1. Trang chủ
  2. » Thể loại khác

(SPE78168MS) New Model for Predicting the Rate of Sand Production

9 588 1

Đang tải... (xem toàn văn)

THÔNG TIN TÀI LIỆU

Thông tin cơ bản

Định dạng
Số trang 9
Dung lượng 387,03 KB
File đính kèm SPE-78168-MS.rar (385 KB)

Các công cụ chuyển đổi và chỉnh sửa cho tài liệu này

Nội dung

Abstract A number of robust predictive methods for establishing sanding thresholds have been developed over the past decade. Having identified when the onset of sanding occurs, recent research efforts have focused on determining the rate at which sand will be produced once these thresholds are exceeded. In this paper a new analytic model for predicting the rate of continuous (steadystate) sand production is described. This sanding rate model is consistent with the threshold prediction model, and utilizes as its basis the nondimensionalized concepts of loading factor (nearwellbore formation permeability, viscosity, density and flow velocity). Interpreted this way, the results of laboratory sand production experiments are used to derive an empirical relationship between loading factor, Reynold’s number and the rate of sand production. A second empirical sand production ‘boost factor’ incorporates the effects of water production. The derived model is compared with field data from a total of six wells from two fields, for a wide range of flowing conditions. The predictions are a good match to the field data, typically overestimating the fieldmeasured data by a factor of less than four. However, as the model is for continuous sanding only, this degree of overprediction is considered acceptable for field application, as it provides some compensation for shortlived transient sand production at rates higher than steadystate values.

Trang 1

Copyright 2002, Society of Petroleum Engineers Inc

This paper was prepared for presentation at the SPE/ISRM Rock Mechanics Conference held

in Irving, Texas, 20-23 October 2002

This paper was selected for presentation by an SPE/ISRM 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 or

International Society of Rock Mechanics 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, International Society of Rock Mechanics, its officers, or members Papers

presented at SPE/ISRM 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, TX 75083-3836, U.S.A., fax 01-972-952-9435

Abstract

A number of robust predictive methods for establishing

sanding thresholds have been developed over the past decade

Having identified when the onset of sanding occurs, recent

research efforts have focused on determining the rate at which

sand will be produced once these thresholds are exceeded In

this paper a new analytic model for predicting the rate of

continuous (steady-state) sand production is described This

sanding rate model is consistent with the threshold prediction

model, and utilizes as its basis the non-dimensionalized

concepts of loading factor (near-wellbore formation stress

normalized by strength) and Reynold’s number (a function of

permeability, viscosity, density and flow velocity)

Interpreted this way, the results of laboratory sand production

experiments are used to derive an empirical relationship

between loading factor, Reynold’s number and the rate of sand

production A second empirical sand production ‘boost factor’

incorporates the effects of water production The derived

model is compared with field data from a total of six wells

from two fields, for a wide range of flowing conditions The

predictions are a good match to the field data, typically

overestimating the field-measured data by a factor of less than

four However, as the model is for continuous sanding only,

this degree of overprediction is considered acceptable for field

application, as it provides some compensation for short-lived

transient sand production at rates higher than steady-state

values

Introduction

Over the past decade considerable research efforts have been

expended in developing robust methods for predicting the

onset of sand production as a function of rock strength,

drawdown and reservoir pressure The most notable contributions to this area of work have been by Shell; References 1 and 2 provide a good overview of this decade of effort In recent years, attention has now focused on establishing methodologies for predicting the rate at which sand is produced once the sanding threshold is exceeded The principal motivation for this work is to determine whether sand production can be managed at surface, or if downhole sand control is needed There are pros and cons to both approaches – both management and exclusion

