 2

    019

    Jan 2

    020

    Jan 2

    021

    Jan 2

    022

    Jan 2

    023

    Jan 2

    024

    Jan 2

    025

    Jan 2

    026

    Jan 2

    027

    Jan 2

    028

    Jan 2

    029

    Jan 2

    030

    Jan 2

    031

    Jan 2

    032

    Date

    Cum

    ulat

    ive

    Oil,

    Mst

    b

    1/1/2019

    Base Case M26_BASE_3P2I Displayed in Red

    J05_SENS_HI_OWC

    J05_SENS_LOW_VREP

    J14_SENS_LO_FACIES_LO_VCL

    Western & Central Cumulative Oil Difference from Base Case (M26_BASE_3P2I) at 1/1/2019

    -30,000 -25,000 -20,000 -15,000 -10,000 -5,000 0 5,000 10,000 15,000 20,000 25,000 30,000

    1

    Sens

    itivi

    ty S

    cena

    rios

    Cumulative Oil Difference from Base Case, Mstb

    J14_SENS_INJ_OFF_CJ14_SENS_INJ_OFF_WJ14_SENS_LO_FACIES_LO_VCLJ07_SENS_LO_CONNECTIVITYJ07_SENS_HI_CONNECTIVITYJ02_SENS_LO_GASLIFTJ02_SENS_HI_GASLIFTJ02_SENS_LO_THPJ02_SENS_HI_THPJ05_SENS_LOW_VREPJ05_SENS_MID_VREPJ02_SENS_LO_PROD_WELL_RATESJ02_SENS_HI_PROD_WELL_RATESJ02_SENS_LO_INJ_WELL_RATESJ02_SENS_HI_INJ_WELL_RATESJ02_SENS_LO_WELL_SKINJ02_SENS_HI_WELL_SKINJ02_SENS_LO_WATER_ENDPTJ02_SENS_HI_WATER_ENDPTJ03_SENS_LO_WATER_COREYJ03_SENS_HI_WATER_COREYJ03_SENS_LO_OIL_COREYJ03_SENS_HI_OIL_COREYJ06_SENS_LO_VISCJ06_SENS_HI_VISCM30_SENS_LO_CRITGASM30_SENS_HI_CRITGASM30_SENS_LO_CONNATEM30_SENS_HI_CONNATEM30_SENS_LO_SORM30_SENS_HI_SORM30_SENS_LO_ROCKCRM30_SENS_HI_ROCKCRJ06_SENS_LO_FAULT_SEALJ06_SENS_MID_FAULT_SEALJ06_SENS_LO_VCLJ06_SENS_HI_VCLJ05_SENS_LO_FACIESJ05_SENS_LOW_NTGJ05_SENS_HIGH_NTGJ05_SENS_LOW_OWCJ05_SENS_HI_OWCM30_SENS_LO_KHM30_SENS_HI_KHM30_SENS_LO_KVKHM30_SENS_HI_KVKHM26_BASE_3P2I_54614M26_BASE_3P2I_42001M26_BASE_3P2I_25252M26_BASE_3P2I_79086M26_BASE_3P2I

  • 8 SPE 110066

    Aspects of Monte Carlo Simulation or Study: The Monte Carlo simulation generates a realization

    of the reservoir properties that is conditional to ALL input data.

    The Monte Carlo simulation varies the input parameters within their defined ranges and uses the combination of all input parameters.

    Some input parameters such as permeability are best modeled using a log-normal distribution, others use a triangular distribution.

    Monte Carlo requires a (proxy) function to be created that can be used to predict the simulation results.

    Without utilizing a proxy function with Monte Carlo techniques, 50,000 numerical reservoir simulation runs each lasting 8 hours would take approximately 45 years of run time!

    In order to perform a Monte Carlo study using reservoir

    simulation in a timely manner, a proxy function is used in lieu of running the numerical simulation model many times. This proxy function is (hopefully) able to predict the expected simulation model results in between the end points for each sensitivity parameter without having to re-run the simulator, as it would not be realistic to run the simulation model 50,000 times to conduct the Monte Carlo study. Note that the function will not extrapolate the sensitivity parameters beyond the defined probabilistic range.

    The Monte Carlo proxy function was calculated using the built-in solver utility in Microsoft Excel to minimize the error in the function between the actual simulation results and those predicted by the function using the same input parameters.

