Sunteți pe pagina 1din 9

PO179724 DOI: 10.

2118/179724-PA Date: 5-May-16 Stage: Page: 176 Total Pages: 9

Design and Analysis of Leak-Testing


Methodology for Gas Lift Valves
T. Rocha-Valadez, A. R. Hasan, and M. S. Mannan, Texas A&M University; and C. S. Kabir*, Hess Corporation

Summary With safety concerns, periodic testing of GLV integrity is of


Properly operating gas lift valves (GLVs) allow annular gas to significant importance. The current industry practice entails meas-
flow into the tubing and prevent any backflow. However, erosion, uring liquid level in the annulus with the acoustic well-sounding
corrosion, and other effects may cause a GLV to leak, thereby (AWS) method (McCoy et al. 2009; Taylor et al. 2014). An
posing a serious safety issue. Suspected leaking GLVs are often increasing liquid level, estimated by AWS, enables the detection
retrieved for inspection to address safety issues. Proactive testing of faulty in-situ valve(s) when it creates a positive pressure differ-
is a way to minimize costly interventions. Accordingly, nonintru- ential toward the annulus across the GLV. Recently, Rocha-Vala-
sive test procedures have emerged. For example, tests entailing dez et al. (2015) suggested several improvements to GLV test
monitoring of annular transient-pressure response can establish a procedures with or without AWS. The new test procedures recom-
GLV’s integrity by avoiding valve retrieval for inspection. mend drawing down the annular gas to develop a positive pressure
This study details the development of a forward model that differential, thereby allowing fluid flow from the tubing into the
helps design methodology for GLV testing. The model recom- annulus through the suspected GLV. This DD is followed by a
mends drawing down the annular gas to develop a positive pres- pressure-BU test. These pressure-transient tests enable establish-
sure differential to allow fluid flow from the tubing into the ing the integrity of the GLV and the extent of damage, if any.
annulus through the suspected GLV. This drawdown (DD) is fol- However, no forward model exists for design of GLV testing.
lowed by a pressure-buildup (BU) test. These pressure-transient The objective of this study is to present a forward model for
tests allow the determination of the integrity and extent of GLV design and interpretation of GLV integrity tests. To that end, this
damage. The forward model also allows the estimation of the time study expands on our recent approach presented earlier by Rocha-
needed for each segment of the transient-test sequence, thereby Valadez et al. (2015) for interpreting test data for assessing GLV
helping to schedule such tests at a certain operational frequency. integrity. This new model not only helps design a GLV integrity
test, but also enables test interpretation. Specifically, the model
allows estimation of test duration and quantifies the degree of
Introduction
GLV leakage with a type-curve approach. Field data helped vali-
Inherent risks of gas lift operations, involving a large inventory of date the proposed model.
gas and high-operating pressures, dictate that detection and quan-
tification of leaks in GLVs are imperative. Prevention of uncon-
trolled release of hydrocarbons forms the backbone of barrier Methodology
philosophy in modern well construction and operation. Quoting The basis of the proposed test methodology is depicted in Fig. 1.
Norwegian guidelines, Carlsen et al. (2010) stated that “well bar- As Fig. 1a shows, even if the GLV has a leak, liquid from the tub-
riers are envelopes of one or several dependent well barrier ele- ing will not flow into the annulus if there is no pressure differen-
ments preventing fluids or gases from flowing unintentionally tial. If a positive pressure differential is created across a faulty
from the formation into another formation or to surface.” They GLV by venting some of the annular gas, liquid from tubing will
cited several reasons for GLV failures, including flow cutting in flow into the annulus, as Fig. 1b indicates. If the valve at the
sealing seats caused by well unloading of completion fluids, the casinghead is closed after a period of venting, liquid coming in
valve-spring fatigue failure, damage during installation, erosion from the tubing will then compress the gas in the annulus, result-
during multiphase or wet gas injection, and damage to valve seat ing in pressure increase with shut-in time; Figs. 1c and 1d illus-
caused by chattering in a low-pressure-drop situation. trate this point. Then, one can analyze the rate of casing-pressure
Gilbertson et al. (2013) developed a thermally actuated GLV rise to obtain a quantitative measure of GLV damage.
for offshore wells to add redundancy to well systems and signifi- This methodology recommends venting through an appropriate
cantly to increase well safety. This valve avoids the use of either choke such that sonic velocity occurs, allowing venting rate to be
active control or passive pressure-induced actuation that is per- known (Fig. 1b). To avoid a casing-integrity issue arising from
formed in current valve systems. By modeling heat transfer and excessive pressure differential, venting is terminated when the
laboratory prototype construction, they showed that this safety casinghead pressure reaches a certain level, prisk (Fig. 1c). Obvi-
valve is applicable to wells with at least a 13  C temperature dif- ously, if the initial annular pressure was not high enough, venting
ference across the valve. would cease when casinghead pressure nears atmospheric pres-
To assess the well-integrity risk, Dethlefs and Chastain (2012) sure (or venting-line pressure), as shown in Fig. 1d.
reported formulation of a qualitative model. Their modeling The test methodology and model formulation are described in
approach of risk assessment entails comparison between well-bar- the subsequent sections. However, to get a perspective, we present
rier failure-mode scenarios, and qualitatively assesses well-barrier how one can use the gathered data to determine the extent of
failure risk for every segment of a well, among others. One may GLV damage, if any. When a GLV is damaged and fluid flows
apply this risk-assessment approach to a given well, a group of through it from the tubing into the annulus, we may visualize the
similar wells, or an entire well portfolio in a field. Julian et al. flow process as one through an orifice, with e fraction of the total
(2014), citing long-standing operational experiences in Prudhoe open cross-sectional area of the valve. In this context, e may be
Bay, Alaska, suggested that the barrier-qualified GLVs now meet viewed as a damage coefficient, meaning a value of e ¼ 0% repre-
performance metrics. sents a fault-free GLV, whereas e ¼ 100% implies a wide-open
GLV check valve.
Fig. 2 shows pressure or liquid level (dependent variable) rise
*
Retired; now with CS Kabir Consulting in the annulus as a function of time for various valve-damage
Copyright V
C 2016 Society of Petroleum Engineers coefficient values, e. The solid lines represent simulated values of
Original SPE manuscript received for review 1 July 2015. Revised manuscript received for the dependent variable for various degrees of GLV damage. The
review 24 October 2015. Paper (SPE 179724) peer approved 23 November 2015. open circles show data from an actual test. Fig. 2, therefore,

