You are on page 1of 16

PAPER

Detecting and Correcting Pipeline Leaks


Before They Become a Big Problem
AUTHORS ABSTRACT
Ron Cramer Timely pipeline leak detection is a significant business issue in view of a long his-
Shell Global Solutions US, Inc., tory of catastrophic incidents and growing intolerance for such events. It is vital to flag
Houston, Texas containment loss and location quickly, credibly, and reliably for all green or brown
David Shaw field critical lines in order to shut down the line safely and isolate the leak. Pipelines
Technical Toolboxes Consulting, are designed to transport hydrocarbons safely; however, leaks have severe safety,
LLC, Houston, Texas economic, environmental, and reputational effects. This paper will highlight robust,
reliable, and cost-effective methods, most of which leverage real-time instrumenta-
Robert Tulalian
tion, telecommunications, SCADA, DCS, and associated online leak detection appli-
Shell Nigeria
cations. The purpose of this paper will be to review the underlying leak detection
Pabs Angelo business issues, catalogue the functional challenges, and describe experiences
Shell Norway with available technologies. Internal and external techniques will be described, includ-
ing basic rate of change of flow and pressure, compensated mass balance, statistical,
Maarten van Stuijvenberg
real-time transient modeling, acoustic wave sensing, fiber optic cable (distributed
IPCOS, Luiven, Belgium
temperature, distributed acoustic sensing), and subsea hydrophones. The paper
will also describe related credibility, deployment, organizational, and maintenance
Background issues with an emphasis on upstream applications. The scope will include leak de-

T here is a very large global up- tection for pipelines conveying various flowing fluids—gas, liquid, and multiphase
stream pipeline network for the trans- flow. Pipeline environments will include subsea and onshore. Advantages, disadvan-
portation of hydrocarbon fluids. These tages, and experiences with these techniques will be described and analyzed.
include hydrocarbon liquids, gases, Keywords: pipelines, leak, fiber, SCADA
and multiphase fluids. In the authors’
opinion, pipelines offer the safest and
most efficient means of transporting very low public tolerance for hydro- ilarly, even very carefully operated and
liquids. It is estimated, for example, carbon spills. Thus, even though intrumented pipeline systems are sub-
that deaths due to accidents per pipelines provide the safest means of ject to unexpected environmental haz-
ton-mile of shipped petroleum are hazardous fluid transport, they are ards. The July 2011 Exxon Pipeline
87 times greater for trucks, 4 times also subject to unexpected leaks, and rupture, due primarily to a “thousand-
greater for ships, and 2.7 times greater the purpose of this paper is to review year flood,” fouled 70 miles of river-
for rail than for pipelines (Furchtgott- means of quickly identifying and mit- bank along the scenic Yellowstone
Roth, 2013a) The July 2013 Quebec igating the impact of such incidents. River in Montana, killing fish and
rail incident resulting in many fatali- Leaks occur on pipelines even wildlife and prompting a massive,
ties, significant property damage, and when the most careful inspection and months-long cleanup (Brown, 2013).
business loss provides further corrobo- maintenance is performed system- The underlying leak detection
ration of the relative safety of pipelines atically, notably even on brand new (LD) business requirement is therefore
(Furchtgott-Roth, 2013b). constructions. The first Keystone tar fast, reliable, and credible flagging of
Over the last 30 years, the annual sands pipeline, constructed less than a containment loss for pipelines trans-
number of pipeline leaks has decreased year previously, sprang its 12th leak porting hazardous hydrocarbons in
significantly. However, the potential for on May 29th, 2011, spilling approxi- order to initiate response actions to
leaks still presents a significant risk, and mately 2,100 gallons of raw tar sands minimize spill volume and to mitigate
since the Macondo incident, there is a crude oil in Kansas (Swift, 2011). Sim- HSSE consequences. Fast containment

January/February 2015 Volume 49 Number 1 31


loss detection is important to facilitate ample, API 1149 (1993), a highly ■ pipeline instrumentation accuracy
rapid response and hence minimize sensitive LDS, will also tend to pro- limits the LD sensitivity (for sys-
consequent hazard and damage before duce more false alarms, reducing its tems that monitor pipeline flow
it becomes a “big problem.” For ex- reliability. Similarly, a highly accurate changes, sensor repeatability is
ample, the July 2010 Kalamazoo spill system will tend to work less well in key);
leaked 840,000 gallons of oil into unexpected situations, reducing its ■ limited ability to locate leaks
the river prior to remediation, even robustness. Some of the more impor- accurately;
though leak alarms were flagged tant identified factors that affect per- ■ limited ability to handle pipeline

at the time of the incident, some 17 h formance include the following (API transients due to start-up, shut-
before (National Transportation Safety 1149, 1993): down, valve closures, intermittent
Board, 2010). ■ multiphase flow (generally, LD on flow, and slack line flow—these all
single-phase, incompressible fluids tend to cause false alarms and to
is more reliable); limit reliability;
Performance Requirements ■ instrumentation placement; ■ uncertainty due to fluid compo-