In the sand management scenario, the biggest risk and challenge is being able to reliably estimate the amount and concentration of the produced sand This is important for sizing facilities sand handling capabilities, as well as ensuring that erosion limits for chokes and surface pipework are not exceeded From a HSE perspective, this is especially critical

in high rate gas wells, as well as in high rate oil wells, particularly where gas-oil ratios are high From an operating cost perspective, the consequences of severe sand production and choke erosion could be very costly in subsea wells, especially in deepwater On the positive side, the cased and perforated completion option usually employed with sand management does permit avoidance of producing from notably sanding prone intervals through selective or optimized perforating Cased and perforated completions also maintain access to the producing interval to shut-off water or to recomplete in other secondary producing horizons This has allowed significant increases in reserves recovery in a number

of fields worldwide

The alternative to sand management is sand exclusion When properly implemented, downhole sand control will exclude the bulk of the formation sand from being produced (It is noted, however, that some fines, smaller than the filter media apertures, may still be produced to surface even for successfully installed sand control; this is particularly true of transient fine sand production) The downside of this option is typically a significant increase in up-front well completion cost, and oftentimes, a lower well productivity than a comparable cased and perforated completion Occasionally, sand control completions may also ‘fail’ during the well life, either mechanically, so permitting the influx of formation sand, or suffer degrading inflow performance due to plugging The ability to easily intervene in sand control completions to shut-off water is often difficult, as the preferred completion

SPE/ISRM 78168

New Model for Predicting the Rate of Sand Production

S.M Willson (BP America Inc.), Z.A Moschovidis, J.R Cameron (PCM Technical Inc.) & I.D Palmer (BP America Inc.)

Trang 2

option – typically open-hole gravel packs, screen completions

and frac-packs – may allow the water to by-pass the treated

interval

Therefore, there is often a significant cost benefit – both

for capital and operating expenditure – if sand management

can be successfully implemented However, to reliably do this

in a new project development it is necessary to be able to

produce a credible prediction of the rate at which the sand

might be produced The derivation and validation of such a

model is described in the sections following

Description of Sand Rate Model

1 Impact of Stress Concentration Effects

In the development of sand rate prediction models, it is

important that the basic framework is consistent with the

sanding threshold models applied in other applications This

ensures continuity in approach between predicting the onset of

sanding and its severity once it occurs

The following formulation has been used for the onset of

sanding calculations; i.e the calculation of the critical

bottom-hole flowing pressure resulting in sand production, CBHFP It

is based on a simple apparent strength criterion, together with

assumed linear-elastic behavior, applied to a formation

element next to a circular hole The hole could be the

wellbore (for open hole completion) or a perforation (for cased

hole completion) The orientation of the wellbore or the

perforation is reflected in the calculation of the principal

stresses perpendicular to the hole in terms of suitably

transformed in situ principal stresses

Given the far field total stresses on a plane perpendicular

to the axis of a hole, S1 and S2, (S1 > S2), the tangential stresses

on the surface of the hole (see Figure 1) are given by:

0 1

2

1 3 S S p ( 1 A ) Ap

and similarly

0 2

1

2 3 S S p ( 1 A ) Ap

where it is assumed that the wellbore pressure is

communicated in the formation (i.e during production of a

permeable interval); pw is the wellbore pressure, p0 is the

reservoir pressure far field and A is a poro-elastic constant

given by:

) 1

(

) 2

1

(

ν

α

ν

=

and α is Biot’s constant given by:

b

r C

C /

1 −

=

where ν is the Poisson’s ratio and Cr and Cb are the grain and

bulk rock compressibility, respectively

To avoid sand production the largest effective tangential stress (St2 - pw) should be smaller than the effective strength of the formation, U, next to the hole, i.e.:

U p

Figure 1: Tangential stresses at the wall of a hole

Solving the inequality for pw and introducing the notation CBHFP (Critical Bottom Hole Flowing Pressure) it follows that:

) 2 ( )

2 (

3

0 2

1

A

A p A

U S S CBHFP

pw

=

The critical drawdown pressure (CDP) is defined as the drawdown from the reservoir pressure to cause failure (and sand production) of the reservoir Using the definition, the

Introducing this in (6) we find the functional relation between the reservoir pressure, p0, and CDP

)]

2 ( 3

[ 2

1

2 1

)]

3 ( 2 [ 2

1

2 1

p A

In particular the CRP (critical reservoir pressure), defined as the reservoir pressure that would not tolerate any drawdown,

is given by (7a) for CDP=0: CRP = ( 3 S1 − S2− U ) / 2 Note that S1 and S2 depend linearly on the reservoir pressure po Therefore, (6) should not be used with constant