176 May 2016 SPE Production & Operations

ID: jaganm Time: 21:08 I Path: S:/PO##/Vol00000/160007/Comp/APPFile/SA-PO##160007


PO179724 DOI: 10.2118/179724-PA Date: 5-May-16 Stage: Page: 177 Total Pages: 9

pT pA pa pT pA pa pT prisk pa pT pA = pa

mc

LgA

gas

gas

gas

gas
LTotal LgA LTotal LgA LTotal LgA LTotal
Liquid

Liquid

Liquid

Liquid
LlA

LlA

LlA

(a) Initial conditions, t = 0 (b) After, t = Δt (c) Stopping condition (d) Stopping condition
pGT = pGA pGT > pGA pA ≤ prisk p A ≥ pa

Fig. 1—Changes in liquid level in the system caused by flow through the GLV during DD, followed by a BU test.

indicates that this GLV is faulty, and approximately 35% of the to calculate liquid-ingress rate into the annulus. We recommend
check-valve area is “open.” the use of a flow-prover at the casinghead so that critical velocity
The test methodology consists of opening the casinghead is attained for the exiting gas, allowing estimation of its rate. Note
valve, drawing down the annular gas (DD test), followed by a that the inside-to-outside pressure ratio for the annular gas is so
pressure-BU test, conducted by closing the casinghead valve. high that attaining critical gas velocity through any choke will be
The methodology relies on calculating the mass of a known easily attained.
volume of gas in the annulus with the gas law and its pressure The pressure differential across the GLV affects the flow from
and temperature. the tubing into the annulus. Because any gas in the tubing fluid
For a buildup test when gas mass does not change with time, moves up quickly to the wellhead after well shut-in, we assume
rate of change of casinghead pressure is used to calculate the rate that only liquid flows from the tubing into the annulus through a
of change in gas volume, which is equal to the liquid ingress rate faulty GLV. In addition, we note that, after the tubinghead is
through the GLV. For a DD test, the rate at which gas is escaping closed, fluid will still flow (afterflow) from the reservoir into the
the annulus needs to be known to set up a proper material balance well, given the compressible wellbore fluid. The afterflow will ei-
ther raise the tubinghead pressure or flow across the GLV into the
annulus. If the tubinghead pressure remains essentially constant,
all of the fluid afterflow from the reservoir enters the annulus by
means of the faulty GLV.
100%
50% Note that one can use the approach just described in conjunc-
96 35% tion with the AWS data acquisition to provide more confidence
25%
in the estimates. The steps for the test procedure are enumer-
15%
ated next:
Dependent Variable

10%
1. Begin the test by turning off gas injection and closing the
tubinghead valve. Allow some time for the gas bubbles in
the tubing to escape before shutting in the valve. We
5% assume that the pressure in the annulus and that in the tub-
95 ing at GLV depth will essentially be the same.
2. Begin the DD test by opening the casinghead valve and let-
ting gas flow through a choke (of known size) so that the
0%
critical sonic velocity is attained, thereby allowing estima-
tion of the exiting gas-flow rate.
3. Terminate the DD test when the annular pressure reaches
Data
60% ( prisk) of its original value to maintain a safe operating
94 pressure. The shortest possible time to reach prisk is calcu-
0 10 20 30 40 50 60 70 lated with the known gas rate and assuming that no liquid
Time inflow from the tubing occurs into the annulus; that is, GLV
damage coefficient is assumed to be zero. This step ensures
Fig. 2—Dependent-variable (pressure or liquid level) profiles sufficient pressure differential across the GLV for the BU
for several e values. test that follows.

May 2016 SPE Production & Operations 177

ID: jaganm Time: 21:08 I Path: S:/PO##/Vol00000/160007/Comp/APPFile/SA-PO##160007


