General Method For The Consistent Volume Assessment of Complex Hydrocabon Traps

You might also like

Download as pdf or txt
Download as pdf or txt
You are on page 1of 13

Journal of Petroleum Geology, Vol.

35(1), January 2012, pp 85-98

85

A GENERAL METHOD FOR THE CONSISTENT VOLUME ASSESSMENT OF COMPLEX HYDROCARBON TRAPS

A. Beha1*, J.E. Christensen1 and R.Young2

Complex hydrocarbon traps are those in which a number of different trapping elements have to work favourably and simultaneously in order to allow access to the full hydrocarbon volume potential of the prospect. The probability of success varies across the prospect and the volume distribution. A consistent assessment of the probability of success relative to the volume uncertainty of hydrocarbon traps is essential for unbiased prospect characterisation and vital for decision making, portfolio management and for delivering predicted value. It is relatively straightforward to assess the probability of geological success and volume uncertainty of simple anticlines or domal structures. Unfortunately, simple four-way closures are usually drilled early in the exploration of a basin or play and become increasingly less common with exploration maturity. As a consequence, prospects available for drilling in mature exploration areas are typically complex traps, i.e. they possess a combination of different trapping elements. In such cases, trapping of the full volume potential requires that many geological elements work concurrently. Complex traps are often perceived to carry a comparatively lower probability of geological success. However, the introduction of additional factors in a multiplicative chance estimation model may not reflect the true probability of finding hydrocarbons at the prospect location. Evaluations which involve multiplying additional chance factors may lead to an under-estimation of the probability of geological success and an over-estimation of the hydrocarbon volume. A solution to this problem is to calculate the probability of occurrence of each possible success scenario or outcome.This paper describes a straightforward method for assessing volume uncertainty in complex traps which is independent of the model or method used to estimate the probability of geological success. A faulted four-way closure is used as an example for the detailed description of the suggested workflow.
INTRODUCTION The uncertainty of estimated hydrocarbon volumes in a prospect is most commonly captured using
1 DONG E&P, Agern All 24-26, DK-2970 Hrsholm, Denmark. 2 Rose and Associates, LLP, 4203 Yoakum Boulevard, Suite 320, Houston, 77006, USA. * Corresponding author: anbeh@dongenergy.dk

stochastic evaluation methods. Although the mathematical foundations of many of these methods are comparable, the estimated volumes of a potential hydrocarbon trap will vary depending on factors including the type of assessment method used, bias rooted in a particular exploration company and, to a
Key words: complex hydrocarbon traps, prospect assessment, probability of well success, fault seal failure, risk analysis.

2012 The Authors. Journal of Petroleum Geology 2012 Scientific Press Ltd

86

Volume assessment of complex hydrocarbon traps

certain degree, the bias of individual explorationists within the company. A widely accepted technique is to link an assessment of the probability of geological success (POS) with the assessment of the potential hydrocarbon volume (see for example, Young et al., 2005; Citron et al., 2006). However, guidelines for preferred probability distributions for all volume parameters as well as definitions of the model for estimating the probability of geological success may vary from company to company. Estimates of the POS and trap size may therefore differ significantly, even between companies in the same joint venture evaluating the same acreage with the same data. A variety of mathematical modelling tools for prospect assessment is available. A typical model output will include the POS and a range of possible success-case hydrocarbon volumes. The latter is usually expressed by the mean value of the uncertainty distribution and a pre-defined subset of percentiles of the cumulative probability curve of which P90, P50 and P10 are the most common. (Throughout this paper, the authors use the greater than or equal to convention, i.e. P99 is the smallest outcome and P1 is the largest). The methodology described below focuses on a best practice for volume assessment of complex hydrocarbon traps. It can be used independently of a specific model for estimating the chance of geological success, and hence is relatively easy to implement. The only prerequisite required is a definition of the link between the POS and the volume uncertainty. In this paper, the POS is defined as the probability that there will be a minimal but sustained flow of hydrocarbons. The POS therefore represents the chance of equalling or exceeding the P99 of the volume distribution, prior to any truncations related to the minimum commercial field size. The POS of a prospect can be determined from an assessment of a standard set of parameters (e.g. Reservoir, Trap, Hydrocarbon Charge, and Seal). These parameters vary between companies and the assessment of their probabilities can be conducted in different ways. However, the resulting number will determine the predicted success rate of the prospect, i.e. the probability of a hydrocarbon-bearing trap occurring if an infinite number of identical prospects could be drilled. POS assessment will not be discussed further in this paper, but readers are directed towards Otis and Schneidermann (1997) for further details. Buoyancy forces cause hydrocarbons to accumulate in the crest of a valid trap. In order to define the test of a prospect as technically successful, only a relatively small gross rock volume at the crest of the prospect is required to be effectively sealed and filled with hydrocarbons. The potential failure of

