A New Method for Leak

Detection in Gas Pipelines

Kegang Ling, University of North Dakota; Guoqing Han and Xiao Ni, China University of Petroleum;
Chunming Xu, China National Offshore Oil Corporation; and Jun He, Peng Pei, and Jun Ge, University of North Dakota

Summary lower cost. It also has the advantages of monitoring the system con-
Two types of approaches—physical inspection and mathematical- tinuously and noninterference with pipeline operations. One of the
model simulation—are used to identify a leak in a gas pipeline. The limitations of the modeling method is that it requires flow param-
former method can result in an accurate detection of the location eters, which are not always available. Leak detection from math-
and the size of the leak, but comes with the expense of production ematical modeling also has a higher uncertainty than that from
shutdown and the high cost/long time to run the physical detection, physical inspection.
which is very crucial in a long-distance gas pipeline. The latter ap- Many researchers have conducted investigations on gas transient
proach detects a gas leak by solving the governing equations, thus flow in pipelines to detect leaks. Huber (1981) used a computer-
leading to quick evaluation at much lower costs, but with higher based pipeline simulator for batch tracking, line balance, and leak
uncertainties. Our literature review indicates that a simple, prac- detection in the Cochin pipeline system. The instruments installed in
tical, and reliable method to detect a gas leak under the conditions the pipeline and the simulator in the central control office made on-
of unknown inlet or outlet gas rate, or unknown inlet or outlet pres- line, real-time surveillance of the line possible. The resulting model
sure, is highly desirable. was capable of determining pressure, temperature, density, and flow
In this study, we develop single and multiple rate test methods profiles for the line. The simulator was based on mass balance, and
to detect leaks in a gas pipeline. By conducting multiple rate tests, thus required a complete set of variables to detect the leak.
the location and size of leaks can be detected. The new method can Shell used physical methods to detect leaks in a 36-in.-diam-
be applied under the conditions of no inlet or outlet rate available or eter, 78-mile-long submarine pipeline near Bintulu, Sarawak (van
no inlet or outlet pressure available. Because these conditions are der Marel and Sluyter 1984). The leaks were detected accurately
not uncommon in gas-pipeline transportation, our method provides by optical and acoustical equipment mounted on a remotely oper-
a quick and low-computational-cost approach to detect leaks corre- ated vehicle, which was guided along the pipeline from a distance
sponding to different scenarios. of 0.5 m above the pipeline. The disadvantages of this detection
method are time consumption (15 days to finish detection), and the
Introduction pipeline needed to be kept at a high pressure to obtain a relatively
Because of its efficiency, cleanliness, and reliability, natural gas high signal/noise ratio. Sections of the pipeline were covered by a
supplies nearly one-fourth of all energy used in the United States thick layer of selected backfill. This ruled out the use of the optical
and is expected to increase by 50% within the next 20 years (An- technology. It is also noted that the maximum water depth was 230
derson and Driscoll 2000). New gas-delivery infrastructure is con- ft. Applications in a deepwater environment have not been tested.
structed to transport more natural gas to terminals far away from the Luongo (1986) studied the gas transient flow in a constant-
production site. At the same time, existing gas-delivery infrastruc- cross-section pipe. He linearized the partial-differential equation
ture is aging rapidly. Ensuring natural-gas-infrastructure reliability and developed a numerical solution to the linear parabolic partial-
is one of the critical needs for the energy sector. Therefore, the reli- differential equation. In his derivation, friction factor was calcu-
able and timely detection of leakage from a newly-built gas pipeline lated from steady-state conditions (i.e., constant friction factor for
during startup, and the failure of any part of the old pipeline, is crit- transient flow). Luongo (1986) claimed that his linearization al-
ical to the flow assurance of the natural-gas infrastructure. gorithm can save 25% in the computational time without a major
Traditionally, there are two types of approaches to detecting sacrifice in accuracy when compared with other methods. The gov-
leaks in a gas pipeline; one is physical inspection to identify the erning equations used by Luongo (1986) required a complete data
location and size of the leak, and the other is mathematical mod- set of pressure and flow rate.
eling with numerical simulation. Physical inspection consists of Massinon (1988) proposed a real-time transient hydraulic model
gas sampling; soil monitoring; flow-rate monitoring; and acoustic-, for leak detection and batch tracking on a liquid-pipeline system
optical-, and satellite-based hyperspectral imaging. Usually, the on the basis of the conservation of mass, momentum, and energy,
physical inspection can result in an accurate detection of the lo- and an equation of state. Although this model can detect leaks in a
cation and size of a leak, but this comes with the expense of pro- timely manner, it required intensive acquisition of complete data
duction shutdown and the high cost/long time to run the physical sets, both in the space domain (the pipeline lengths between sen-
detection, which is very crucial in a long-distance gas pipeline. The sors are very short) and in the time domain (time interval between
mathematical-modeling approach detects a gas leak by solving the two consecutive measurements is short), which are impossible for
governing mass-conservation, momentum-conservation, and en- many pipelines.
ergy-balance equations, thus leading to a quick evaluation at much Mactaggart (1989) applied a compensated volume-balance
method at a cost less than a transient-model-based leak detection
for sour-gas-leak detection. The method is cost effective, but is ap-
Copyright © 2015 Society of Petroleum Engineers
plied only to well-instrumented pipelines. Pressure and rate at the
This paper (SPE 1891568) was accepted for presentation at the SPE/AAPG/SEG inlet and the outlet of the pipeline are required for this analysis.
Unconventional Resources Technology Conference, Denver, 25–27 August 2014, and
revised for publication. Original manuscript received for review 10 June 2014. Revised
Scott et al. (1999) modeled the deepwater leak in a multiphase
manuscript received for review 26 December 2014. Paper peer approved 28 January 2015. production flowline. Their method can detect a multiphase leak, but