    Graphs of results from the reservoir simulation versus results from the Monte Carlo proxy function were used to verify the validity of the function as illustrated in Fig. 9. Ideally all function points should lie on a straight line sloping from the origin at 45 degrees (directly proportional). Fig. 9Evaluating the quality of the Monte Carlo function.

    Monte Carlo Results The results from the Monte Carlo were generated in the form of a distribution which is shown in Fig. 10. Fig. 10Results from the Monte Carlo simulation.

    The Monte Carlo study gives rise to the following observations:

    The Monte Carlo Probabilistic P50 case was approximately 4.6 MMstb less than the current Base Case model.

    This indicates that the parameters chosen to define the base case were more optimistic than the P50 Monte Carlo results.

    Parameters of greatest uncertainty: o Reservoir pressure maintenance (voidage

    replacement with water injection) o Oil Water Contact (Western FB) o NTG Distribution

    Visually, the tornado plot supports the results (shown in Fig. 11).

    The revised tornado plot with the addition of the P50 Monte

    Carlo results is shown in Fig. 11. Fig. 11Base case compared to the Monte Carlo P50.

    Western Cumulative Oil Regression Fit

    0

    10

    20

    30

    40

    50

    60

    0 10 20 30 40 50 60

    Simulator Cum Oil MMstb at 1/1/2019

    Func

    tion

    Cum

    Oil

    MM

    stb

    at 1

    /1/2

    019

    Good Match

    Western FB Simulated vsFunction Predicted Cum Oil

    Western Cumulative Oil Regression Fit

    0

    10

    20

    30

    40

    50

    60

    0 10 20 30 40 50 60

    Simulator Cum Oil MMstb at 1/1/2019

    Func

    tion

    Cum

    Oil

    MM

    stb

    at 1

    /1/2

    019

    Good Match

    Western FB Simulated vsFunction Predicted Cum Oil Western & Central Cumulative Oil Difference from Base Case (M26_BASE_3P2I) at 1/1/2019

    -30,000 -25,000 -20,000 -15,000 -10,000 -5,000 0 5,000 10,000 15,000 20,000 25,000 30,000

    1

    Sens

    itivi

    ty S

    cena

    rios

    Cumulative Oil Difference from Base Case, Mstb

    J14_SENS_INJ_OFF_CJ14_SENS_INJ_OFF_WJ14_SENS_LO_FACIES_LO_VCLJ07_SENS_LO_CONNECTIVITYJ07_SENS_HI_CONNECTIVITYJ02_SENS_LO_GASLIFTJ02_SENS_HI_GASLIFTJ02_SENS_LO_THPJ02_SENS_HI_THPJ05_SENS_LOW_VREPJ05_SENS_MID_VREPJ02_SENS_LO_PROD_WELL_RATESJ02_SENS_HI_PROD_WELL_RATESJ02_SENS_LO_INJ_WELL_RATESJ02_SENS_HI_INJ_WELL_RATESJ02_SENS_LO_WELL_SKINJ02_SENS_HI_WELL_SKINJ02_SENS_LO_WATER_ENDPTJ02_SENS_HI_WATER_ENDPTJ03_SENS_LO_WATER_COREYJ03_SENS_HI_WATER_COREYJ03_SENS_LO_OIL_COREYJ03_SENS_HI_OIL_COREYJ06_SENS_LO_VISCJ06_SENS_HI_VISCM30_SENS_LO_CRITGASM30_SENS_HI_CRITGASM30_SENS_LO_CONNATEM30_SENS_HI_CONNATEM30_SENS_LO_SORM30_SENS_HI_SORM30_SENS_LO_ROCKCRM30_SENS_HI_ROCKCRJ06_SENS_LO_FAULT_SEALJ06_SENS_MID_FAULT_SEALJ06_SENS_LO_VCLJ06_SENS_HI_VCLJ05_SENS_LO_FACIESJ05_SENS_LOW_NTGJ05_SENS_HIGH_NTGJ05_SENS_LOW_OWCJ05_SENS_HI_OWCM30_SENS_LO_KHM30_SENS_HI_KHM30_SENS_LO_KVKHM30_SENS_HI_KVKHM26_BASE_3P2I_54614M26_BASE_3P2I_42001M26_BASE_3P2I_25252M26_BASE_3P2I_79086M26_BASE_3P2I

    Mean of Combined West & Central Cumulative Oil Distribution

    Combined Central and Western Block, Cumulative Oil Recovery Regression Fit

    50,000 Monte Carlo runs performed

    Combined Central and Western Block, Cumulative Oil Recovery Regression Fit

    50,000 Monte Carlo runs performed

    P50 of combined Western and Central cumulative oil distribution

  • SPE 110066 9

    Estimation of Reserves for Case 3 The current project has not yet been sanctioned by the operator. If it were, then the following reserves would be recommended to be booked (subject to economic constraints). Proved (1P)

    1P reserves should only be booked based upon a compliant drive recovery factor and 1P independently verified (SEC compliant) OOIP.