PO179724 DOI: 10.2118/179724-PA Date: 5-May-16 Stage: Page: 178 Total Pages: 9

4. Begin BU test by monitoring the annular wellhead pressure LT ¼ LgA þ LlA : . . . . . . . . . . . . . . . . . . . . . . . . . . . ð4Þ
after closing the casinghead valve (Step 3). Perform perti-
nent gas-compression calculations and liquid-influx
estimation. Hydrostatic Pressure. The density of a gas can be related to its
5. Compare the pressure profile from DD and/or BU test with pressure and temperature by q ¼ ( pM)/(ZRT). Therefore, the
theoretical pressure isopleths, described later. The match hydrostatic pressure, dp, of a gas column of vertical height dz is
obtained with the test data and theoretical pressure isopleths given by the following equation:
allows estimation of the damage coefficient, e, which is a pM
reflection of the extent of GLV damage. dp ¼ qdz ¼ dz; . . . . . . . . . . . . . . . . . . . . . . . . ð5Þ
ZRT
where M ¼ 29cg is the molecular weight of the gas and cg is the
specific gravity (SG) of gas with respect to air. Note that dz is the
Model Formulation vertical depth that can be related to measured depth (MD) L, as
The test begins when pressure in the tubing and in the annulus dz ¼ dL (cos h), where h is the deviation angle from the vertical.
across the GLV is equal. If there is a leak in the GLV, the liquid Eq. 5 may be easily integrated over a short depth with average
originally in the tubing will enter the annulus at depletion of the temperature and Z-factor for the depth, yielding
annular gas because of the pressure differential. Our experience  
with a number of such tests indicates that the wellhead tubing 29cg ðz2  z1 Þ
p2 ¼ p1 exp : . . . . . . . . . . . . . . . . . . . ð6Þ
pressure pT remains practically constant even when liquid moves RZT
from the annulus into the tubing. This response is not surprising
because the compressible tubular fluid column masks any subtle We compute the annular-gas pressure from the known pressure
perturbation-created downhole. The operating envelope of the at the casinghead, and march downward in n steps to the gas/
tubinghead pressure during the complete test period (DD, BU, liquid interface. Therefore, the gas pressure at this position pGA in
and AWS measurements) remained within 415 6 15 psi for Well terms of the hydrostatic pressure of the liquid and gas columns is
1,418 6 15 psi for Well 2, and 1,474 6 30 psi for Well 3, varying given by
only approximately 2% from the average value.  
Xn
29gcg ðzgA;iþ1  zgA;i Þ
We observed that, if a leak exists, the liquid column above the pGA ¼ qL zlA g þ pgA;i exp : . . . .ð7Þ
GLV depth in the tubing remains unchanged and that the liquid i¼0 Z i RT i
entering the annulus originates from the formation. Therefore, the
tubing pressure at GLV depth can be assumed constant. We also We use the same iterative solutions approach to estimate the
assume that the flow of gas from the annulus through a choke mass of gas in the annulus, mgA at each timestep. However, there
onto a flowline will occur at the critical rate. Critical gas-flow is an added complexity for computations for t > 0, because the
conditions exist when the inlet-to-outlet pressure ratio is approxi- length of the gas column changes with time as liquid enters the
mately 1.7, which is very likely to be satisfied during these tests. annulus from the tubing. Because pGA depends on liquid level
Our model is based on mass and volume balance of the annular above it, and the liquid influx rate, which, in turn, depends on
gas and liquid, and estimation of hydrostatic pressures exerted by pGA, the computation becomes iterative. We need to solve for
the gas and liquid columns. mass of gas remaining in the annulus, represented by Eq. 8 listed
next, simultaneously with Eq. 7 for annular pressure at the GLV
to obtain pGA, zLA, and mGA:
Mass Balance. The mass conservation for the liquid and the gas
Xn
29cg ðLgA;iþ1  LgA;i ÞAA ðpgA;i þ pgA;iþ1 Þ
in the system is represented by Eqs. 1 and 2 noted next: mGA ¼ : . . . . ð8Þ
Mass of liquid in the annulus, i¼0 2Z i RT i
ð
mlA ¼ ql LlAo AA þ m_ l dt ¼ ql LlA AA . . . . . . . . . . . . . . ð1Þ Computational Scheme. One of the objectives of this model is
to allow for estimating the time t60 needed for the annular pres-
Mass of gas in annulus, sure to go down to 60% of its initial value, prisk, through a known
ð (critical) rate of gas DD. The idea is to allow annular pressure to
decrease to a minimum allowable annular wellhead pressure
mgA ¼ mgAo  m_ g dt: . . . . . . . . . . . . . . . . . . . . . . . ð2Þ (MAAWP). If annular wellhead pressure decreases beyond prisk,
the risk of damaging the casing becomes a possibility. The time at
In these equations, m represents mass, m_ is mass rate, L represents which prisk is reached will vary according to the damage coeffi-
length above GLV, and subscripts A, c, l, g, and o, represent annu- cient of the GLV. We assume that the damage coefficient is zero;
lus, critical, liquid, gas, and values at initial conditions, that is, GLV is intact (e ¼ 0), to obtain a conservative estimate of
respectively. t60. The liquid flow rate m_ l through GLV is calculated with the
The initial mass of gas in the annulus is calculated with the following general expression for flow through orifices:
gas law. However, because the pressure in the annulus varies sig- pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
nificantly over the well depth, we discretize the annulus into n m_ l ¼ eA 2ql ð pGT  pGA Þ: . . . . . . . . . . . . . . . . . . . . ð9Þ
sections and calculate the mass and pressure in each section itera-
tively. Eq. 2, therefore, is rewritten in the following form that Because pressure is a continuous variable and the methodol-
sums the mass of gas in each cell: ogy proposed discretizes this variable, the annular pressure at the
ð GLV depth should use an average pressure between the previous
Xn
29cg ðLgA;iþ1  LgA;i ÞAA ðpgA;i þ pgA;iþ1 Þ
mgA ¼  m_ g dt: timestep and the current one, that is, ( pGA,i þ pGA,i-1)/2.
i¼0 2Z i RT i The magnitude of pressures at the casinghead is much higher
than the outside pressure where gas is discharged, thereby making
                   ð3Þ the gas flow from the casinghead critical. Therefore, we calculate
Eqs. 1 and 3 represent the mass balance of each component of the the casinghead gas-flow rate m_ g with the following critical
system and are valid at any given time. To obtain the initial gas flow equation:
mass in the annulus, we use Eq. 3 by treating the entire length vffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
u  k þ 1
above the GLV as gas. At any other time, because of liquid u
t 2 k1
ingress from the tubing, the lengths of the liquid and gas columns m_ g ¼ Co A qg pA : . . . . . . . . . . . . . . . . . .ð10Þ
kþ1
add up to the same total length, LT:

178 May 2016 SPE Production & Operations

ID: jaganm Time: 21:08 I Path: S:/PO##/Vol00000/160007/Comp/APPFile/SA-PO##160007


PO179724 DOI: 10.2118/179724-PA Date: 5-May-16 Stage: Page: 179 Total Pages: 9

pT pA Valve closed pT pA

mc = 0 mc = 0

gas

gas
LgT
LgT

LgA

Liquid

Liquid
LTotal LgA LTotal
LlT
LlT

LlA

LlA

(a) At, t = 0 (b) Stopping conditions at


pGT > pGA t = tfinal, pGT = pGA

Fig. 3—Changes occurring in the system caused by flow through GLV during the BU test.

In Eq. 10, Co is the valve’s orifice discharge coefficient; A is The procedure to estimate time needed for 40% annulus pres-
its cross-sectional area; and qg, k, and pA are the gas density, its sure reduction is the following:
specific heat ratio, and its pressure at the casinghead, respectively. 1. Estimate the mass of gas in the annulus after a given time-
All these properties are known from the top-down calculation and step Dt for a e ¼ 0.
should be used at each timestep. 2. For the next timestep, use an initial pA value, smaller than
In the following section, we discuss the computational scheme in the previous timestep.
used. At initial conditions, pressure equilibrium exists, and there 3. Verify that pA satisfies gas properties for mass mgA ¼ pV/
is no flow through the GLV. At these conditions, one can establish (29cgRT ) and length of gas chamber LgA.
the pressure at GLV depth in both the annulus and the tubing. 4. Repeat Steps 3 and 4 until convergence is achieved.
The procedure to calculate the mass and pressure of the gas 5. Estimate new pGA with Eq. 7 along with zTotal ¼ zgA þ zlA
column is as follows: and use Eq. 8 to estimate amount of mass of liquid that
1. We assume that in the beginning there is no liquid above entered the annular section.
the GLV in the annulus; thus, the length of the gas column 6. Use density of the oil (qL) and geometry of the annulus to
is the same as the depth of the GLV, LgA ¼ LT. Because the estimate the height occupied by liquid (LlA and zlA).
casinghead initial pressure pA is known, we estimate the 7. Calculate the change in gas and liquid columns with Eqs. 3
pressure at GLV depth pGA and the total mass of annular and 4 and use as input for next timestep.
gas mgA with Eqs. 7 and 8. 8. Repeat Steps 2 through 8 until t ¼ tfinal, pA ¼ prisk or pA ¼ 0
2. With the known annular initial mass, we proceed to esti- psig.
mate these quantities at later times allowing for gas bleed- The procedure for a BU test is the same as that for a DD test,
off. The reduction in the amount of annular gas, with except that the gas mass, mgA, remains constant during the entire
consequent pressure reduction, allows liquid to flow from test. In Appendix A, we present a flow diagram for the methodol-
the tubing into the annulus through a faulty GLV. This and ogy presented previously.
several important stages of the test are sketched in Figs. 1
and 3 for DD and BU tests, respectively.
The damage coefficient can range from 0  e  emax. We used Model Validation
e ¼ 0 to calculate t60. For a damaged GLV when liquid coming in We compared our simulated pressure vs. time profiles from BU
from the tubing will support the annulus pressure, t60 calculated test with those performed in three gas lift wells with AWS data.
that this way will lead to lesser DD, leading to safer The DD-tests data acquired during the tests could not be used to
test environment. validate the model because the bleedoff rate for gas discharge
remained unmeasured. The gas-discharge rate is a required pa-
rameter in our model and the reason for recommending a flow-
meter at the casinghead for DD tests.
We used the liquid level measured by AWS and annular pres-
sure data at the beginning of the BU test as starting conditions;
the proposed model was used thereafter to track all changes. The
starting conditions of Wells 1, 2, and 3 are given in Table 1.
We simulated pressure BU over a time for several GLV dam-
age coefficients, and compared the estimated values with the pres-
Table 1—Initial well conditions for BU tests. sure reported from the field tests. Figs. 4 through 6 show the

May 2016 SPE Production & Operations 179

ID: jaganm Time: 21:08 I Path: S:/PO##/Vol00000/160007/Comp/APPFile/SA-PO##160007


PO179724 DOI: 10.2118/179724-PA Date: 5-May-16 Stage: Page: 180 Total Pages: 9

1,400
1,520
75% 60% 50%
1,395 40%

Annular Wellhead Pressure (psi)


Annular Wellhead Pressure (psi)

50%
35% 1,510
25%
1,390

1,500
1,385

1,490
1,380

1,480
1,375

1,470
1,370
0 10 20 30 40 50 60 0 5 10 15 20 25 30 35

Time (minutes) Time (minutes)

Fig. 4—Comparison of pressure profiles during BU Tests, Well 1. Fig. 5—Comparison of pressure profiles during BU Tests. Well 2.