S1 and S2 values for cases where reservoir depletion effects are considered

Relation of Effective Formation Strength, U, to Measured Strength In the sanding models employed by BP, the

collapse pressure of a so-called thick-walled cylinder test

S 1

S 2

S t1

S t2

S 1

S 2

S t1

St2

Trang 3

(TWC) is used as the fundamental strength measure for

unsupported boreholes and perforations This is consistent

with the original methodology described by Veeken et al1

The standard dimensions for the TWC samples used by BP are

1½” OD × ½”ID × 3” long These are slightly larger than the

sample dimensions adopted by Shell1

A relationship between the effective in-situ strength of the

formation, U, and the TWC strength is necessary since the

TWC test does not directly replicate perforation collapse

pressures The standard TWC test is performed on specimens

where the ratio OD/ID = 3 At in situ conditions, the effective

strength would be represented by a TWC strength where

OD/ID tends to infinity There is an ID scaling issue too, as

perforation tunnels may easily exceed 0.5” diameter when

deep penetrating perforating charges are used in low-strength

sandstones Scaling relationships to account for these effects

have been published by van den Hoek at al2 They found that

for Castlegate sandstone, with an OD/ID ratio of infinity, the

maximum size effect factor varies between 3.0 and 3.8,

depending upon the amount of post-failure softening

Comparable internal research by BP investigated the TWC

collapse resistance of a number of sandstones having a variety

of OD/ID ratios, and different values of ID (see Figure 2)

Figure 2 Increasing TWC Collapse Pressure in Castlegate

Sandstone with Different OD/ID Ratios

The trend of results for varying OD/ID ratios is presented in

Figure 3, which compares the relative strengths of experiments

run in large specimens (OD/ID = 14) with those at smaller and

standard OD/ID ratios Overall, these laboratory results are in

good agreement with the analytical results of van den Hoek at

al2 The testing showed that relative to the collapse pressure

of the standard specimen, TWCsp, the equivalent formation

strength, U, of a specimen with an OD/ID ratio of infinity

would be equivalent to:

U = 2 × 1.55 × TWCsp = 3.10 × TWCsp (8)

Note that in the above, the factor of 2 is introduced to compute

the effective (or ‘boosted’) formation strength by virtue of the

linear-elastic model assumptions inherent in the derivation of

critical bottom hole flowing pressures

Figure 3 Scaling Factors for TWC Collapse Pressures, Normalizing by Collapse at Large OD/ID Ratios

Concept of a Loading Factor Having defined the

appropriate expressions for predicting sanding thresholds (i.e the onset of sanding), it is convenient to non-dimensionalize the stress state acting on a perforation tunnel or borehole by considering the concept of a “Loading Factor”, LF, where, to

be consistent with (5), LF is defined as:

U p S

where St2 is the maximum tangential total stress acting on the formation or perforation We note that for LF<1 the formation

is not failed, while for LF>1 the formation is failed and sand is produced To be consistent with the field, i.e with (6), substituting (2) and (8) into (9) it can be shown that LF must also be equal to:

TWC

p p A p S S

* 10 3

) (

2

2 Impact of Fluid Flow Effects

Intuitively, once perforation tunnels have been stressed sufficiently that a mechanically-weakened zone and disaggregated sand grains exist around the perforation tunnels,

it is reasonable to assume that these could be produced to the surface with sufficient production flow This is in contrast to the analysis of sanding thresholds, where fluid flow rate has only a negligible effect in rocks with moderate cementation The analytical approach adopted to assess fluid flow effects in the sanding model draws on extensive work undertaken to assess required underbalance surge flow rates for perforation clean-upe.g 3,4 In these previous studies, the removal of shock-damaged and mechanically-weakened debris due to non-Darcy flow or turbulence in the region adjacent to the perforation cavity was correlated with the non-dimensional Reynold’s number, defined by:

3000

3500

4000

4500

5000

5500

6000

6500

7000

7500

0.0 2.0 4.0 6.0 8.0 10.0 12.0 14.0

OD/ID TWC Sample Ratio

ID = 0.30"

ID = 0.50"

ID = 0.63"

ID = 1.00"

0.8 0.9 1 1.1 1.2 1.3 1.4 1.5 1.6 1.7

TWC Sample OD/ID Ratio

Saltwash South

Trang 4

βρV 10

31735

.