2015  ••   Oil and Gas Facilities

April 2015 97
1
needs to identify the flow regime first. The flow regime can change small leaks. The disadvantage is that a large number of sensors
along the pipeline. A multiphase leak also affects the flow regime and were installed in the line, thus increasing the cost significantly.
makes its prediction very difficult. Zhou and Adewumi (2000) in- Balda Rivas and Civan (2013) used mass-balance and transient-
cluded the kinetic-energy term in the governing equation and solved flow models to detect leaks in liquid pipelines. The response times
the partial-differential equation numerically. They formulated an ex- to the transient-flow operation were used to estimate leak location.
plicit five-point, second-order-accurate total-variation-diminishing Their model required intensive measurements of all variables.
scheme to capture the behavior of transient flow. Boundary and ini- In summary, existing methods are classified into two larger cat-
tial conditions needed to be given in the simulation. egories: physical method and mathematical model. Physical detec-
Sadovnychiy et al. (2005) discussed the development of a tion has the advantages of accuracy and high certainty. The online,
remote-detection system consisting of an infrared camera, a video real-time surveillance of pipelines and leak detection can be real-
camera, a laser spectrometer, and a global-positioning system for ized if monitoring equipment is installed in the pipeline. Because the
early detection of leaks in oil and gas pipelines. The remote system physical method requires installation and maintenance of substantial
can detect small leaks. The disadvantages include reliance of the levels of costly equipment on the pipeline, it may be excluded be-
system on multiple sensitive and delicate instruments, which are cause the high operating cost is not affordable and the long time taken
susceptible to harsh environments or severe weather; and the need to detect the leak is unacceptable because of the continuous loss of
to develop an automation system for data acquisition, transmission, revenue, damage to facilities and environment, and possible loss of
integration, and interpretation. life. Sometimes, a harsh environment or severe weather can make the
Reddy et al. (2006) built a dynamic simulation model by use installation of detection instruments in the pipeline and/or physical
of a transfer-function model for online state estimation and leak inspections impossible. In some cases, remote locations that are dif-
detection in a gas pipeline. The model reduced the computational ficult to access make physical inspection unrealistic. The mathemat-
time, while obtaining accurate state estimation from noisy mea- ical model has the advantages of low cost and quick leak detection.
surements. The computation required all available measurements Shutdown of the operation may not be required. The continuous on-
of pressure and flow rate. line, real-time monitoring of the pipeline and leak identification are
Wang and Carroll (2007) analyzed the real-time data with a tran- possible if the required data can be measured and transmitted to the
sient model to detect gas- and liquid-pipeline leakage. Stochastic central office simultaneously. The disadvantages of the mathemat-
processing and noise filtering of the meter reading were used to re- ical model are low accuracy and high uncertainty. High-quality and
duce the impact of noise. The correlations for diagnosing the leak complete data sets are key factors of detecting leaks successfully. In
location and amount are derived on the basis of the online real-time practice, the mathematical model can be used to narrow down the
observation and the readings of pressure, temperature, and flow possible leak interval before the physical inspection is conducted.
rate at both ends of the pipeline. This literature review indicated that only a few studies provided
Gajbhiye and Kam (2008) used a mechanism model to de- a practical method for detecting gas leaks in pipelines without inlet
tect leakage in a subsea pipeline under fixed pressure boundaries. or outlet flow rate or pressure. It is worthwhile to develop an ap-
The model compared the inlet and outlet flow-rate changes with proach to locate a leak and evaluate its size under these conditions
fixed-pressure boundary conditions to detect leakage. Although the because, as oil and gas exploration and production move toward
model can be applied to single- and multiphase flows, it needed offshore, deep water, polar regions, and remote/frontier locations,
pressures and flow rates at both ends of the line. it is not uncommon that metering equipment or pressure gauges
Elliott et al. (2008) showed the efficiency of leak detection by would not be installed at the inlet and/or outlet of the pipeline in
a spherical acoustic device called a SmartBall®, which has the ad- these fields. Even for onshore fields or fields with easy access, op-
vantages of low cost, ease of deployment, and the ability to locate erators may choose not to install metering equipment to cut costs.
pinhole leaks immediately to within 1 m. This technique is limited In some gas-gathering systems, metering equipment is not installed
by pipeline geometry. A long pipeline also requires long inspec- in the branches that connect to the trunk lines. The flow rates are
tion time to detect leakage. Launching and receiving SmartBalls measured only in the trunk lines. In addition, the metering equip-
are necessary, which may not be applicable in some conditions. ment and pressure gauges installed in the pipeline may be nonfunc-
Hauge et al. (2009) used an adaptive Luenberger-type esti- tioning. Although the percentage of these uncommon issues is low,
mator to locate and quantify leakage given inlet velocity, pressure, the absolute number of pipelines without inlet or outlet pressure or
and temperature and outlet velocity and pressure. The model was flow rate can be large, considering the large number of oil and gas
built in OLGA, a commercial software from Schlumberger (2014), pipelines operated in the field. Furthermore, these issues might be-
which can handle multiphase flow and incorporated temperature come common in the future when fields in harsh environments are
dynamics. Pressure and rate at two ends of the pipeline are required developed. It is also noted that leakage in offshore and deepwater
for numerical calculation. pipelines is difficult to locate and quantify. Therefore, we propose
Bustnes et al. (2011) applied a commercial real-time transient a new method to solve these issues.
model to detect leakage in a Troll field oil pipeline. The model
can accept American Standard Code for Information Interchange Model Development
(ASCII) input data and required no prior knowledge of the soft- In this work, leak detection and localization are realized by cou-
ware-calculation method. However, the accuracy of leak detection pling the gas-pipeline flow with the gas-leak flow. Multiple rate
for this method is low. Uncertainty because of transient pipeline tests are conducted to solve the governing flow equations to eval-
operation is also an issue. uate gas leak. Modeling of gas flow in pipelines and gas-leak flow
Eisler (2011) reviewed the leak-detection technologies applied are discussed in Appendices A and B, respectively.
to Artic subsea pipelines and recommend fiber-optic-cable tech-
nology for pipelines under such conditions. This method can de- Leak Detection for One Pipeline. Single and multiple rate tests
tect leaks promptly, but with a high cost for equipment, installation, are required to obtain flow parameters to solve the governing equa-
and maintenance. tions to locate the gas leak and evaluate the gas-leak rate (or leak
Vrålstad et al. (2011) compared five different leak-detection sys- size) for different scenarios. The application of multiple-rate tests
tems that are suitable for continuous monitoring of a subsea tem- to different scenarios is discussed in the following subsections.
plate and elucidated the advantages and limitations of the different Three assumptions are made in the analyses:
detection principles. Continuous monitoring by permanently in- • Single gas phase flows in the pipeline.
stalled systems, flow-measurement devices, or inspection/sur- • Temperature profile along the pipeline is known.
veying by sensors attached to mobile units were applied to detect • Gas leak occurs in only one location.