additional trapping elements down-dip from the crest of the structure will not reduce the probability of finding hydrocarbons at the prospect location. Rather, they will influence the probability of deeper hydrocarbon-water contacts and hence the presence of larger volumes of hydrocarbons. This is important because the results of the different assessment methods can be significant. METHODOLOGY In this paper, the volume uncertainty distribution is analyzed for a fictitious complex four-way trap. All volume calculations were performed on a personal computer by a Monte Carlo simulator. This method relies on the repeated random sampling of various parameters and computes a predefined quantity of combinations (trials). The key for such calculations is the definition of uncertainty ranges for all the relevant input parameters. The result of the simulation is a range of possible outcomes that can be displayed as a histogram of success cases and a corresponding exceedance probability (cumulative probability) curve. The basis for calculating the volume of recoverable hydrocarbons in a trap is expressed by equation 1: Recoverable volume = [GRV x N/G x Porosity x HC saturation x Recovery Factor] / FVF (1) The gross rock volume (GRV) is calculated by combining (i) area versus depth measurements for the top surface of the reservoir from the crest to the lowest closing contour; (ii) an estimate of reservoir thickness; and (iii) the definition of possible hydrocarbon-water contacts. Other input parameters, such as net-to-gross ratio (N/G), porosity, hydrocarbon saturation (HCsaturation), formation volume factor (FVF) and recovery factor can also be given uncertainty ranges. The determination of valid input ranges and uncertainty distribution functions for these parameters may be anchored in well-defined company policies. Modelling of possible volume outcomes aims to address all the uncertainties related to the amount of hydrocarbons in a potential trap. The interpretation of geological and geophysical data allows appropriate uncertainty ranges to be determined for all relevant factors influencing the volume potentially trapped. The first step of a prospect volume simulation is the translation of observations from the geological model into numerical data. Parameters in the method of estimating the probability of geological success, for instance, are assessed for their probability of adequacy. This can be translated into an either / or distribution which has only two valid values: 0 for

A. Beha et al.

87

failure and 1 for success. Weighting of the two values expresses how likely the parameter is to be present (working) or not. Each parameter of the chance estimation model has a range of possible consequences for the hydrocarbon volume associated with it. Migration is a good example, and is commonly estimated during the assessment of chance of success and volumes. The estimate of efficient migration of hydrocarbons into a potential trap is an either / or decision: either hydrocarbons did migrate in, or they did not. But of equal importance is the assessment of hydrocarbon quantity if migration into the trap occurred. Well-calibrated petroleum systems models can provide answers to this question. In most cases, a range of possible scenarios needs to be included, and inputs for the Monte Carlo model are consequently displayed as a histogram and an exceedance probability curve. Trap complexity can be treated in a similar fashion, and observations from the geological model need to be translated into numerical information. This task comprises an analysis of how likely a trap element is to fail. Thus, specific information for each potential trap element is required to compute the predicted frequency of failure and the consequences of it. For instance, a fault located down dip from the crest requires careful investigation to assess its sealing capability. There will be no consequences for the trap if the fault is sealing. However, failure of the fault results in a leak point at the shallowest intersection between the fault plane and the top reservoir surface. If the fault is leaking, no hydrocarbons will be trapped below this depth. Translation of the observation into numerical modelling language is straightforward. The simulation input will be a weighted-values distribution with two options: either (i) presence of a leak point at the intersection between top reservoir surface and fault plane at the predicted chance of fault failure frequency; or (ii) no leak at a 1 chance of fault failure frequency. Theoretically, each trap element down-dip from the crest of the structure requires an individual input distribution. The method suggested here aims to express this complexity in a single uncertainty distribution for the Monte Carlo simulation. The workflow starts with an introduction to the prospect example and to the geological observations and interpretations. These will help to determine the most appropriate modelling concept. Before the final volume model is designed, the process will describe the method of simplifying the trap complexity to a single continuous uncertainty distribution as input for the final model. The identification of leak points and attempts to predict their frequency using stochastic methods are dealt with in this phase.

The first model in the workflow is very simplistic and was built for the purpose of testing the theoretical predictions with a Monte Carlo simulator. The assumptions were verified by the modelling, and were therefore applied as a single uncertainty distribution input to a more comprehensive volume model. This final volume model contains all the uncertainty parameters which affect the depth of the hydrocarbonwater contact. CASE STUDY: A FAULTED FOUR-WAY CLOSURE Assumptions and definitions A possible workflow for the suggested method is shown for a fictitious structural trap with two faults intersecting the top reservoir grid (Figs 1, 2). The two faults can be sealing or may provide pathways for hydrocarbons to leak out of the system. The crest of the structure is at 2000 m and the lowest closing contour is defined by the 2150 m depth contour. A constant reservoir thickness of 60 m is assumed over the entire prospect area. Uncertainty with regard to the trapped volume is very common when hydrocarbon prospects are evaluated because many parameters are associated with uncertainty ranges. When selecting a depth versus area representation of the prospect GRV (gross rock volume), the parameters with the most influence on hydrocarbon volume are those which affect the depth of the hydrocarbon-water contact. In the faulted four-way closure example, uncertainty related to hydrocarbons available for filling the trap will have to be assumed because no information is available from petroleum systems analysis. For the volume assessment, it is assumed that the charged volumes can range from very small to large. Leak point uncertainty on a leaking fault also needs to be taken into account when assessing the position of the hydrocarbon-water contact, because failure of either of the faults will influence the maximum depth of this contact. If for example the NE fault has no sealing capacity, no hydrocarbons can be trapped below 2050 m. For the sake of simplicity, no uncertainty is assumed for the integrity of the top seal in this example. Therefore the two independent parameters amount of available hydrocarbons and potential fault leak points control the distribution of the hydrocarbon-water contact and thus the degree of trap fill. The two parameters hydrocarbon-water contact and leak point depth are sampled individually during the Monte Carlo simulation, and the shallower of the two depths will determine the actual hydrocarbon-water contact of the individual simulation trial.