1

e

k

×

Here, k is the permeability (in mD); β, the non-Darcy flow

coefficient (having dimensions of ft-1); V is the velocity of the

fluid crossing the lateral surface of the perforation or well (in

inches/second); ρ is the density (in lb/ft3); and µ is the

viscosity (in cP) Various correlations have been proposed in

porosity, φ, and/or saturation, Sw In this work, as well as in

Tariq3, a correlation of the form β = constant/ke is used The

range of the exponent, e, in the literature varies from 1.03 to

1.65 The relation used by Hovem et al4 is used specifically

here:

2 1

10/

10

65

.

=

Therefore, by non-dimensionalizing the fluid flow

contribution, variations in formation permeability, fluid flow

rate, viscosity, etc., can be easily captured in the analysis

necessary for effective perforation clean-up during

underbalanced flow Subsequent discussions will show that

similar high values of Reynold’s number are needed for

massive sand production rates At Reynold’s numbers less

than 0.1, the sand production rate is dominated by the loading

factor

Sources of Experimental Data

An important feature of the prediction model described here is

that the sand rate magnitude is established from an empirical

interpretation of laboratory sand production tests, rather than

relying upon empirically derived relationships from field

sanding events; e.g as done in Reference 5 This permits

laboratory sanding experiments to be performed on reservoir

core to derive field-specific sanding relationships; however,

the results presented in this paper have been derived from

generic relationships based on earlier sand production

experiments

Extensive laboratory testing programmes have been

undertaken in the past decade, primarily to establish sand

production thresholds; e.g References 6 through 9 In these

programmes, the effects of fluid flow rate, seepage forces and

stress levels have been investigated separately, thus providing

ideal input data so that these individual contributions can be

properly quantified The TerraTek CEA#11 testing, in

particular, investigated the sanding response of formations

having unconfined compressive strengths in the range 500 psi

to 2000 psi, so making it directly applicable to common field

situations where sand production is a concern

The results of a typical “stress-to-failure” experiment from

the CEA#11 testing programme is shown in Figure 4 In the

experiment shown, the flow-rate was kept constant (typically

at 50 cc/sec) and the confining pressure increased step-wise

(with associated ‘hold’ periods) and the sand production rate

monitored until approximately stable and constant sand production rates were observed Short-lived transient sand production is seen each time the stress level is increased, though this decays to a lower continuous sanding level after a period of time The model data used in this study are only those constant sanding rate values observed at the end of each successive period where the confining pressure is held constant In the example shown, the constant sanding rate is seen to increase gradually as the applied confining stress is applied until a catastrophic sanding event is seen at 7000 psi confining stress

Figure 4 Typical Result of Stress-to-Failure Sand Production Experiment

Using the normalized parameters of Loading Factor and Reynold’s Number, it has been possible to consistently combine results from different sandstones (e.g accounting for strength and permeability variability), as well as factoring in the effects of fluid viscosity and flow rate It is specifically noted here that the model scope is limited to weakly compressible fluids (oils and water) and not to highly compressible gas flows However, the authors see no reason why the methodology cannot be extended further if sufficient calibrating sand production tests were performed

For the data available, from the CEA#11 sand production JIP project6,7,8 and from Papamichos9, the empirically-derived surface shown in Figure 5 was fitted to the data This relates the constant sand production rate (in pounds per thousand barrels, pptb) to the Loading Factor and Reynolds Number for those tests flowing oil only

To address the impact of water-cut on sand production rate, other sand production experiments that were conducted using two-phase flow were analyzed In this step, the continuous sand production rate at a specified water-cut and stress level was compared with that of a test flowing dry oil only at similar conditions This permitted the derivation of a

“water-cut boost factor” to raise the level of sand production from that evaluated from the function pertaining to no water production The form of this correlation is shown in Figure 6

It is recognized that the method used to increase sand

0.1 1 10 100 1000

0 200 400 600 800 1000 1200 Cumulative Flow Through Sample (litres)