98
2 Oil and Gas Facilities  
Facilities •  
• April 2015
The gas-leak rate is the difference between the inlet and the outlet
gas rates. The location of the gas leak can be identified by di-
mensionless analysis. To develop a general solution, we introduce
three dimensionless variables: leak location, gas-leak rate, and
pressure drop.
Dimensionless leak location is defined as the ratio of distance
between leak locale and pipeline inlet to pipeline length, which is
expressed as
Lleak , ..........................................................................(1)
Lleak , D =
L
where Lleak,D is the dimensionless leak location and Lleak is the leak
location (measured from the inlet of the pipeline to the leak locale).
Dimensionless gas-leak rate is defined as the ratio of the gas-
leak rate to the gas rate at the inlet of the pipeline, which is
qleak Table 1—Input data for the leak detection in synthetic examples.
qleak , D = , ...........................................................................(2)
qinlet
3. Assuming gas leakage at different locations with different
where qleak,D is the dimensionless gas-leak rate, qleak is the gas-leak leak sizes, calculate the pressure drops that correspond to the
rate, and qinlet is the gas rate at the inlet of the pipeline. different leak locations and leak sizes. Also calculate the di-
Dimensionless pressure drop is defined as the ratio of pressure mensionless leak locations, dimensionless gas-leak rates, and
drop through the pipeline under gas-leak conditions to pressure the dimensionless pressure drops.
drop through the pipeline without leak, which is expressed as 4. Plot the dimensionless pressure-drop/gas-leak-rate/leak-lo-
cation type curves on the basis of the data gained in Steps 1,
∆pleak , ........................................................................(3) 2, and 3, as in Fig. 1.
∆pD =
∆pno leak 5. Calculate the pressure drop, dimensionless pressure drop,
gas-leak rate, and dimensionless gas-leak rate for Single Rate
where ∆pD is the dimensionless pressure drop, ∆pleak is the pressure Test 1.
drop through the pipeline under gas-leak conditions, and ∆pno leak is 6. Connect the intersection points between the dimensionless
the pressure drop through the pipeline without leak. Synthetic ex- pressure-drop plane from Step 5 and the type-curve plane ob-
amples are used to better illustrate the detection procedure. Table 1 tained in Step 4 to yield Line AB.
lists the data used for the different scenarios described in the fol- 7. Connect the intersection points between the dimensionless
lowing subsections. gas-leak-rate plane from Step 5 and the type-curve plane to
Scenario 1: One Pipeline With Known Inlet and Outlet Rates yield Line CD.
and Known Inlet and Outlet Pressures. Only one flow-rate test is 8. Project Point E, which is the intersection point of lines AB and
required to locate the gas leak. The analysis procedure is CD, onto the x–y-plane to obtain Point F. Project Point F onto
1. Run Single Rate Test 1 and record the inlet and outlet gas the x-axis to obtain the dimensionless gas-leak location, G,
rates and pressures. as shown in Fig. 1. Then, the leak location can be calculated.
2. Calculate the pressure drop in the pipeline, assuming that 9. Calculate the difference between the inlet and outlet rates to
there is no leak in the pipeline, by use of Eq. A-1. It should obtain the gas-leak size. The gas-leak-flow equations (Eqs.
be noted that the pressure drop without gas leak is the max- B-2 and B-3) can be used to verify the leak location, provided
imum compared with gas-leak cases. that the external pressure at the leak point is known.
Dimensionless Pressure Drop

1.0

0.8
C
B
0.6
E
A 0.4

0.2
0.0
0.0
0.20

F 0.50 s
D les
s ion Rate
1.00 G 0.80 en ak
0.75
0.42 Dim s-Le
G a
Dimensionless Leak Location 0.08 1.00
0.00
(Measured From the Inlet of
Pipeline)
0.0–0.2 0.2–0.4 0.4–0.6 0.6–0.8 0.8–1.0

Fig. 1—Plot of dimensionless pressure-drop/gas-leak-rate/leak-location for Inlet-Gas Rate 1.

2015  ••   Oil and Gas Facilities

April 2015 99
3
 Dimensionless Pressure Drop
1.0

0.8

0.6
B1
0.4
A1
0.2
0.0
0.0
0.20
B'1
0.50 s
D
n les te
A'1 si o a
1.00 0.80
m en ak R
0.75 i
D s- L e
0.42
0.08 1.00 Ga
Dimensionless Leak Location
0.00
(Measured From the Inlet of
Pipeline)
0.0–0.2 0.2–0.4 0.4–0.6 0.6–0.8 0.8–1.0

Fig. 2—Plot of dimensionless pressure-drop/gas-leak-rate/leak-location for Inlet-Gas Rate 1.