88

Volume assessment of complex hydrocarbon traps

P(fault seal) = 0.4

20 50 21 00

P(fault seal) = 0.7

Fig. 1. Map view of a faulted four-way dip closure.Two faults cut into the reservoir below depths of 2050 m and 2100 m, respectively. The fault seal probabilities are estimated to be 40% for the NE fault and 70% for the SW fault. A - B marks the line of the cross-section in Fig. 2.

The range of hydrocarbon volume available to fill the trap will be addressed below. First, however, leak point uncertainty is considered. The fault to the NE (NE fault) cuts the top reservoir at a depth of 2050 m. If the fault has no sealing capacity, hydrocarbons cannot be trapped below this depth. The second fault (SW fault) cuts the top reservoir at a depth of 2100 m. It is assumed that there is no dependency between the two faults, i.e. if one fault fails to act as a seal there is no increased probability of failure of the second fault. Quantifying the probability of seal integrity for both faults is critical for volume assessment of the prospect. The probability that the NE fault seals is estimated to be 40%, and the probability that the SW fault seals is 70%. These estimates are independent of the probability of geological success (POS) estimated for the prospect. The prospect POS is defined as the probability of finding the minimum, P99, success-case volume of hydrocarbons which accumulates at the crest of the structure regardless of the seal integrity of either fault. Hence, the two estimates of sealing integrity for the SE and NE faults will have no impact on the assessment of the prospect POS. But the authors suggest that the reduced probability of encountering hydrocarbons at a greater depth down-dip from the crest is determined by the uncertainty distributions of the potential leak points. This requires the identification and definition of all possible trap configurations and the assessment of the likelihood of their occurrence.

Identification and definition of possible trap configurations Three different leak points can exist in this prospect example (Fig. 2): one at a depth of 2050 m if the NE fault allows hydrocarbons to leak out of the trap; one at 2100 m if the NE fault is sealing and the SW fault leaks; and one at 2150 m if both faults are sealing. For this last option, the leak point is equivalent to the lowest closing contour of the structure, i.e. the spill point depth. In order to create a stochastic volume model with a Monte Carlo simulator, the software needs to know how often it is supposed to generate the different leak points which determine the potential limitation of the hydrocarbon-water contact. This information is provided by the assessment of fault seal probabilities. From the trap configuration, it is clear that the deepest leak point at 2150 m depends on the sealing capacity of both faults. But it may not be valid to assume that this is the least likely result. The workflow may appear ambiguous and counterintuitive but it follows stochastic principles. The map view of the prospect (Fig. 1) shows that the highest point where the NE fault intersects the top of the reservoir section occurs at a depth of 2050 m. If the NE fault is not sealing, no hydrocarbons can be trapped below this depth. Note however that this outcome is an aggregate of two possible scenarios (the first where the NE fault leaks and the SW fault leaks; and the second where the

21 50

A. Beha et al.

89

A
2000 m 2050 m 2100 m 2150 m

max contact 'NE fault' leaking

max contact 'NE fault' sealing AND 'NW fault' leaking max contact 'NE fault' sealing AND 'SW fault' sealing

Fig. 2. Cross-section view of the faulted four-way dip closure showing the maximum possible hydrocarbonwater contact depths for the three possible outcomes.

a)
N
Leaking
m

b)
N
Sealing
m 50 20

c)
N
Sealing
50 20 m 00 21 m

20

50

00

Leaking OR Sealing

Leaking
m 50

Sealing
m 50 21

21

21 21 50 m

Fig. 3. Map view of the three possible outcomes of the prospect example. (a) The small volume outcome (dark grey) is applicable if the NE fault is leaking. The probability of the scenario where the NE fault is leaking and the SW fault seals is different from the probability of the scenario where both faults are leaking at the same time. However, whether the SW fault seals or leaks has no influence on the leak point depth if the NE fault leaks. (b) The medium volume outcome (medium grey) is applicable in the scenario where the NE fault is sealing and the SW fault leaks. (c) The large volume outcome (light grey) is only used for the scenario where both faults are sealing.