0 1000 2000 3000 4000 5000 6000 7000 8000

Sand Production Rate Confining Pressure

Drilled Hole 'Stress to Failure' Test

Trang 5

production after water-cut is a crude approximation of an

effect that is a function of many different physical processes –

capillary pressure reduction, increased seepage pressures due

to relative permeability effects, fluid viscosity effects, as well

as possible mechanical strength reduction post

water-breakthrough However, within the confines of this simple

analytical model for predicting sand production rate, the

water-cut boost factor shown below is seen as an expedient

compromise

Figure 5 Fitted Surface to Experimental Sand Production Rate in

Terms of Loading Factor & Reynold’s Number (Dry Oil Flow Only)

Figure 6 Experimental Data and Analytical Function to Account

For Two-Phase Flow Sand Production Rate Increases

We have field data from sand producing fields, moreover,

which suggests that this boost factor correlation to account for

water-cut effects is not unreasonable Figure 7 shows four

such example wells where the sand production rate is

compared with the measured water-cut These plots are not

quite like-with-like comparisons, as in Figure 6, as drawdowns

and depletion values are not entirely constant for the data

combinations compared Nevertheless, the overall trend of the

observed sand production variation with water-cut is not

inconsistent with that derived from the experimental data

Final Form of Sand Rate Model

From the preceding discussions the final form of the sand rate

model is thus defined It comprises three basic components:

Definition of the Load Factor, LF = f(in-situ stresses, well

trajectory, reservoir pressure, drawdown & depletion, TWC strength)

Definition of Reynold’s Number, R e = f(permeability, flow rate per perforation, viscosity, density, perforation number and size)

Definition of Sand Production Rate, SPR = f(LF, Re, water-cut)

Figure 7 Field Data From Four Wells Showing Approximately 10-Fold Increases in Sanding Rate After Water-Breakthrough

The analytical expressions for the load factor and Reynold’s number can be applied on a foot-by-foot basis using petrophysical wireline data Contributing to the Loading Factor, the rock strength profile is typically derived first using standard predictions of unconfined compressive strength, UCS, and then using a laboratory-derived relationship between measured UCS and TWC strengths Profiles of in-situ horizontal stresses can be derived by knowing the overburden pressure, pore pressure, formation Poisson’s ratio and any contribution of tectonic stresses (e.g assessed from leak-off tests, minifrac tests, step-rate tests, or from water injection data)

Contributing to the evaluation of the Reynold’s number, formation permeability can be assessed using standard correlations between porosity and permeability evaluated at appropriate net mean stress conditions from routine core analysis Formation fluid properties are typically known From the above, the sanding rate can be evaluated for any combination of drawdown and depletion As the individual contribution per foot (or half-foot, depending on logging data frequency) is assessed, then it is easy to assess the consequences of selective perforating should the highest permeability formations not be perforated

0 5

1 0

1 5

Water-Cut ( %)

0 5

1 0

1 5

2 0

Water Cut (%)

0 5

1 0

1 5

2 0

Water Cut (%)

0 5

1 0

1 5

2 0

Water Cut (%)

0 0.05 0.1 0.15 0.2 0.25

Reynolds No.