Scenario 2: One Pipeline With Known Inlet Rate, Known Inlet 5. Connect the intersection points between the dimensionless
and Outlet Pressures, and Unknown Outlet Rate. Two rate tests, pressure-drop plane from Step 4 and the type-curve plane ob-
or Inlet-Gas Rates 1 and 2, are needed to locate the leak. The detec- tained in Step 3 to yield Line A1B1 in Fig. 2 for Test 1 and
tion procedure is as follows: Line A2B2 in Fig. 3 for Test 2.
1. Run the two rate tests and measure the inlet-gas rates and 6. Project Lines A1B1 and A2B2 onto the x–y-plane to obtain
inlet and outlet pressures. Lines A′1B ′1 and A′2B ′2 in Figs. 2 and 3.
2. Calculate pressure drops in the pipeline, assuming that there 7. Project Point F′12, which is the intersection of lines A′1B ′1 and
is no gas leak in the pipeline. A′2B ′2, onto the x-axis to obtain the dimensionless gas-leak loca-
3. Calculate the dimensionless variables and plot the type tion, Point G′12. With that, the gas-leak location can be calculated.
curves for Tests 1 and 2, as shown in Figs. 2 and 3, 8. Calculate the outlet rate and the leak rate.
respectively. Scenario 3: One Pipeline With Known Inlet and Outlet Rates,
4. Calculate the pressure drops and dimensionless pressure Known Inlet Pressure, and Unknown Outlet Pressure. The ap-
drops for the two rate tests in Step 1. proach to Scenario 3 is similar to that for Scenario 2, but the dif-

Dimensionless Pressure Drop

1.0

0.8

0.6
B2
0.4

A2 0.2
0.0
0.0
F'12 B'2
0.20
B'1
0.50 s
D les
A'1 s ion Rate
1.00 0.80 en ak
0.75 A'2 G'12
0.42 Dim s-Le
0.08 1.00 Ga
Dimensionless Leak Location 0.00
(Measured From the Inlet of
Pipeline)
0.0–0.2 0.2–0.4 0.4–0.6 0.6–0.8 0.8–1.0

Fig. 3—Plot of dimensionless pressure-drop/gas-leak-rate/leak-location for Inlet-Gas Rate 2.

100
4 Oil and Gas Facilities  
Facilities ••   April 2015
 Dimensionless Pressure Drop
C'1
1.0
0.8
C1
0.6

0.4
D'1 0.2

0.0
0.0
0.20

0.50 ss
D1
i o nle ate
s
1.00
0.75
0.80
m en ak R
0.42 D s-Le
i
0.08 1.00 Ga
Dimensionless Leak Location 0.00
(Measured From the Inlet of
Pipeline)
0.0–0.2 0.2–0.4 0.4–0.6 0.6–0.8 0.8–1.0

Fig. 4—Plot of dimensionless pressure-drop/gas-leak-rate/leak-location for Inlet-Gas Rate 1.

ference is that dimensionless gas-leak rates are used instead of 5. If the calculated inlet pressure in Step 4 is different from that
dimensionless pressure drops (Figs. 4 and 5). in Step 2, use it as the new inlet pressure until the calculated
Scenario 4: One Pipeline With Known Inlet and Outlet Rates, inlet pressures converge. The leak location at the converged
Known Outlet Pressure, and Unknown Inlet Pressure. The steps inlet pressures is the solution.
for Scenario 4 are similar to those for Scenario 3; but, because the Scenario 5: One Pipeline with Known Outlet Rate, Known
inlet pressure is unknown, the construction of the dimensionless Inlet and Outlet Pressures, and Unknown Inlet Rate. The proce-
type curves and the detection of the leak location require an itera- dure for Scenario 5 is similar to that of Scenario 4.
tion approach. The procedure is as follows: Discussions on and Comparing Leak Detection in Scenarios 1
1. Run the two rate tests. Through 5. Leak detection in the preceding five scenarios requires
2. Calculate the inlet pressures from the outlet pressures and the accurate measurement of pressures and rates. A high-dimension-
inlet-gas rates, with the assumption that there is no gas leak less-gas-leak-rate case is easier to detect than a lower one. The min-
in the pipeline. imum leak size that can be detected is controlled by the resolutions
3. Estimate the leak location by use of the steps in Scenario 3. of instruments. If the leak is too small, the upstream and down-
4. Calculate the new inlet pressures on the basis of the esti- stream instruments cannot capture the changes and/or the noise
mated leak location. overshadows the signal. Consequently, the leak may occur without

C'2 C'1
Dimensionless Pressure Drop

1.0
C2
H'12 0.8

0.6

0.4
D'2
I'12 0.2
D'1
0.0
0.0
0.20

0.50
D2 ss
i o nle ate
s
1.00 0.80 en ak R
Dim s-Le
0.75
G'12 0.42
0.08 1.00 Ga
Dimensionless Leak Location 0.00
(Measured From the Inlet of
Pipeline)
0.0–0.2 0.2–0.4 0.4–0.6 0.6–0.8 0.8–1.0

Fig. 5—Plot of dimensionless pressure-drop/gas-leak-rate/leak-location for Inlet-Gas Rate 2.

2015  ••   Oil and Gas Facilities

April 2015 101
5
 Pipeline 1 Downstream

Pipeline 2
Upstream
Downstream

Pip
eli
ne
n

Downstream

Fig. 6—n parallel pipelines share the same junction upstream.