21

NE fault leaks and the SW fault seals). The nature of the SW fault has no influence on the leak point depth but both scenarios are valid and should be considered when the probability of possible outcomes is assessed. If the NE fault is sealing and the SW fault leaking, hydrocarbons can potentially be trapped down to a leak point at 2100 m. Finally, if both faults are sealing, hydrocarbons can fill the entire trap to its lowest closing contour at 2150 m. Fig. 3 illustrates possible fault leak points and the lowest closing contour in map view, highlighting the maximum area of each potential outcome. The next section focuses on the critical determination of leak point probabilities. Theoretical prediction of the scenario probabilities Four scenarios with three different leak points have been identified (Table 1). It is important to note however that the probability of the scenarios occurring is not equivalent to the probability of success (POS)

of the prospect. Stochastic combination of the two elements NE fault sealing and SW fault sealing results in four scenarios, as shown in Table 1. Success here indicates that the respective fault acts as a seal; failure indicates that the faults are nonsealing and that hydrocarbons are leaking out of the trap at the shallowest intersection between the fault plane and the top reservoir surface. The two elements, and their probabilities of occurrence, contribute to the probability calculation shown in Table 2. In Scenario 1, both NE and SW faults are sealing; in Scenarios 2 and 3, one fault is sealing and the other is not; in Scenario 4 there is sealing failure of both elements. The scenario probabilities are calculated by multiplying the probabilities of the respective conditions. The probability of Scenario 1 (Table 2) is therefore the product of all success probabilities (0.4*0.7 = 0.28). In Scenario 2, the failure probability of the SW fault is (1- the probability of success), i.e. (1.0 - 0.7

00

90

Volume assessment of complex hydrocarbon traps

Table 1. All possible combinations (scenarios) for the two independent prospect trapping elements are identified in this table. Note that the failures or successes of the elements have no impact on the geological success, but on the uncertainty of the leak points.
Scenario name Scenario 1 Scenario 2 Scenario 3 Scenario 4 NE fault sealing SW fault sealing

x x -

x x x Success - Failure

Table 2.The probability calculation of all valid scenarios combined with the implication for the leak point depth. The basis for the scenario probability calculations is the assessment of adequacy of the two identified trapping elements.The NE fault sealing has a probability of 0.4 and the SW fault sealing is estimated to be valid at a rate of 0.7.
Scenario name Scenario 1 Scenario 2 Scenario 3 Scenario 4 NE fault sealing 0.4 0.4 (1.0 - 0.4) (1.0 - 0.4) SW fault sealing 0.7 (1.0 - 0.7) 0.7 (1.0 - 0.7) P(scenario) 0.28 0.12 0.42 0.18 Leak point depth 2150 m 2100 m 2050 m 2050 m

= 0.3). The probability of Scenario 2 is therefore (0.4*0.3 = 0.12). Similarly, the probability of Scenario 3 is calculated to be 0.42. The probability of Scenario 4 is determined by multiplying the failure probabilities of both faults ((1.0 - 0.4) * (1.0 - 0.7)) = 0.18. Note that the sum of the four scenarios is 1.0. The final column in Table 2 summarises leak point depths for each scenario. As noted above, if the NE fault fails, then the failure or success of the SW fault has no implication for the leak point. Therefore a 2050 m leak point is assigned to both scenarios 3 and 4 in which the NE fault is not sealing. The sum of the probabilities of Scenario 3 and 4 determine the probability of the leak point at 2050 m; the probability of the leak point at 2100 m is given by the probability of Scenario 2 only. Finally, the probability of the deepest leak point at 2150 m is equal to the probability of Scenario 1. The probability of the leak point depths equals the frequency at which the Monte Carlo simulator is required to determine the identified leak points for the calculation of the prospect volume uncertainty. Fig. 4 shows the calculation of weighting probabilities in the form of a scenario tree, a type of display which is commonly employed when various different scenarios are being evaluated. The POS of the prospect is included in the tree; however, the POS only determines the success rate of the prospect and

therefore the rate at which the Monte Carlo simulator calculates successful trials. Once a success has been identified, the POS neither influences the scenario weighting nor the volume calculation. In other words, the weighting of scenarios is normalised to the success rate of the prospect. In Fig. 4, all valid combinations of trapping elements below the crest of the prospect are depicted as individual branches. The weight of each branch is calculated by multiplying the probabilities of the events leading to the particular branch after passing the initial POS criterion. To the right, the applicable leak point depth is attached to each branch. A reference to the scenario determination method described above and shown in Tables 1 and 2 is given by the scenario names. Simulation of the predicted scenario weighting In the next step, a test model is created in the Monte Carlo software for the purpose of replicating and validating the theoretical probability predictions and testing the implications of each scenario on the leak point depth. The calculation of volumes with a Monte Carlo simulator requires the definition of uncertainty ranges for various input parameters. The gross rock volume (GRV) of the example prospect is defined by the depth versus area data in Table 3 and an estimate of reservoir thickness in Table 4. For the sake of simplicity, the

A. Beha et al.

91

Scenario tree
NE fault Success 0.4 Success
POS

Scenario name P(scenario)

Leak point depth 2150 m

0.7 0.3

SW fault Success SW fault Failure SW fault Success SW fault Failure

Scenario 1

0.28

Scenario 2

0.12

2100 m

0.6 NE fault Failure

0.7 0.3

Scenario 3

0.42

2050 m

Prospect
1-POS

Scenario 4

0.18

2050 m

Failure

Sum of P(scenario) =

1.0

Fig. 4. A scenario tree showing all possible outcomes as branches, the weight of each branch and the implied leak point depth for the example prospect. Reference to the scenarios in Tables 1 and 2 is established by the scenario names. Table 3. Depth versus area input data for the description of the prospect shape. For simplicity, no area uncertainty has been modelled.