0 0.5 1 1.5 2 2.5

3

3.5

4

LF

0.1

1

10

100

1000

Water Cut %

0.1

1

10

100

1000

Water Cut %

Trang 6

Field Case History Analyses of Sand Production

Rate Prediction

The methodology for predicting sanding rates is now applied

to two fields in the section following The first field example

(two wells) is producing dry oil at modest to high rates

(between 2,000 and 20,000 bopd); the second field example

(four wells) has historically produced at high rates – initially

at up to 38,000 bopd, though with time total production rates

have declined to approximately one-half this initial amount as

water production has increased to over 90% These two field

examples (6 wells in total) therefore provide a quite thorough

testing of the model Table 1 provides a summary of

formation properties over the perforated intervals analyzed

However, as the model takes into account the

half-foot-by-half-foot variability of in-situ properties, the minimum and

average values are not necessarily representative of those

formation characteristics dictating the overall rate of sand

production

For the fields analyzed, strength profiles were established

by conducting both unconfined and thick-walled cylinder

measurements over the whole range of formation quality, so

characterizing the extent of strength variability In-situ

stresses were determined from using procedures described

previously

The results of the sand rate prediction analyses are shown

in Figures 8 and 9 for Field A wells, and in Figures 10 to 13

for Field B wells Figures 8 and 9 simply compare the

observed and predicted sand production rates for dry oil

production only Figures 10 to 13 also show the measured

water-cut, which in some cases varies significantly over the

period analyzed

The data presented in Figures 8 through 13 were collected

when the specific wells were flowed through a test separator

Thus, good measurements were made of oil, water and sand

rates, as well as surface flowing pressures from which

bottomhole flowing pressures were estimated from nodal

analysis Therefore, the data used to validate the sanding

models are the best typically available offshore

Overall, the prediction model derived from the laboratory

sanding experiments is able to reproduce the measured

sanding response, though typically over-predicting that

measured by a factor of two to four For those wells

producing dry oil (Figures 8 and 9) very good agreement is

reached The predicted rates of 3 pptb to 4 pptb typically

provide an upper limit to that measured For those wells

producing water (Figures 10 to 13) the match is still quite

acceptable, though there is more scatter in both the measured

sanding data and the predictions The principal cause of this is

the representation of the produced water in the model If a

well is producing with 50% water-cut, the model assumes a

50/50 split in oil and water over the entire perforated interval

This maximizes the applied water-production sanding “boost

factor” shown in Figure 6 The reality could be quite

different, with perhaps the top half of the perforated interval

producing dry oil, whereas water coning has caused the lower

half to water-out The distribution of the water influx in the

wells analyzed is not known, however, and the analysis approach adopted is known to be conservative

Figure 14 shows the predicted sand influx distribution for well B / 1 for the following specified producing conditions: 29,690 bpd gross liquid production; 77% water-cut; 592 psi drawdown and 265 psi depletion The overall predicted sand production for the entire perforated interval is 119 lbs/day, equivalent to a sanding rate of 4 pptb Also shown in this figure is the formation permeability distribution This correlates well with porosity and inversely with formation strength (high permeability, low strength) The figure shows a high permeability streak from 9927 ft to 9930 ft TVD.SS is predicted to produce 12 lbs of sand per day, approximately 10% of the overall predicted total Therefore, if sand production rate and erosional constraints were of concern in this well, then it may be prudent to omit perforating this 3 feet long interval It would be possible to make such an assessment in the time available between logging and perforating a well, if the required correlations between strength, porosity and permeability were established beforehand

Conclusions

1 The non-dimensionalized approach described to combine and interpret laboratory sand production experimental data can be used as a basis for deriving credible sand production rate prediction methods

2 The “Loading Factor” concept allows the derived sanding rate model to be consistent with existing models for predicting the on-set of sand production

3 The “Reynold’s Number” concept to include fluid flow effects is well documented from perforation clean-up research, and the empirical “sand production boost factor”

to account for the effects of water production is corroborated by field evidence

4 Applied to field examples from sand producing wells, the derived analytical model is seen to perform well when compared with the measured data The over-prediction of the continuous sanding rate, by a factor of typically two to four, is seen as acceptable when using these data for sizing facilities sand handling capabilities

Acknowledgements

We thank BP America Inc for permission to publish this paper The efforts of Dr Joe Hagan in overseeing the work associated with the TWC scaling relationships are specifically acknowledged

References

1 Veeken, C.A.M et al: “Sand Production Prediction Review:

Developing an Integrated Approach”, paper SPE 22792, presented at the 1991 SPE Annual Technical Conference and Exhibition, Dallas, 6-9 October

2 van den Hoek et al: “A New Concept of Sand Production

Prediction: Theory and Laboratory Experiments”, SPE Drilling

& Completion, Vol 15, No 4, December 2000, pp 261-273

Trang 7

3 Tariq,S.M “New, Generalized Criteria for Determining the

Level of Underbalance for Obtaining Clean Perforations”, paper

SPE 20636, presented at the 1990 SPE Annual Technical

Conference and Exhibition, New Orleans, 23-26 September

4 Hovem,K., Jøransen,H., Espedal,A., & Willson,S.M “An

Investigation of Critical Parameters for Optimum Perforation

Clean-Up”, paper SPE 30084, presented at the European

Formation Damage Conference, The Hague, The Netherlands,

15-16 May 1995

5 Papamichos,E & Malmanger,E.M “A Sand Erosion Model for

Volumetric Sand Predictions in a North Sea Reservoir”, paper

SPE 54007, presented at the 1999 SPE Latin American and

Caribbean Petroleum Engineering Conference, Caracas,

Venezuela, 21-23 April

6 Halleck,P.M., “An Experimental Investigation of Sand Production: CEA Project #11 Final Report”, prepared by TerraTek, Inc., June 1991