notice, or it would be very difficult to identify the leak and to lo- but it is unknown in which pipeline the leak is located, the analysis
cate the leak point. Generally, the confidence in the level of leak of the leak in the basic units is a critical step. Therefore, leak detec-
detection in Scenario 1 is higher than that in Scenarios 2 through tion for parallel pipelines with junctions, as shown in Figs. 6, 7, and
5. High-resolution pressure gauges and metering equipment, which 8, is useful from realistic and feasible aspects.
can provide high-quality data, are critical to accurate leak detec- Scenario 6: Parallel-Pipeline Setup, as Shown in Fig. 6. Fig.
tion, especially for Scenarios 2 through 5, in which one of the pres- 6 shows that n parallel pipelines share the same junction upstream.
sures or rates is unknown. It is more difficult to locate leaks in The leak-detection approaches for cases with different given data
Scenarios 2 through 5. To clearly identify the intersection points are described in the following.
(or leak locations) in these scenarios, the difference between Rates Scenario 6A: Known Inlet and Outlet Rates and Pressures. Mul-
1 and 2 should be as large as possible. It is noted that the selections tiple rate tests are required to identify which pipeline contains the
of Rates 1 and 2 are limited by pipeline operating specifications leak. The steps for detecting a leak for each rate test are similar to
and the sensitivities of pressure gauges and metering equipment. those in Scenario 1. It should be noted that each pipeline has its
The shapes of the type curves in Figs. 1 through 5 indicate that own type curves. The pipeline that gives the same leak location
it is easier to detect a leak if the leak occurs close to the center under different rates is the one with the leak, while the pipelines
of the pipeline. It is also clear that the leak locale can be detected that give different leak locations under different rates are excluded.
with a higher confidence as the number of flow-rate tests increases. Scenario 6B: Known Inlet Rate, Known Inlet and Outlet Pres-
Therefore, three or more rate tests instead of two rate tests can be sures, and Unknown Outlet Rate. For this scenario, multiple rate
applied to reduce the uncertainty in leak detection. The rate num- tests are required. The steps for detecting a leak are similar to those
bers required to detect leaks, as mentioned in the preceding, are the in Scenario 2. The identification of a leaking pipeline is similar to
minimum numbers required for different scenarios. that in Scenario 6A.
Leak Detection in Multiple Pipelines. Scenarios 1 through 5 are Scenario 6C: Known Inlet and Outlet Rates, Known Inlet Pres-
for single-pipeline leak detection. It should be noted that gas-trans- sure, and Unknown Outlet Pressure. The steps for detecting a leak
portation networks can be complex systems. A gas-transportation are similar to those in Scenario 3. The identification of a leaking
network can be considered as a combination of numerous single pipeline is similar to that in Scenario 6A.
pipelines and parallel pipelines connected through junctions and/ Scenario 6D: Known Inlet and Outlet Rates, Known Outlet Pres-
or nodes. Three types of parallel-pipeline setups are shown in Figs. sure, and Unknown Inlet Pressure. The steps for detecting a leak
6, 7, and 8, and are used to illustrate the applications of the pro- are similar to those in Scenario 4. The identification of a leaking
posed method to a complicated pipeline system in the field. Most pipeline is similar to that in Scenario 6A.
gas-pipeline systems can be decomposed into basic units that are Scenario 6E: Known Outlet Rate, Known Inlet and Outlet Pres-
similar to these three setups. If a leak occurs in a pipeline system, sures, and Unknown Inlet Rate. The steps for detecting a leak are

Pipeline 1 Upstream

Pipeline 2
Downstream Upstream

Pip
eli
ne
n
Upstream

Fig. 7—n parallel pipelines share the same junction downstream.

102
6 Oil and Gas Facilities  
Facilities •  
• April 2015
 Pipeline 1

Pipeline 2
Upstream Downstream

Pipeline n

Fig. 8—n parallel pipelines share the same junctions, both upstream and downstream.

similar to those in Scenario 5. The identification of a leaking pipe- Limitations of the Proposed Method and Future-Work
line is similar to that in Scenario 6A. Recommendation
Scenario 7: Parallel-Pipeline Setup, as Shown in Fig. 7. Fig. The proposed method can detect and evaluate a single leak in a
7 shows that n parallel pipelines share the same junction down- pipeline system. If the gas leak occurs in a pipeline network, the
stream. The leak-detection approaches are similar to those in Sce- network needs to be decomposed to basic units, as shown in Figs.
narios 6A through 6E. 6, 7, and 8, before application of the proposed method. However,
Scenario 8: Parallel-Pipeline Setup, as Shown in Fig. 8. Fig. pipeline networks in operation can be very complicated, and there
8 shows that n parallel pipelines share the same junctions, both can be two or more leak points in the same pipeline or different
upstream and downstream. Again, the leak-detection approaches pipelines within the systems. Future work should focus on ex-
are similar to those for Scenarios 6A through 6E. panding the application of the proposed method to more-compli-
cated scenarios, such as multiple leaks in pipeline networks, and
Identifying Multiple Leaks From a Single Leak in a Pipeline, experimental tests should be conducted to verify application of the
With Known Inlet and Outlet Pressures and Rates. Leak sce- method in such scenarios.
narios in a gas pipeline were mainly single leaks. Two leak points
in the same pipeline were observed in a few field cases. More than Conclusions
two leak points in a pipeline were observed rarely. Assuming flow- The following conclusions can be drawn from this study:
rate and pressure data can be measured, multiple rate tests can be • The proposed method provides a straightforward way to lo-
used to identify multiple leaks from a single pipeline. If multiple cate leaks and estimate leak size.
rate tests provide different leak locations, there are two or more • The new model can distinguish a single leak from multiple
leak points. If multiple rate tests result in the same leak location, leaks in a single pipeline or parallel pipelines, which is es-
the leak occurs at a single point. Identifying the leak-point number sential in selecting appropriate technologies to locate leaks
for two or more leak locations in the same pipeline or multiple quickly to minimize loss.
leaks in different pipelines connected in a system is very compli- • The new method can detect a leak without inlet or outlet flow
cated and should be the direction of future work. rate, which cannot be detected by mass-balance approaches.
• The new model can locate a leak point without inlet or outlet
Field Application pressure. Therefore, it is useful for an offshore or remote/fron-
The proposed method was used to detect a leak in an offshore gas tier pipeline in which pressure data cannot be monitored or
pipeline. A 22-in.-diameter, 157.2-km-long pipeline was used to transferred in real time.
transport gas produced from offshore fields to an onshore terminal. • We also proposed a method to locate a leak in parallel pipelines,
Inlet pressures ranged from 10 to 12 MPa during normal operation, which is critical to leak detection in a gas-pipeline system.
with gas-flow rates varied between 13 and 17 million m3/d. A leak
occurred after several years of operation. The inlet-gas-flow rate at
the offshore platform was 15.73 million m3/d and the outlet-gas-flow Nomenclature
rate at the onshore terminal was 13.92 million m3/d, which means a A = cross-sectional area of choke
flow-rate difference of 11.5% between inlet and outlet of the pipe- C = constant for unit conversion
line. The operator excluded the possibility of a false alarm, con- CD = choke-discharge coefficient
sidering the high flow-rate difference. The pipeline leak-detection Cp = fluid heat capacity at constant pressure
procedure was executed. A leak-detection method that used acoustic Cv = fluid heat capacity at constant volume
technology was selected, and leak detection through launching D = pipe diameter
acoustic pigs was executed. The actual leak detected by physical in- d1 = pipe or tank diameter
spection occurred 105.354 km away from the inlet of pipeline. The d2 = choke diameter
leak location calculated by the proposed method was 105.537 km eD = relative roughness
away, which is close to the actual leak point. This indicated that the f = friction factor
proposed method can be used to narrow the range of leak location k = Cp/Cv is the specific-heat ratio of fluid
before the confirmation by physical inspection. The difference be- L = pipe length
tween the model and physical detection may be caused by inaccu- Lleak = leak location (measured from the inlet of the pipeline to
rate measurements of temperature, pressure, and flow rate; change the leak locale)
in the pipeline inner diameter resulting from scaling, corrosion, and Lleak,D = dimensionless leak location
erosion; possible liquid condensed in the pipeline; and inaccurate MW = molecular weight
estimation of gas properties. Therefore, the proposed method can be NRe = Reynolds number
used to narrow the possible pipeline interval involving the gas leak p = gas pressure in pipe
and reduce the range that will be examined by physical inspection. pdown = downstream pressure