1,086.0 Fig. 4 shows that the pressure BU for Well 1 is similar to the
profile resulting from an orifice of 35% of the cross-sectional area
of the GLV. A 35% damaged area would be the equivalent to hav-
1,085.5 ing a completely open valve with a diameter of 0.3 in. or
Annular Wellhead Pressure (psi)

5% 7.25 mm. Fig. 5 represents the pressure response for Well 2. The
data indicated a damage of 60% for the GLV, equivalent to hav-
1,085.0
3%
ing an orifice of 0.39 in., approximately 80% of the original diam-
eter. The severity of the calculated leak matches the field test with
AWS for this well.
1,084.5 Fig. 6 showing the test results for Well 3 suggests essentially no
pressure increase for the entire duration of the test. The pressure
1% rise of less than 1.5 psi (magnified by the small range of the y-axis)
1,084.0 could have resulted from temperature or other operational fluctua-
tions. This test is an indication that if a leak exists, it is minimal.
0% The flow-rate estimations for all wells for various damage-
1,083.5
coefficient values are presented in Table 2. A comparison of the
best fit, according to the pressure-profile contours, is given in
Table 3. In Table 2, the damage coefficients emphasized in bold
1,083.0
0 10 20 30 40 50 60 type correspond to the profile that had the best match to the
reported pressure data.
Time (minutes) Well 1 and Well 3 flow-rate estimates match well with those
reported in field measurements. This test helped us conclude that
Fig. 6—Comparison of pressure profiles during BU Tests, Well 3. a leak exists in Well 1 with a 35% GLV damage. The difference
in Well 3 test data and model estimation most likely corresponds
to temperature fluctuations or noise in the equipment. We con-
clude that there is probably no leak in the GLV of this well. In
pressure contours for these BU tests. In these figures, the open contrast, Well 2 shows a severe leak that represents a damage
circles represent data whereas the solid lines represent annular flow area of 60%. The difference between our estimate and the
pressure calculated with various e values. For all these wells, the measured data can be related to the fact that the test was termi-
GLV aperture was 0.5-in. inner diameter (ID). nated because of severe liquid leak after just 5 minutes of testing,

Table 2—Estimated flow rates for several GLV damage coefficients in three wells.

180 May 2016 SPE Production & Operations

ID: jaganm Time: 21:08 I Path: S:/PO##/Vol00000/160007/Comp/APPFile/SA-PO##160007


PO179724 DOI: 10.2118/179724-PA Date: 5-May-16 Stage: Page: 181 Total Pages: 9

Table 3—Comparison of flow rates from this model and AWS measurement.

Table 4—Parameters at the end of DD-1.

allowing insufficient time for a proper test. On the basis of the during DD test and serves as input for the subsequent BU test,
performance of these tests, we recommend BU times of at least wherein flow rates are calculated to assess the GLV damage.
15 minutes. For this well, the initial wellhead annular pressure pA is 1,727
psig, the initial-liquid level depth LlA is 8,740 ft, and calculated
Case Studies pressure at GLV depth pgA is 2,312 psig. These input parameters
return an initial amount of annular gas of 493 Mscf. From that
Two case studies are presented in this section. The first one shows point forward, the gas is released through a 0.375-in.-ID choke
the general application of the test procedure for a vertical well. In with a discharge coefficient Co ¼ 0.61 under the critical-flow con-
this case, we assume that only initial conditions are known and dition. The DD continues until the pressure reaches 60% of
that there are no data to compare with the simulated values. This the original pA, which is approximately 1,037 psig in this case.
case study shows how the simulated pressure and liquid-level pro- Table 4 shows important parameters for the well at the end of the
files can be examined to assess actual data. DD test, which also serves as the initial condition for the BU test.
The second case is based on Well 1, for which we have the Note that Table 4 shows simulation results for 144 minutes of
most-complete set of data. Data from the BU test from this well DD for all e values. This is the t60 (MAAWP) only for e ¼ 0 (a
were already used to validate the model. In this section, we show faultless GLV); for all other e values, the annular pressure drop
the full simulated pressure profiles for both the DD and BU tests, will be less than 60% after 144 minutes of DD, resulting in a safer
as well as choke size needed to achieve comparable pressure pro- conduct of the GLV test. Figs. 7 through 9 show the variation in
file as the reported data. pressures at the casinghead pA and at the GLV depth pgA with
time. In each figure, the left side corresponds to the DD segment
Case 1. This case study involves the application of the forward of the test whereas the right side corresponds to the BU test.
model in a vertical well that avoids the need for an AWS test. During a GLV test, the casinghead pressure vs. time data for
This vertical well contains the GLV at a depth of 8,740 ft. We both DD and BU tests can be plotted and compared in Fig. 7
assume that, at the beginning of the DD test, the liquid level is against the simulated values. Matching field data will help obtain
just below the GLV. The mass of gas in the annulus is tracked a e value, which is a measure of valve damage. Alternatively, one

1,800
ε, % 2,400
1,700 0
Annular Wellhead Pressure (psi)

5 2,200
Annular Wellhead Pressure (psi)

1,600 10
15
25 2,000
1,500
50
100 1,800
1,400 Start of BU

1,600 ε, %
1,300
0
5
1,200 1,400 10
15
1,200 25
1,100 50 Start of BU
100
1,000 1,000
0 100 200 300 400 0 100 200 300 400

Time (minutes) Time (minutes)

Fig. 7—Annular wellhead-pressure profile for full GLV test. Fig. 8—Annular-pressure profile at GLV depth for full GLV test.