7 Willson,S.M “CEA 11, Phase II, An Experimental Investigation of Phenomena Affecting Sand Production in Low-Strength Sandstones,” Final Report, Vol 1, Summary of Results, prepared by TerraTek, Inc., August 1993

8 Willson,S.M “CEA 11, Phase III, An Experimental Investigation of Phenomena Affecting Sand Production in Low-Strength Sandstones,” Final Report, Vol 1, Summary of Results, prepared by TerraTek, Inc., April 1995

9 Papamichos,E., “A Volumetric Sand Production Experiment,”

2000 Balkema, Rotterdam, ISBN 90 5809 155 4

Table 1 Summary of In-Situ Properties For Wells Analyzed

(TVD.SS) &

Well Deviation

(psi)

Initial Reservoir Pressure (psi)

Initial Horizontal Stress (psi)

43º deviation

Average Minimum Maximum

112

10

353

2481

A / 2

8949 ft to

9440 ft 39º deviation

Average Minimum Maximum

98

10

640

2955

160

4972

1647

9325 4658 7320

0º deviation

Average Minimum Maximum

219

20

1048

3281

B / 2

9253 ft to

9291 ft 58º deviation

Average Minimum Maximum

675

35

2044

1592

967

5146

4025

8230 4055 6174

B / 3

9686 ft to

9725 ft 13º deviation

Average Minimum Maximum

450

40

1133

2544

1923

6512

5675

8658 4180 7110

B / 4

9426 ft to

9527 ft 35º deviation

Average Minimum Maximum

601

21

3761

2456

493

6366

2874

8456 4121 6975

Figure 8 Predicted vs Measured Sand Production Rate – Field A, Well 1

0.0 1.0 2.0 3.0 4.0 5.0 6.0

Q (bpd)

Avg Measured Sand Rate (pptb) Predicted Rate (pptb)

Trang 8

Figure 9 Predicted vs Measured Sand Production Rate – Field A, Well 2

Figure 10 Predicted vs Measured Sand Production Rate – Field B, Well 1

Figure 11 Predicted vs Measured Sand Production Rate – Field B, Well 2

0.0 2.0 4.0 6.0 8.0 10.0

12.0

14.0

Oil Rate(bpd)

Avg Measured Sand Rate (pptb) Predicted Rate (pptb)

0 3 6 9 12 15 18 21 24 27 30

29000 30000 31000 32000 33000 34000 35000 36000 37000 38000

Tot Liquids (bbd)

0 10 20 30 40 50 60 70 80 90 100

0 3 5 8 10 13 15 18 20 23 25

Tot Liquids (bpd)

50 55 60 65 70 75 80 85 90 95 100

Trang 9

Figure 12 Predicted vs Measured Sand Production Rate – Field B, Well 3

Figure 13 Predicted vs Measured Sand Production Rate – Field B, Well 4

Figure 14 Predicted Distribution of Sand Production For Well B / 1 For Specified Producing Conditions

0 4 8 12 16 20

9000 10000 11000 12000 13000 14000 15000 16000 17000 18000 19000 20000 21000 22000

Tot Liquids (bpd)

50 60 70 80 90 100

Sand Production (pptb) Predicted Rate (pptb) WaterCut (%)

0 2 4 6 8 10 12 14 16 18 20

12000 14000 16000 18000 20000 22000 24000 26000 28000 30000 32000

Tot Liquids (bpd)

0 10 20 30 40 50 60 70 80 90 100

0.10 1.00 10.00 100.00 1000.00 10000.00

9750 9775 9800 9825 9850 9875 9900 9925 9950 9975 10000

Depth (feet TVD.SS)

0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5

Formation Permeability (mD) Sand Production Rate (lbs/ft)

Ngày đăng: 26/09/2016, 11:43

TỪ KHÓA LIÊN QUAN

TÀI LIỆU CÙNG NGƯỜI DÙNG

TÀI LIỆU LIÊN QUAN

🧩 Sản phẩm bạn có thể quan tâm

w