2015  ••   Oil and Gas Facilities

April 2015 103
7
 pinlet = inlet pressure Hauge, E., Aamo, O.M., and Godhavn, J.-M. 2009. Model-Based Monitoring
poutlet = outlet pressure and Leak Detection in Oil and Gas Pipelines. SPE Proj Fac & Const
ppr = pseudoreduced pressure 4 (3): 53–60. SPE-114218-PA. http://dx.doi.org/10.2118/114218-PA.
pup = upstream pressure Huber, D.W. 1981. Real-Time Transient Modem for Batch Tracking,
psc = standard-condition pressure Line Balance and Leak Detection. J Can Pet Technol 20 (3): 46–52.
q = gas-flow rate PETSOC-81-03-02. http://dx.doi.org/10.2118/81-03-02.
qleak,D = dimensionless gas-leak rate Jain, A.K. 1976. Accurate Explicit Equation for Friction Factor. Journal of
qleak = gas-leak rate the Hydraulics Division 102 (5): 674–677.
qinlet = gas rate at the inlet of the pipeline Luongo, C.A. 1986. An Efficient Program for Transient Flow Simulation in
T = average temperature equal to (Tinlet + Toutlet)/2 Natural Gas Pipelines. Presented at the PSIG Annual Meeting, New
Tdown = down temperature Orleans, 30–31 October. PSIG-8605.
Tpr = pseudoreduced temperature Mactaggart, R.H. 1989. A Sour Gas Leak Detection System Implementa-
Tsc = standard-condition temperature tion. Presented at the PSIG Annual Meeting, El Paso, Texas, USA,
Tup = upstream temperature 19–20 October. PSIG-8908.
u = gas-flow velocity Massinon, R.V.J. 1988. A Real Time Transient Hydraulic Model for Leak
z = gas compressibility Detection and Batch Tracking on a Liquid Pipeline System. Pre-
z = average gas compressibility equal to (zinlet + zoutlet)/2 sented at the Annual Technical Meeting, Calgary, 12–16 June.
γg = gas specific gravity PETSOC-88-39-93. http://dx.doi.org/10.2118/88-39-93.
∆pD = dimensionless pressure drop Reddy, H.P., Narasimhan, S., and Bhallamudi, S.M. 2006. Simulation and
∆pleak = pressure drop through the pipeline under gas-leak State Estimation of Transient Flow in Gas Pipeline Networks Using a
conditions Transfer Function Model. Ind. Eng. Chem. Res. 45 (11): 3853–3863.
∆pno leak = pressure drop through the pipeline without gas leak http://dx.doi.org/10.1021/ie050755k.
∆z = outlet elevation minus inlet elevation (note that ∆z is Sadovnychiy, S., Bulgakov, I., and Valadez, J. 2005. System for Remote
positive when outlet is higher than inlet) Detection of Pipeline Leakage. Presented at the SPE Latin American
ε = absolute roughness and Caribbean Petroleum Engineering Conference, Rio de Janeiro,
μ = gas viscosity 20–23 June. SPE-94958-MS. http://dx.doi.org/10.2118/94958-MS.
ρ = gas density Schlumberger. 2014. OLGA Dynamic Multiphase Flow Simulator. http://
www.software.slb.com/products/foundation/Pages/olga.aspx.
Acknowledgments Scott, S.L., Lei, L., and Jinghai, Y. 1999. Modeling the Effects of a Deep-
The authors are grateful to the Petroleum Engineering Department water Leak on Behavior of a Multiphase Production Flowline. Pre-
at the University of North Dakota. This research is supported in sented at the SPE/EPA Exploration and Production Environmental
part by the North Dakota Experimental Program to Stimulate Com- Conference, Austin, Texas, USA, 1–3 March. SPE-52760-MS. http://
petitive Research, under award number EPS-0814442. dx.doi.org/10.2118/52760-MS.
van der Marel, M. and Sluyter, E.A. 1984. Leak Detection Survey of a 36
References Inch Diameter 78 Mile Long Submarine Pipeline. Presented at the
Anderson, R. and Driscoll, D. 2000. Pathways for Enhanced Integrity, Re- SPE Offshore Southeast Asia Show, Singapore, 21–24 February. SPE-
liability and Deliverability. Report No. DOE/NETL-2000/1130, U.S. 12446-MS. http://dx.doi.org/10.2118/12446-MS.
Department of Energy: Office of Fossil Energy and the National En- Vrålstad, T., Melbye, A.G., Carlsen, I.M. et al. 2011. Comparison of Leak-
ergy Technology Laboratory, Washington, D.C. (September 2000). Detection Technologies for Continuous Monitoring of Subsea-
Balda Rivas, K.V. and Civan, F. 2013. Application of Mass Balance and Production Templates. SPE Proj Fac & Const 6 (2): 96–103.
Transient Flow Modeling for Leak Detection in Liquid Pipelines. Pre- SPE-136590-PA. http://dx.doi.org/10.2118/136590-PA.
sented at the SPE Production and Operations Symposium, Oklahoma Wang, S. and Carroll, J.J. 2007. Leak Detection for Gas and Liquid Pipe-
City, Oklahoma, USA, 23–26 March. SPE-164520-MS. http://dx.doi. lines by Online Modeling. SPE Proj Fac & Const 2 (2): 1–9. SPE-
org/10.2118/164520-MS. 104133-PA. http://dx.doi.org/10.2118/104133-PA.
Bustnes, T.E., Rousselet, M., and Berland, S. 2011. Leak Detection Per- Weymouth, T.R. 1912. Problems in Natural Gas Engineering. Trans ASME 34.
formance of a Commercial Real Time Transient Model for Troll Oil Zhou, J. and Adewumi, M.A. 2000. Simulation of transients in natural
Pipeline. Presented at the PSIG Annual Meeting, Napa Valley, Cali- gas pipelines using hybrid TVD schemes. Int J Numer Methods
fornia, USA, 24–27 May. PSIG-1114. Fluids 32 (4): 407–437. http://dx.doi.org/10.1002/(sici)1097-0363
Eisler, B. 2011. Leak Detection Systems and Challenges for Arctic (20000229)32:4<407::aid-fld945>3.0.co;2-9.
Subsea Pipelines. Presented at the OTC Arctic Technology Con-
ference, Houston, 7–9 February. OTC-22134-MS. http://dx.doi. Appendix A: Gas Flow in a Pipeline
org/10.4043/22134-MS. Gas flow in a nonhorizontal pipeline can be calculated by the Wey-
Elliott, J., Fletcher, R., and Wrigglesworth, M. 2008. Seeking the Hidden mouth (1912) equation:
Threat: Applications of a New Approach in Pipeline Leak Detection.
Presented at the Abu Dhabi International Petroleum Exhibition and
q=
3.23Tsc (p
2
inlet − e s poutlet
2
)D5
, ...................................... (A-1)
Conference, Abu Dhabi, 3–6 November. SPE-118070-MS. http:// psc f  gTzLe
dx.doi.org/10.2118/118070-MS.
Gajbhiye, R.N. and Kam, S.I. 2008. Leak Detection in Subsea Pipeline: where
A Mechanistic Modeling Approach with Fixed Pressure Boundaries.
Presented at the Offshore Technology Conference, Houston, 5–8 May. Le =
(e s
−1 L ) ..................................................................... (A-2)
OTC-19347-MS. http://dx.doi.org/10.4043/19347-MS. s
Gonzalez, M.H., Eakin, B.E., and Lee, A.L. 1970. Viscosity of Natural and
Gases: Monograph on API Research Project 65. New York: Amer-
ican Petroleum Institute. 0.0375 g z
s= , ................................................................ (A-3)
Guo, B. and Ghalambor, A. 2005. Natural Gas Engineering Handbook, first
Tz
edition. Houston: Gulf Publishing Company.
Hall, K.R. and Yarborough, L. 1973. A New Equation of State for Z-Factor where q is the gas-flow rate (scf/hr); D is the pipe diameter; e =
Calculations. Oil & Gas Journal 71 (25): 82. 2.718; Tsc is the standard-condition temperature; psc is the standard-