Depth vs. Area definition


Depth [m]
2000 2025 2050 2075 2100 2125 2150

Distribution type
Constant Constant Constant Constant Constant Constant Constant

Area [km ]
0.00 0.41 1.34 2.82 4.90 6.93 8.90

reservoir thickness has been kept constant as have the areas at the individual depths (this therefore assumes no uncertainty regarding the mapped structure). However, uncertainty distributions have been applied to all other volume parameters such as net-to-gross ratio, porosity, hydrocarbon saturation, formation volume factor, gas-oil-ratio and recovery factor. These parameters are not discussed in detail but are the same for all volume models in this study, and contribute to the calculation of the volume uncertainty distribution of each model. Input ranges and the distribution type of the volume parameters (whether constant, normal or uniform) are defined in Table 4.

In the first simulation, the complexity of the trap is reflected in two input distributions, one for each of the two faults. Potential leak points are identified and the failure frequency is predicted from analogue data. For the simulator to know how many trials it should create for a sealing or leaking fault, the observations need to be translated into probabilistic language. Fig. 5 shows the prospect in a depth versus square-root-of-area display on the right of the diagram, and the translation of fault failure probabilities and their implication on the leak point on the left. One input distribution is assigned to each potentially leaking fault. The assessment of

92

Volume assessment of complex hydrocarbon traps

Table 4. Uncertainty ranges and distribution shapes of all contributing volume parameters for the stochastic calculation of the prospect potential. Note that all distribution shapes are basic functions such as normal, uniform or constant. More complex shapes may be appropriate in many parts of the world where more available data allows for fine tuning of the histograms.

Volume parameter input distributions


Parameter
Reservoir thickness [m] Net/Gross ratio [%] Porosity [%] HC Saturation [%] Formation Factor [bbl/STB] Gas Oil Ratio [scf/STB] Recovery Factor [%]

Distribution type
Constant Normal Normal Normal Normal Uniform Normal

Mean 60.0 65.0 19.0 55.0 1.25 1200 47.5

Minimum 60.0 40.0 13.0 45.0 1.10 1000 35

P90 60.0 52.8 16.1 50.1 1.18 1040 41.4

P50 60.0 65.0 19.0 55.0 1.25 1200 47.5

P10 60.0 77.2 21.9 59.9 1.32 1360 53.6

Maximum 60.0 90.0 25.0 65.0 1.40 1400 0.6

the NE fault translates into a weighted-values distribution where the random number generator will produce 40% trials with a leak point depth of 2150 m (NE fault sealing) and 60% trials with a leak point depth of 2050 m (NE fault leaking). The distribution of possible outcomes for the SW fault looks very similar but has different frequencies and consequences (leak points). Thus, in 70% of the trials the simulator generates a leak point depth of 2150 m (SW fault sealing), and in 30% of the trials the leak point depth is generated at a depth of 2100 m (SW fault leaking). The Monte Carlo simulator samples both distributions simultaneously and independently for each trial. The hydrocarbon-water contact input distribution is kept constant at the deepest closing contour (2150 m) for this first modelling exercise, and no uncertainty on available hydrocarbons is therefore considered. This initial model is designed only to validate the theoretically predicted consequences of the observed trap complexity. Despite the fact that the input of the hydrocarbonwater contact in the model is constant at the lowest closing contour of the prospect, variation of the resulting hydrocarbon-water contact can be expected due to differences in the fault leak points and in their probability of occurrence. Both fault leak distributions will be sampled simultaneously, and the shallowest leak point determines the resulting hydrocarbon-water contact of the individual trial. For example, if the simulator draws a 2150 m deep leak point from the NE fault distribution (replicating a sealing NE fault in this trial), and in the same trial simulates a failure of

the SW fault (leak at 2100 m), the hydrocarbon-water contact of the system cannot exceed a depth of 2100 m even though the input hydrocarbon-water contact distribution has a constant value at 2150 m. The resulting hydrocarbon-water contact of this trial is therefore 2100 m. All resulting hydrocarbon-water contacts of the simulation are displayed in a histogram in the centre of Fig. 5. The resulting distribution of hydrocarbon-water contacts differs significantly from the input hydrocarbon-water contact distribution and allows for the validation of the predicted outcomes through the following procedure. The Monte Carlo simulator was utilised to produce 10,000 successful trials. The starting point of each trial is a filled-to-spill trap with a constant hydrocarbon-water contact at 2150 m, and only the definition of leak point depths and their probability of occurrence will force the contact to deviate from this depth in the simulation. The purpose of the first Monte Carlo simulation is to verify the predicted probabilities of the different scenarios. Four different scenarios were described. Two have the same leak point depth and hence will produce the same hydrocarbon-water contact. Scenarios 3 and 4 will both limit the maximum possible hydrocarbon-water contact to a depth of 2050 m (Fig. 4). The sum of the predicted probabilities of Scenarios 3 and 4 (0.42 + 0.18 = 0.6) is the estimated frequency at which this will occur. Thus, out of the 10,000 simulated samples, ca. 6000 are expected to have a contact depth of 2050 m. Scenario 2 would allow for a deeper leak point at 2100 m and the predicted probability is 0.12 (Fig. 4). Thus, of 10,000