May 2016 SPE Production & Operations 181

ID: jaganm Time: 21:08 I Path: S:/PO##/Vol00000/160007/Comp/APPFile/SA-PO##160007


PO179724 DOI: 10.2118/179724-PA Date: 5-May-16 Stage: Page: 182 Total Pages: 9

6,000
ε, %
0 Start of BU
6,500 5
10
15
25
Liqiuid Level (feet)

7,000
50
100
7,500

8,000

8,500

9,000
0 50 100 150 200 250 300 350 400

Time (minutes)

Fig. 9—Liquid-level profile for full GLV test.


Table 5—Data used for GLV leak estimation with proposed
methodology.
can use the “preliminary” results from a DD test to design a better
and shorter BU test. A DD test that indicates severe leak in a
GLV can be used to terminate the test because valve retrieval and
estimation from pressure data with the methodology shown earlier
replacement have become obvious.
by Rocha-Valadez et al. (2015).
We think that critical flow conditions prevailed during the
Case 2. Well 1 was selected for this case study because it has the AWS test for Well 1; the model’s ability to match data with the
most-complete set of data. Unfortunately, the gas rate exiting the critical gas-discharge rate instills confidence in solution quality.
annulus was not measured nor the choke size at the casinghead As mentioned earlier, critical flow is likely to occur with the mag-
reported. However, the high inside/outside pressure ratio allows nitude of pressure ratio between the annulus and the flowline.
us to presume that critical flow was achieved. The simulated Finally, with specific parameters and variability between valves,
results are based on critical flow for a choke valve with a diameter such as the flow coefficient, we cannot be sure that a valve of
such that the initial conditions for the BU test at the end of the 0.225-in. ID was used. However, we showed that with the perti-
DD test was matched. The initial pA was 1,748 psig, and the faulty nent information, the process of bleeding off a well followed
GLV was at a depth of 8,286 ft, giving a pressure in the annulus by pressure BU test can be modeled and the damage to a
at GLV depth pGA of approximately 2,400 psig. This result is sim- GLV evaluated.
ilar to the estimated pGT obtained with AWS data, as reported in
Table 1. This outcome reaffirms that the initial hydrostatic equi-
librium assumption is reasonable. The data reported in the test are Discussion
shown in Table 5. We note that there are some discrepancies This paper introduces a forward model that is anchored in analyz-
between the end time for DD and start time for BU tests in the ing the annular-pressure transients. A type-curve matching
reported data. approach of annular pressures conveniently determines the magni-
With data in Table 5, we computed the casinghead pressure tude of GLV damage, if any. This forward model forms the back-
for the DD test followed by the BU test. Fig. 10 suggests good bone of the test-design procedure, which the methodology section
agreement between the model estimates (solid line) and data details. In summary, we recommend a constant-rate DD of the an-
(open circles). Fig. 11 shows the liquid-level profile when the two nular gas followed by a BU test. The pressure profiles obtained
tests are performed successively. Fig. 11 plots the liquid-level from DD and/or BU test should then be compared with

1,800 6,500
ε = 35%, Dchoke = 0.225 in., Co = 0.61
ε = 35%, Dchoke = 0.225 in., Co = 0.61
6,750
Annular Wellhead Pressure (psi)

1,700
7,000
Liqiuid Level (feet)

1,600 7,250
Start of BU
7,500 6,630
Liqiuid Level (feet)

1,500 7,750 6,665


Start of BU
6,700
8,000 6,735
1,400
6,770
8,250
240 250 260 270
Time (minutes)
1,300 8,500
0 50 100 150 200 250 300 0 50 100 150 200 250 300

Time (minutes) Time (minutes)

Fig. 10—Pressure profile generated with proposed model Fig. 11—Liquid-level profile with proposed methodology with
mimics field data, Well 1. two different thermodynamic models.

182 May 2016 SPE Production & Operations

ID: jaganm Time: 21:08 I Path: S:/PO##/Vol00000/160007/Comp/APPFile/SA-PO##160007