104 Oil8 and Gas

Oil and
Facilities  
Gas Facilities
•   April 2015
• April 2015
condition pressure; pinlet is the inlet pressure; poutlet is the outlet d2 0.3167
pressure; T is the average temperature equal to (Tinlet + Toutlet)/2; z CD = + + 0.025  log ( N Re ) − 4  , .........................(B-4)
d1  d2  0.6
is the average gas compressibility equal to (zinlet + zoutlet)/2; γg is
 d 
the gas specific gravity; L is the pipeline length; ∆z is the outlet el- 1
evation minus the inlet elevation (note that ∆z is positive when the
where d1 is the pipe diameter, d2 is the choke diameter, and NRe is
outlet is higher than the inlet); and f is the friction factor, which can
the Reynolds number. The calculations of gas z-factor, density, and
be calculated by the Jain (1976) correlation:
viscosity are given in Appendix C.
1  21.25 
= 1.14 − 2 log  eD + 0.9  , ........................................ (A-4) Appendix C: Gas z-Factor and Viscosity
f  N Re  Hall and Yarborough (1973) presented an accurate correlation to
where eD is the relative roughness, which is defined as the ratio of estimate the z-factor of natural gas. This correlation is summarized
the absolute roughness to the pipe internal diameter, as follows:

ε , .............................................................................. (A-5) Ap pr
eD = z= , .............................................................................(C-1)
D Y
and NRe is the Reynolds number, which can be expressed as a di- where Y is the reduced density to be solved from
mensionless group,
Y +Y 2 +Y3 −Y 4
Du , ........................................................................ (A-6) − Ap pr − BY 2 + CY D = 0
N Re =

(1 − Y )3
where ε is the pipe absolute roughness, u is the gas velocity, ρ is the and
gas density, and μ is the gas viscosity. 0.06125   1  ,
2