A. Beha et al.

93

Number of trials
2000 4000 6000

2000

2040

2050

Depth [m]

2080 2100 2120

2150

NE fault seal probability and failure consequence

SW fault seal probability and failure consequence

HC Water contact [m] Result

2160

Depth [m]

Square root of area

Fig. 5. The results of the test model for the validation of the predicted likeliness of the leak point depths. On the left, the observations from the fault seal analysis (frequencies and consequences of fault seal failure) have been translated into Monte Carlo simulator input distributions for both faults.The output hydrocarbon-water contacts and their frequencies as proxy for the three possible outcomes are displayed next to the depth versus area representation of the example prospect.

simulations, the model will produce around 1200 samples with a hydrocarbon-water contact at 2100 m. Similarly, the predicted probability of Scenario 1 (0.28) indicates that the simulator will produce ca. 2800 samples with a hydrocarbon-water contact at 2150 m. In these ca. 2800 cases, the output hydrocarbon-water contact will be the same as the input depth. The resulting (output) hydrocarbon-water contact is therefore a suitable criterion for filtering the successful trials of the test model and validating the prediction. Consequently, three sample groups are defined. Sample group 1 filters all 10,000 trials for outcomes with a hydrocarbon-water contact at 2050 m. The software identified 6005 samples in this group and therefore confirms the predicted frequency (6005 10,000 = 0.6005) of this outcome. Some 1192 samples were identified in the second group with the filter set to a hydrocarbon-water contact of 2100 m, likewise confirming the predicted frequency. The remaining 2803 trials were assigned to the third sample group, which filtered the trials for a hydrocarbonwater contact at 2150 m. The absence of duplicates or remaining unclassified samples indicates that no scenario was left out of consideration. In this first Monte Carlo simulation, one input distribution was required to model the frequency and consequences of each leaking fault. The resulting hydrocarbon-water contacts of the model confirmed the stochastically-determined predictions, permitting a simplified input of leak point uncertainty in subsequent volume modelling calculations. The resulting hydrocarbon-water contact distribution in

the centre of Fig. 5 is a proxy for the leak point frequency, and can be defined by a single weightedvalues distribution. Thus, one input distribution can simulate two (or any number of) faults. This is especially valuable because not all simulation software provides the option of defining as many leak point uncertainty distributions as required. In order to simplify the input of potential leak points at various depths and frequencies, the workflow allows for the translation of discrete geological events (inefficient fault seal at specified frequency of occurrence) and their consequences (leak points at specific depths) into a single uncertainty distribution. For the realistic assessment of prospect volume potential, more complex modelling is usually necessary. So far, all the calculations started with a filled to spill condition. However, potential limitations in charge require the additional uncertainty of the hydrocarbon-water contact. Advanced prospect assessment software provides the option of defining the amount of available charge as an input. In this case, the simulator can determine the amount of hydrocarbons for each sample from a given range of possible hydrocarbon charges. This feature is not considered in the present workflow. Instead, the uncertainty of the hydrocarbon-water contact, as a result of potentially limited hydrocarbon charge, is applied to the model. Final volume modelling of the complex four-way trap Uncertainty modelling of the hydrocarbon-water contact may be performed differently in different

94

Volume assessment of complex hydrocarbon traps

Input
Model 1
2020

Result
Model 2 Model 1 Model 2
2000 Minimum HC column length of 20 m

Fig. 6.This diagram displays the input and output uncertainty distributions in relation to the depth versus area representation of the prospect.Two models are constructed to evaluate the significance of the trap complexity on the volume distribution. Model 1 ignores the fault leak possibilities and hence only the uniform hydrocarbon-water contact is applied to this model. As expected, the resulting hydrocarbon-water contact distribution looks very similar to the input. In model 2, the weighted values distribution is applied to the spillpoint parameter in order to include the trap complexity in the volume calculation.This has a considerable impact on the resulting hydrocarbon-water contact.