PO179724 DOI: 10.2118/179724-PA Date: 5-May-16 Stage: Page: 183 Total Pages: 9

theoretical-pressure isopleths, shown in Figs. 7 and/or 8, to esti- zgA ¼ true vertical depth (TVD) of gas column in the annulus,
mate the GLV damage coefficient, e. ft
The proposed model is equally capable of designing tests and zlA ¼ true depth of liquid column in the annulus, ft
analyzing test data for assessing the integrity of suspected GLVs. zTotal ¼ total TVD from casing head to GLV, ft
The nonintrusive nature and short duration of the test make the e ¼ GLV damage coefficient or fraction of valve area open
recommended procedure very pragmatic. Indeed, whenever gas caused by damage, dimensionless
injection is suspended because of operational reasons in a gas-lift k ¼ heat-capacity ratio, dimensionless
well, this test can be run with minimal expenditure in terms of qgA ¼ density of gas in annulus, lbm/ft3
time and effort to determine the health of the system’s GLVs. In ql ¼ density of liquid, lbm/ft3
this context, we recognize that gas leakage does not implicitly h ¼ well-inclination (to horizontal) angle, degrees
suggest that a faulty GLV is responsible (i.e., because integrity of
tubing and/or cement sheath may also play a role). But, only test-
ing followed by valve replacement can resolve this question. References
Note that the loss of cement-integrity triggering sustained cas- Altun, G., Langlinais, J., and Bourgoyne, A. T. 2001. Application of a
ing pressure in an annulus has a different approach to testing; one New Model to Analyze Leak-off Tests. SPE Drill & Compl 16 (2):
can find details elsewhere (for example, see Huerta et al. 2009; 108–116. SPE-72061-PA. http://dx.doi.org/10.2118/72061-PA.
Watson and Bachu 2009; Rocha-Valadez et al. 2014a, 2014b). Carlsen, J. A., Stokka, Ø., and Kleppa, E. 2010. Taking the Gas Lift
The casing-integrity problem may also arise from the failure of Valves to a New Level of Reliability. Presented at the Offshore Tech-
casing shoes. To diagnose such a problem, the formation leakoff nology Conference, Houston, 3–6 May. SPE-20820-MS. http://
test, suggested by Wang et al. (2011) and Altun et al. (2001), can dx.doi.org/10.2118/20820-MS.
reveal whether the mud pressure at a certain depth exceeds Dethlefs, J. and Chastain, B. 2012. Assessing Well-Itegrity Risk: A Quali-
the ability of the formation to contain it at the casing shoe tative Model. SPE Drill & Compl 27 (2): 294–302. SPE-142854-PA.
after cementing. http://dx.doi.org/10.2118/142854-PA.
Gilbertson, E., Hover, F., and Freeman, B. 2013. A Thermally Actuated
Gas-Lift Safety Valve. SPE Prod & Oper 28 (1): 77–84. SPE-161930-
Conclusions PA. http://dx.doi.org/10.2118/161930-PA.
1. This study presents a novel forward model that allows test Huerta, N. J., Checkai, D. A., and Bryant, S. L. 2009. Utilizing Sustained
design and quantifies the degree of leakage across the GLV fol- Casing Pressure Analog to Provide Parameters to Study CO2 Leakage
lowing the test. The system-response curves involving annular- Rates Along a Wellbore. Presented at the SPE International Confer-
pressure change and/or liquid level as a function of time consti- ence on CO2 capture, Storage, and Utilization, San Diego, California,
tute the type curves corresponding to various degrees of GLV USA, 2–4 November. SPE-126700-MS. http://dx.doi.org/10.2118/
damage, reflecting the severity of leak. 126700-MS.
2. Although intended as a forward model for test design, this tool Julian, J. Y., Jackson, J. C., and White, T. M. 2014. A History of Gas Lift
also lends itself for interpretation of test data. In fact, data Valve and Gas Lift Mandrel Damage and Subsequent Retrofit Gas Lift
from three wells validated the modeling approach used in this Straddle Installation in Alaska. Presented at the SPE/CoTA Coiled
study. In these tests, leak severity ranged from zero to 60%. Tubing and Well Intervention Conference and Exhibition, The Wood-
Assessing leak severity may help prioritize the valve-replace- lands, Texas, USA, 25–26 March. SPE-168304-MS. http://dx.doi.org/
ment schedule in a given asset. 10.2118/168304-MS.
3. This paper presents a stepwise nonintrusive test approach for McCoy, J. N., Rowlan, O. L., and Podio, A. 2009. Acoustic Liquid Level
collecting valid test data. Generally speaking, collection of Testing of Gas Wells. Proc., SPE Production and Operations Sympo-
wellhead pressure as a function of time allows estimation of sium, Oklahoma City, Oklahoma, USA, 4–8 April. SPE-120643-MS.
leak severity of the GLV. http://dx.doi.org/10.2118/120643-MS.
Rocha-Valadez, T., Hasan, A. R., Mannan, M. S. et al. 2015. Assessing In-
Nomenclature tegrity of the Gas-Lift Valves by Analyzing Annular-Pressure-Tran-
sient Response. Journal of Petroleum Science and Engineering 133:
A ¼ GLV bellow cross-sectional area, ft2
177–183. http://dx.doi.org./10.1016/j.petrol.2015.05.002.
AA ¼ annular cross-sectional area, ft2
Rocha-Valadez, T., Hasan, A. R., Mannan, M. S. et al. 2014a. Assessing
AT ¼ tubular cross-sectional area, ft2
Wellbore Integrity in Sustained-Casing-Pressure Annulus. SPE Drill
g ¼ gravitational acceleration, ft/sec2
& Compl 29 (1): 131–138. SPE-169814-PA. http://dx.doi.org/10.2118/
gG ¼ geothermal gradient, 0.0353  F/ft
169814-PA.
LgA ¼ MD of gas column in the annulus, ft
Rocha-Valadez, T., Metzger, R. A., Hasan, A. R. et al. 2014b. Inherently
LlA ¼ MD of liquid column in the annulus, ft
Safer Sustained Casing Pressure Testing for Well Integrity Evaluation.
LT ¼ total measured-depth length from casinghead to GLV, ft
J. Loss Prevention in the Process Industries 29: 209–215. http://
mgA ¼ mass of gas in the annulus at any given time, lbm
dx.doi.org/10.1016/j.jlp.2014.02.012.
mlA ¼ mass of liquid in the annulus at any given time, lbm
Taylor, C., Rowlan, O. L., and McCoy, J. 2014. Acoustic Techniques to
m_ c ¼ mass flow of gas from choke flow on wellhead, scf/hr
Monitor and Troubleshoot Gas-Lift Wells. Proc., SPE Western North
m_ l ¼ mass-flow rate of liquid through GLV, lbm/hr
American and Rocky Mountain Joint Regional Meeting, Denver,
n ¼ number of cells for calculation to liquid level,
USA, 17–18 April. SPE-169536-MS. http://dx.doi.org/10.2118/
dimensionless
169536-MS.
p ¼ pressure, psia
Wang, H., Soliman, M. Y., Shan, Z. et al. 2011. Understanding the Effects
p1, p2 ¼ pressure at Location 1 or 2, psia
of Leakoff Tests on Wellbore Strength. SPE Drill & Compl 26 (4):
pa ¼ pressure of choke-flow outlet (¼14.7), psia
531–539. SPE-132981-PA. http://dx.doi.org/10.2118/132981-PA.
pA ¼ annulus wellhead pressure, psi
Watson, T. L. and Bachu, S. 2009. Evaluation of the Potential for Gas and
pGA ¼ annulus pressure at GLV depth, psi
CO2 Leakage Along Wellbores. SPE Drill & Compl 24 (1): 115–126.
pGT ¼ tubing pressure at GLV depth, psi
SPE-106817-PA. http://dx.doi.org/10.2118/106817-PA.
prisk ¼ minimum allowable annular wellhead pressure, psi
pT ¼ tubing wellhead pressure, psi
R ¼ gas constant (¼10.731), ft3-psi/( R-lbm mol)
T ¼ temperature,  R Appendix A—Methodology for Test Design
Twh ¼ wellhead temperature,  R The methodology discussed in the main body of the paper for a
cg ¼ gas SG DD test is shown as a flow diagram in Fig. A-1. The procedure
Z ¼ gas-compressibility factor, dimensionless for calculating the changes during a BU test is fundamentally the