A= exp  −1.2  1 −  
Appendix B: Gas-Leak Flow T pr   T pr  

Gas leak from the pipeline can be simulated with gas flow through
a restriction, such as a nozzle or an orifice, into a lower-pressure 14.76 9.76 4.58 ,
environment. Choke performance can be used to evaluate gas flow B= − 2 + 3
T pr T pr T pr
under this condition. Gas flows through the choke can be divided
into subsonic and sonic flows, according to the flow regime. Sonic
flow is defined as the point at which the fluid-flow velocity through 90.7 242.2 42.4 ,
C= − 2 + 3
a choke or throated pipe reaches the sonic velocity in the fluid under T pr T pr T pr
the in-situ condition. In other words, the upstream cannot “feel” the
pressure wave propagated from downstream upward because the 2.82 ,
fluid is traveling in the opposite direction with the same velocity D = 2.18 +
T pr
under sonic-flow conditions. From its name, we know that subsonic
flow exists when flow velocity is less than the sound velocity in the where ppr is the pseudoreduced pressure and Tpr is the pseudo-
fluid at the in-situ condition, under which the change of downstream reduced temperature.
pressure can be “felt” by the upstream. Downstream/upstream-pres- Once gas compressibility factor is provided, gas density can be
sure ratio is used to determine the flow regime. It is expressed as calculated by
k
 pdown   2  k −1 , .........................................................(B-1) M W p . ................................................................(C-2)
  =    = 2.7
p
 up  c k + 1 28.96 zT
where pdown is the downstream pressure, pup is the upstream pres- With given z-factor and density, gas viscosity can be estimated by
sure, k = Cp/Cv is the specific-heat ratio of fluid, Cp is the fluid heat use of the correlation by Gonzalez et al. (1970):
capacity at constant pressure, and Cv is the fluid heat capacity at con-
stant volume.  = 10 −4 K exp ( X  Y ), ........................................................(C-3a)
Sonic flow occurs when the downstream/upstream-pressure ratio
is equal to or less than the critical pressure ratio. Otherwise, sub-
sonic flow occurs. The gas rate of sonic flow can be calculated by
K=
(9.379 + 0.01607 M W ) T 1.5 , ........................................(C-3b)
k +1 209.2 + 19.26 M W + T
 k   2  k −1
q = 879C D Apup    , .................................(B-2)
  gTup   k + 1   986.4 
X = 3.448 +  + 0.01009 M W , .................................(C-3c)
where A is the cross-sectional area of the choke, CD is the choke-  T 
discharge coefficient, Tup is the upstream temperature, and γg is the
gas specific gravity. and
Under subsonic-flow condition, gas rate is calculated by
Y = 2.447 − 0.2224 X , .......................................................(C-3d)
 2 k +1

 pdown   pdown   where MW is the molecular weight.
k k
2k .
q = 1, 248C D Apup  p  −  
( k − 1) gTup  pup  
 up

Kegang Ling is an assistant professor in petroleum engineering at the
.............................................................................................(B-3)
University of North Dakota. His research interests are in the area of pro-
The correlation by Guo and Ghalambor (2005) provides a feasible duction optimization. Ling holds a BS degree in geology from the China
way to estimate the choke-discharge coefficient: University of Petroleum, and MS and PhD degrees in petroleum engi-

April 2015
2015   Oil and• Gas
April••  2015 Oil and
Facilities
Gas Facilities
105
9 9
 neering from the University of Louisiana at Lafayette and Texas A&M University of Petroleum, and an MS degree in geology from the Univer-
University, respectively. sity of North Dakota.

Guoqing Han is an associate professor at China University of Petroleum, Peng Pei is a research engineer at the Institute for Energy Studies, Uni-
Beijing. He has expertise in several fields, including artificial-lift design, versity of North Dakota. His research interests include energy-related
flow assurance, and reservoir simulation, in which he has extensive ex- geomechanics and energy-conversion and -utilization processes. Pei
perience in teaching and researching. Han has published more than 20 has authored or coauthored more than 25 papers. He holds a PhD de-
technical papers. He holds BSc and MSc degrees in process automa- gree in geological engineering and an MS degree in mechanical engi-
tion and production engineering, respectively, from China University of neering, both from the University of North Dakota.
Petroleum, Shandong. Han also holds a PhD degree in petroleum engi-
neering from China University of Petroleum, Beijing. Jun Ge is a research engineer at the Energy and Environment Research
Center (EERC) at the University of North Dakota, where his work focuses on
Xiao Ni is a student in the Petroleum Engineering Department at China developing geophysical and geomechanical models of the subsurface and
University of Petroleum, Beijing. His research interests are in the areas running dynamic simulations to determine the long-term fate of produced/
of flow assurance and production optimization. Ni has published sev- injected fluids, including hydrocarbons, CO2, and brine, using oil-and-gas-
eral technical papers. industry simulation software. Before his position at the EERC, Ge served as
an intern at Golder Associates in Redmond, Washington, and as a graduate
Chunming Xu is a senior geophysicist at China National Offshore Cor- research assistant at Texas A&M University. His principal areas of interest
poration. His research interests are in the areas of reserves evaluation and expertise include reservoir geomechanics simulation, hydraulic-frac-
and reservoir characterization. Xu holds a BS degree in geophysics from ture design and propagation modeling, CO2-flooding simulation, numerical
China Petroleum University. modeling of wellbore stability, and rock mechanics applied to petroleum
and geothermal reservoir development. Ge has authored or coauthored
Jun He is a graduate student at the University of North Dakota. His re- several publications in the fields of petroleum engineering and geology. He
search interests are in the areas of reserves evaluation and reservoir holds an MS degree in petroleum engineering from Texas A&M University;
characterization. He holds a BS degree in geology from Southwest Pe- an MS degree in economic geology from Peking University, Bejing; and a
troleum University, an MS degree in petroleum engineering from China BS degree in geology from China University of Geosciences, Wuhan.

106
10 Oil and Gas Facilities  
Facilities ••   April 2015