    For this case, a single deterministic run was made with the above base case simulation model using a depletion drive recovery mechanism. The projection for this case yielded a 15% recovery factor and is presented for comparison with the 2P recovery factor estimates in Fig. 12.

    An independently verified (SEC compliant) 1P OOIP (approximately 100 MMstb) combined with the depletion drive recovery factor (approximately 15% taken from the model) = 15 MMstb.

    Overlying theme is consistent with SEC guidelines - Reasonable Certainty: The concept of reasonable certainty implies that, as more technical data becomes available, a positive, or upward, revision is much more likely than a negative, or downward, revision. 7

    Proved Plus Probable (2P)

    Could use the results of the P50 from the Monte Carlo = 58.1 MMstb.

    Western & Central Oil Recovery Factor Comparison

    0

    10

    20

    30

    40

    50

    60

    70

    Jan 2

    009

    Jan 2

    010

    Jan 2

    011

    Jan 2

    012

    Jan 2

    013

    Jan 2

    014

    Jan 2

    015

    Jan 2

    016

    Jan 2

    017

    Jan 2

    018

    Jan 2

    019

    Jan 2

    020

    Jan 2

    021

    Jan 2

    022

    Jan 2

    023

    Jan 2

    024

    Jan 2

    025

    Jan 2

    026

    Jan 2

    027

    Jan 2

    028

    Jan 2

    029

    Jan 2

    030

    Jan 2

    031

    Jan 2

    032

    Date

    Rec

    over

    y Fa

    ctor

    , per

    cent

    2P Base Case Maximum 2P Case

    Minimum 2P Case

    Natural Depletion Case (1P)

    Fig. 12Recovery factors for 1P and 2P cases. In General for 1P Reserves

    Use model generated depletion recovery factor (approximately 15% for this case) that is consistent with reasonable certainty.

    Use P90 of model-generated recovery factors from simulation sensitivity runs (using only compliant drive mechanisms, e.g., depletion only).

    In some cases, the P50 recovery factor may be appropriate; however, this would strongly depend on many factors, including but not limited to, existing data, range of recovery factors, available analogs, compliance of modeled drive mechanism, and operational considerations. The authors believe that

    although there may be a certain degree of confidence by the operator to achieve the P50 recovery factor, the likelihood of this being accepted by a regulatory body such as the SEC can be very challenging. That is, since one must have compelling and convincing evidence to book reserves based on the P50 value (even though applied to a 1P OOIP), we would at this time consider this to be an extreme upper end (for proved reserves).

    If waterflood response has been demonstrated, the above approach can apply, except cases now include waterflood projection.

    Nearby physical analogs are always useful to help verify the model and its results.

    Case Study 3 - Summary The results obtained from models constructed for immature reservoirs with no sustained production history as a guide for field performance must be compared to analogs with respect to production rates and oil recovery factors. This is precisely why these types of unconstrained models are recommended to be simulated with a wide array of sensitivity scenarios to better understand the range of possible recoveries without unwarranted over-reliance on a single deterministic simulation model.

    The authors of this paper propose the following display

    (Fig. 13) which links the classification of the model input to the output results obtained. For example, if the simulation model used in the probabilistic study contained an OOIP considered to be approximately 2P or most likely case (as in our Case Study 3), then the possible range of reserves (prior to screening for economics) would be between the P90 and P50 depending on the degree of certainty and confidence in the model for the 2P case. As mentioned earlier, each case is unique and is considered on an individual basis, and there is no simplistic approach for all reservoirs. The level of the reviewers experience in dealing with such scenarios and their degree of professional expertise are major factors in ultimately deciding how the reserves from the reservoir are classified and categorized.

    P90 P50 P10

    1PProved (SEC etc) (Meets reasonable

    certainty)

    Proved? Probable?

    (If Proved, then likely to be extreme upper end

    and analog may be required)

    Probable?

    2P Probable (low end)

    ProbablePossible?

    (or high end of Probable)

    3P Possible (low end)

    PossibleResources?