May 2016 SPE Production & Operations 183

ID: jaganm Time: 21:08 I Path: S:/PO##/Vol00000/160007/Comp/APPFile/SA-PO##160007


PO179724 DOI: 10.2118/179724-PA Date: 5-May-16 Stage: Page: 184 Total Pages: 9

process safety, risk assessment, and management in both


Estimate mgA and pGA upstream and downstream sectors of oil and gas industry. He
from known pA and LgA holds BS and MS degrees in chemical engineering from the
at t = 0 Technology Institute of Celaya, Mexico, and a PhD degree
from Texas A&M University.
A. Rashid Hasan is professor of petroleum engineering at Texas
Increase t = ti +Δt
A & M University. He has more than 35 years of teaching, con-
sulting, and research experience. Hasan is an expert in the
area of production engineering; he focuses on modeling
complex transport processes in various components of petro-
Calculate new mgA,
leum-production systems. He has also worked with the
using Eq. 2 National Aeronautics and Space Administration on various
aspects of multiphase flow and thermohydraulic transients.
Hasan has published extensively, and is a coauthor of the
book Fluid Flow and Heat Transfer in Wellbores. He has served
Assume pA for on various SPE committees, including editorial review for SPE
current timestep Production and Operations, SPE Production and Facilities, and
NO
SPE Journal. Hasan was the recipient of the 2011 SPE Produc-
tion and Operations Award, and he achieved SPE Distin-
guished Member status in 2015. He earned his MS and PhD
' =m
mgA gA degrees in chemical engineering from the University of Water-
loo, Canada.
YES
M. Sam Mannan is Regents Professor in the Chemical Engineer-
Estimate pGA and ml ing Department at Texas A&M University and Executive Direc-
with Eqs. 5 and 7, tor of the Mary Kay O’Connor Process Safety Center at the
respectively Texas Engineering Experiment Station. Before joining Texas
A&M University, he was vice president at RMT, a nationwide
engineering-services company. Mannan’s experience is wide-
Calculate liquid
ranging, covering process design of chemical plants and refin-
column heights eries, computer simulation of engineering problems, mathe-
(LgA, LlA, zgA and zlA) matical modeling, process safety, risk assessment, inherently
for the next time step safer design, critical infrastructure vulnerability assessment,
aerosol modeling, and reactive and energetic materials
assessments. He has published 244 peer-reviewed journal pub-
NO lications, four books, eight book chapters, 214 proceedings
papers, and 14 major reports. Mannan is a registered profes-
pA ≤ prisk,0 sional engineer in the states of Texas and Louisiana, a Certified
or t = tfinal Safety Professional, and a Professional Process Safety Engi-
neer. He earned a BS degree in chemical engineering from
YES Bangladesh University of Engineering and Technology in
Dhaka, Bangladesh and earned an MS degree and a PhD
END degree in chemical engineering from the University of
Oklahoma.
Fig. A-1—Solution methodology for estimating GLV-test Shah Kabir is the proprietor of CS Kabir Consulting. He recently
duration. retired from Hess Corporation after working 39 years in the
industry, with stints at Chevron and Schlumberger. Kabir’s ex-
same as that for a DD test, with the constraint that mgA remains perience includes pressure- and rate-transient analyses, reser-
voir engineering, and fluid- and heat-flow modeling in
constant during the whole test, meaning that the gas flow rate out wellbores. Besides coauthoring more than 130 papers, he
is equal to zero. The rest of the calculations remain the same. coauthored the 2002 SPE textbook Fluid Flow and Heat Trans-
fer in Wellbores and contributed to the 2009 SPE Monograph
Tony Rocha-Valadez is an engineer for ExxonMobil in Baytown, Transient Well Testing. Kabir was an SPE Distinguished Lecturer
Texas, where he joined the Planning and Improvement Group during 2006–2007, and became an SPE Distinguished Member
in 2015. Before joining industry, he worked as a research assist- in 2007. He received the 2010 SPE Reservoir Description and
ant in the Mary Kay O’Connor Process Safety Center at Texas Dynamics Award, and has served as an editor for three SPE
A&M University. Rocha-Valadez’ research interests include journals.

184 May 2016 SPE Production & Operations

ID: jaganm Time: 21:08 I Path: S:/PO##/Vol00000/160007/Comp/APPFile/SA-PO##160007

S-ar putea să vă placă și