companies which use the depth versus area methodology to assess the GRV. This element of the analysis is usually critical in any hydrocarbon volume analysis and requires particular attention. Here, a specific methodology for estimating the range, the anchor points or the shape of the hydrocarbon-water contact distribution is not suggested. Rather, a single distribution shape is applied to two volume models as an example. One model is constructed without the elements of complexity of the structure, and the other includes the potential leak points and their predicted frequencies. Finally, differences in the results of the calculations are compared and discussed. The minimum geological success case is defined by a hydrocarbon-water contact of 2020 m. At this depth the example prospect consists only of the fourway element, and it is assumed that a sustainable flow of hydrocarbons to the surface will be guaranteed if hydrocarbons are detected at or below this depth. Hence, the success case definition allows for the assessment of probability of geological success independent from fault seal probabilities. The maximum hydrocarbon-water contact depth is equivalent to the depth of the lowest closing contour at 2150 m. For the sake of simplicity, a uniform distribution between minimum and maximum possible hydrocarbon-water contacts is assumed. Other distribution shapes can be used to describe the uncertainty of the hydrocarbon-water contact without affecting the described method. Volume assessment model 1 ignores the influence of the potentially leaking faults on the hydrocarbon-

HC Water contact / leak point uncertainty [m]

2050

2050

2100

2100

2150

2150

Uniform HC contact distribution

Weighted value leak point distribution

Uniform HC contact distribution

Model 1 (ignoring trap complexity)

Model 2 (incl. trap complexity)

water contact, and therefore only the uniform distribution of the contact is applied. Model 2 incorporates added complexity in terms of the downdip trapping elements. In addition to the uniform uncertainty of the contact, a weighted-values distribution for the leak points is applied. This uncertainty distribution reflects the previously predicted scenarios, their frequency of occurrence, and the implications on the leak points. Fig. 6 illustrates the input distributions of the two volume assessment models in relation to the depth versus area representation of the prospect. The graph also shows the resulting hydrocarbon-water contacts of both models which can be extracted after the Monte Carlo simulator finishes the calculation. The distribution of the resulting hydrocarbon-water contacts of model 1 is, as expected, very similar to the input contact distribution. However, the fault leak probabilities have a significant impact on the resulting contact depths in model 2. This model recognises the trap complexity, and calculates a contact of 2050 m in 4593 of the 10,000 samples; these are visible as a prominent peak in the result histogram of the hydrocarbon-water contact of model 2 (Fig. 6). This peak can be explained in terms of the stochastic interaction of the two independent input uncertainties (hydrocarbon-water contact, and leak point depth). The independent behaviour of these uncertainties will lead to many trials where the simulator picks a contact from the hydrocarbon-water contact input distribution at a depth below 2050 m. This depth (2050 m) is the P77 of the input hydrocarbon-water contact and therefore

Depth [m]
Square root of area

A. Beha et al.

95

(ignoring trap complexity) 1.0 0.9 0.8


Cumulative frequency Cumulative frequency

Model 1

(including trap complexity) 1.0 0.9 0.8 0.7 0.6 0.5 0.4 0.3 0.2 0.1

Model 2

0.7 0.6 0.5 0.4 0.3 0.2 0.1 0 0 20 40 60 80 100 120

0 0

20

40

60

80

100

120

Accumulation size Total Resources [1e6 STB OE]

Accumulation size Total Resources [1e6 STB OE]

Fig.7. Resource histogram and cumulative probability curve for both volume models. Model 1 ignores trap complexity while model 2 includes the fault seal assessment of both faults. The effect of the potential fault seal failures on the hydrocarbon water contact determines the significant difference of the results.

Table 5. Monte Carlo simulation results of the two volume models of the example prospect. Mean volume and most commonly communicated percentiles of the recoverable hydrocarbon volume distribution are displayed in million barrels of oil equivalent [MMboe].

Recoverable hydrocarbon volume [MMboe]


Volume model
Model 1 ignoring trap complexity Model 2 including trap complexity

Mean 27.5 12.7

P99 0.7 0.7

P90 1.8 1.8

P50 19.7 5.6

P10 65.7 36.9

P1 98.4 82.3

a deeper contact is selected in 76% of the trials. At the same time, the simulator draws a leak point at 2050 m (60% of the leak point input distribution). A potentially large column will then be truncated, and the resulting hydrocarbon-water contact of the particular trial is 2050 m. Thus the shallower depth of the two independently sampled input distributions determines the hydrocarbon-water contact of the trial. The difference between models 1 and 2 with respect to the amount of recoverable hydrocarbons is shown in Table 5, which shows the mean volume and the five most commonly used percentiles of the cumulative probability curve. Numbers in the table are given in million barrels of oil equivalent [MMboe] and the conversion factor to million standard cubic metres [MMSm3] is 0.159. The resource histogram and cumulative probability curve of both models are shown in Fig. 7. The significant difference of the resource distribution shapes is solely caused by the different effective hydrocarbon water contact distributions.