    (or high end of Possible)

    Independently Verified In-Place HC

    Recovery Factor Probability / Level of Confidence

    Fig. 13Current thinking for using simulation combined with probabilistic methods to estimate reserves.

  • 10 SPE 110066

    These Examples are a Snapshot of Current Thinking Various reasonable approaches have been proposed by the authors to enable the simulation specialist or reserves evaluator to modify and or adjust simulator results for the purpose of estimating reserves. The proposed methods are just some examples of the current thinking in this emergent area of discussion and any suggestions are not meant to be regarded as a rule.

    The SEC in particular has not provided any public feedback on the published papers in the area of simulation and reserves. (The authors recognize the lack of a forum for the SEC to comment publicly). Generally, the only time one can expect to receive feedback from the SEC is privately when they have an issue with a reserves booking.

    It should be noted that due to the unique nature of each field, the reserves associated with that field must be reviewed on a case-by-case basis. A cookie-cutter approach cannot and must not be used for the purpose of estimating reserves. Approaches such as these proposed by the authors will likely change and/or progress with further discussion. Both modeling and reserves methods will also likely change or mature.

    The authors of this paper welcome further discussion and/or suggestions by other simulation or reserves specialists, particularly with respect to probabilistic modeling.

    Conclusions It is often impractical to build models solely for the purpose of estimating reserves. Models not built for estimation of reserves may still be used for reserves determination as demonstrated in the above case studies, by the modification of results, modification of input, and through the use of probabilistic sensitivity studies. Some models, depending on construction techniques (geology, history match, etc.) may be unsuitable for quantitative reserves forecasting regardless of what adjustments are applied.

    Further work needs to be conducted by both the industry and regulatory bodies to definitively prescribe procedures for the use of models and probabilistic methods in the reserves process.

    The concept of considering a simulation model as an analog is easier to apply to a mature field (containing at least a few years of documented pressure and rate history) with more check-points to validate the comparison rather than to an immature field with little to no production history.

    In the case of the immature reservoir, the range of uncertainty in the results should be adequately reflected by the range of hydrocarbon ultimate recoveries. Furthermore, the range of results from all simulations can be used as a key indicator to highlight the relative position of the selected base case, i.e. is the assumed base case considered to be optimistic or pessimistic in regard to cumulative hydrocarbon production?

    Acknowledgements The authors would like to express their gratitude to fellow Ryder Scott colleagues Don Roesle and Bob Wagner for their comments and suggestions during the preparation of this paper. We also acknowledge the assistance of Pamela Nozza and Carl Schorken for their help in the preparation of this manuscript and conducting the simulation runs. References 1. Palke M.R., Rietz D.C.: The Adaptation of Reservoir Simulation

    Models for Use in Reserves Certification Under Regulatory Guidelines or Reserves Definitions, paper SPE 71430 presented at SPE Annual Technical Conference and Exhibition in New Orleans, September 2001.

    2. Rietz D.C., Usmani A.: Reservoir Simulation and Reserves Classifications Guidelines for Reviewing Model History Matches to Help Bridge the Gap Between Evaluators and Simulation Specialists, paper SPE 96410 presented at SPE Annual Technical Conference and Exhibition in Dallas, October 2005.

    3. Hodgin J.E., Harrell D.R.: The Selection, Application and Misapplication of Reservoir Analogs for the Estimation of Petroleum Reserves, paper SPE 102505-PP presented at SPE Annual Technical Conference and Exhibition in San Antonio, September 2006.

    4. 2007 Petroleum Reserves and Resources Classification, Definitions, and Guidelines approved by SPE, WPC, AAPG and SPEE, March, 2007: http://www.spe.org/spe-app/spe/industry/reserves/index.htm

    5. Etherington J.R., Ritter, J.E.: The 2007 SPE/AAPG/WPC/SPEE Reserves and Resources Classification, Definitions and Guidelines. Defining the Standard!, paper SPE 107693 presented at SPE Hydrocarbon Economics and Evaluation Symposium in Dallas, April 2007.

    6. Rietz, D.C. and Palke, M.R. : History matching helps validate reservoir simulation models, Oil and Gas Journal, p. 47, December 24, 2001.

    7. U.S. Securities and Exchange Commision, Division of Corporation Finance: Current Accounting and Disclosure Issues, p. 51, June 30, 2000.

    http://www.sec.gov/pdf/cfcr072k.pdf


Recommended