If trap complexity in model 1 is ignored (i.e. ignoring whether the NE fault and SW fault may or may not seal), this is equal to assuming permanent access to the full GRV of the structure. Table 5 shows that the result of this volume uncertainty calculation gives a mean volume of 27.5 MMboe compared to 12.7 MMboe when trap complexity is incorporated, but a key reminder is that the probability of geological success (i.e. flowing hydrocarbons at a minimal but sustained level) is the same for both. DISCUSSION A consistent definition of the geological success of an exploration project is vital for the volume assessment of the trap, especially when prospects are aggregated within a portfolio and forecasts or commitments are made. Whether the prospect generator is requested to assess the probability of a mean volume case or the minimum economic volume of hydrocarbons or the minimum success-case

96

Volume assessment of complex hydrocarbon traps

P(HC water contact at or deeper than reservoir entry depth)


2000

P(HC water contact at or deeper than reservoir entry depth)


0.23 0.0 0.5 1.0

0.57

0.0

1.0

2050

Depth [m]

2075 2100

2150

Square root of area

Projected well trajectory

Model 1 (ignoring trap complexity)

Model 2 (incl. trap complexity)

Fig. 8.The probability of well success is different than the probability of technical success if the well is not designed to test the prospect in the most crestal position.The graph shows a well that is planned to enter the reservoir at a depth of 2075 m. Contact depths of 2075 m and deeper result in a successful well. The hydrocarbon-water contact probability of exceedance curve of model 1 shows that this is the case in some 57% of the technically successful trials. Hence, the probability of well success is equal to the probability of geological success (POS) multiplied by 0.57. However, in model 2 which includes the potentially leaking faults, the probability of a successful well is significantly reduced to 23% of the geological success cases due to the complexity of the trap.

volume has a significant impact on the total assessment. The definition of the probability of geological success used here permits the use of a relatively straightforward methodology for estimating the potential hydrocarbon volume in a trap. Whether or not trap elements fail in the deeper part of the structure will not influence the probability of finding the minimum success-case volume since this is usually located in the shallowest part of the structure. The methodology presented will determine the volume distribution of complex traps, but will also influence subsequent analyses such as calculating the probability of success of a particular well location if the well is not located in the most crestal position. For example, a location may be chosen to test the minimum economic hydrocarbon volume of the prospect. Once a Monte Carlo simulation has been completed to assess the prospect volume uncertainty, the shallowest reservoir entry point to test a minimum economic volume of hydrocarbons can then be backcalculated. The probability of success of the well location then becomes the POS (the probability of finding the minimum accumulation) multiplied by the percentile associated with the shallowest reservoir entry point to test a minimum economic volume on the cumulative probability curve of the hydrocarbonwater contact.

Fig. 8 illustrates the concept of determining the probability of well success multiplier for a well that is planned to enter the reservoir at a depth of 2075m. Two multipliers are determined: one is based on model 1 which ignores the trap complexity, and the second is based on model 2 which includes the potentially failing trap elements. For model 1, the probability that the hydrocarbon water contact is at 2075m or deeper amounts to some 57%. This implies that in 43% of the geologically successful cases, the contact will be between 2020 m and 2075 m. These cases cannot be proven by the well, and hence the probability of well success is equal to the probability of technical success (POS) multiplied by 0.57, i.e. the fraction of the contact distribution that is at or below 2075 m. The same technique is applied to model 2; in this case, only 23% of the technically successful cases can be tested by a well that is designed to enter the reservoir at 2075 m. In other words, the multiplier determined from model 2 (0.23) is significantly smaller than the multiplier of model 1 (0.57). The likelihood of deeper hydrocarbon-water contacts is reduced in cases where the trap is dependent on additional trapping elements down-flank. Therefore, the probability of well success can be considerably less if the well is designed to test a section that is potentially affected by the failure of a deeper trap element.

A. Beha et al.

97

The probability of success for any well location can be calculated in the same way. CONCLUSIONS The assessment of complex hydrocarbon traps can be challenging and requires more attention to detail compared to simple four-way dip closures. All elements which can potentially reduce the probability of deeper hydrocarbon-water contacts need to be assessed and their implications for the volume model evaluated. The stochastic combination of all elements allows for an adequate volume assessment of the prospect. It is evident from the example presented that the assessment of complex traps may be a counterintuitive process. Without determining all possible scenarios, it is not intuitively obvious that a deep leak point can be statistically more likely than a leak point higher up the structure, although the deeper leak point requires more elements to seal simultaneously. ACKNOWLEDGEMENTS The authors wish to express their gratitude to Gary Citron, David Cook and James MacKay for their reviews of an earlier draft of this article. Their comments were very much appreciated and helped to improve the manuscript. They would also like to thank the referees Ren O. Thomsen and Glenn McMaster for their very constructive suggestions during journal review. REFERENCES
CITRON, G.P., MACKAY, J.A. and ROSE, P.R., 2006.Appropriate Creativity and Measurement in the Deliberate Search for stratigraphic traps. In:Allen, M.R., Goffey, G.P.,Morgan, R.K. and Walker, I.M., (Eds), The Deliberate Search for the Stratigraphic Trap Where Are We Now? Geol. Soc. Lond. Spec. Publ., 254, 27-41. OTIS, R.M. and SCHNEIDERMANN, N., 1997. A process for evaluating exploration prospects. AAPG Bull., 81(7), 10871109. YOUNG, R., McINTYRE, S., MCLANE, M.A., COOK, D.M., MACKAY, J.A. and GOUVEIA, J. 2005. Complex Traps: A method for calculating the chance-weighted value outcomes for a prospect with multiple trapping styles. (Abstr) AAPG Bull.

You might also like