The performance of leak detection ■ instrumentation quality and sition and physical properties—
systems varies widely between tech- accuracy; especially the case for multi-phase
nologies, pipelines, and operational ■ transient operations (generally, LD fluids.
situations. Factors such as sensitivity, on steady-state systems is simpler External LDS detect leaking fluid
accuracy, reliability, robustness, and and more reliable). outside the line using direct sensing
adaptability must be considered when The API also emphasizes that LDSs (Geiger, 2006 and 2012). These LDS
selecting a leak detection system. The are complete systems that include tech- measure physical properties around
American Petroleum Institute (API) nology, processes for utilizing the tech- the pipelines using sensors on or adja-
has published detailed guidelines and nology, and people who use them. cent to the pipeline; for example, hydro-
procedures specifying how performance Performance of the entire system is carbon sensors, acoustic sensors, and
of pipeline leak detection system (LDS) therefore a combination of the perfor- fiber optic cable sensors. The general
is measured (API 1149, 1993; API mance of each component. characteristics of external LDS are
1130, 2007). Important aspects of summarized as follows:
these are as follows: ■ continuous ability to detect and lo-

■ sensitivity (minimum detectable Categories of Technologies cate small, chronic leaks accurately,
leak size and corresponding re- Internal or Computational Pipeline usually by leaking fluid diffusing to
sponse time); Monitoring (CPM) LDS (API 1130, the sensor;
■ accuracy (estimation of the true 2007) works by a combination of ■ require regular sensor maintenance,

leak flow rate and the location of monitoring pipeline variables that re- repair, or replacement;
the leak); fresh online hydraulic models and ■ difficulty in quantifying size and

■ reliability (ability of a leak detec- hence are subject to process uncer- rate of small leaks;
tion system to render accurate tainties and the limitations of pipeline ■ limited area coverage (requires many

decisions about the existence of a instrumentation. These LDSs utilize sensors to cover a large area).
leak on a pipeline during normal field SCADA or DCS sensor data to Figure 1 shows the principal current
operations); monitor pipeline parameters including internal and external LD technologies
■ robustness (ability of a leak detection pressure, temperature, density, and (Thodi et al., 2012).
system to continue to function and flow rate and thus continuously infer
provide useful information under hydrocarbon release by computation.
changing pipeline conditions). The general characteristics of internal
It is important to note that no LDS are summarized as follows: Internal LD by Simultaneous
single technology or even implemen- ■ fast detection of larger leaks (greater Pressure/Flow Monitoring
tation of a single technology will than about 5% of total line flow); of Incompressible Fluids
provide a perfect combination of all ■ inability to detect small (under 1%) When a pipeline containing in-
these performance metrics. For ex- leaks; compressible flowing fluid at greater

32 Marine Technology Society Journal


FIGURE 1 under steady-state conditions for a
given length of pipeline. This is shown
Listing of current internal and external pipeline LD technologies.
in Figure 3.
Under leak conditions, the flow
squared/pressure ratio will change sig-
nificantly. The detection of a leak,
using the flow and pressure, is there-
fore performed by a rate of change
(ROC) computation at steady state.
A filtered or time-averaged value (rep-
resenting steady nonleak conditions)
and nonfiltered flow squared/pressure
value (representing actual operating
conditions) is compared to give a
change of value (COV) indicator.
The filtering provides a built-in per-
sistence that reduces unwanted false
alarms. The COV, when it exceeds a
certain threshold, indicates a poten-
tial leak.

 2 
Q 2IN Q IN
COV ¼ 
ΔPIN ΔPIN
than ambient pressure ruptures, the and flow measurements are available,
pressure decreases. At the same time, then they can be used for LD purposes
the fluid flow into the line increases (see Figure 2). The SCADA (or DCS) system
and the flow out of the line decreases If the fluid can be assumed to be frequently (e.g., every 30 s) samples
—as shown in Figure 2. This is a char- incompressible and inviscid, Bernoulli’s the pressure and flow measurements
acteristic signature of a major line leak. theorem applies, and the flow squared/ and derives the COV. Any significant
Thus, if existing real-time pressure pressure ratio of a pipeline is constant change is flagged immediately as a
high-priority alarm at the central con-
trol room (CCR). The alarm threshold
FIGURE 2 settings are crucial (too tight), the sys-
tem generates numerous false alarms
Typical flow and pressure response to a pipeline leak.
(too wide), and the system may miss
leaks. To get the balance right, it is
recommended to build a pipeline
hydraulic model, simulate leak con-
ditions, and tune the alarm limits to
detect leak levels with negligible false
alarms. A simple traffic light display
can also be used to show the pipeline
alarm status (see Figure 4).
Advantages of this approach in-
clude the following:
■ simplicity—based on just two sen-

sor measurements and SCADA/


DCS, all of which may preexist, in

January/February 2015 Volume 49 Number 1 33


FIGURE 3 availability of the pressure and flow
measurements.
Typical pipeline monitoring parameters.

Internal LD Using
Flow Imbalance for
Incompressible Fluids
With Outlet Glow Derived
which case the LDS can be installed ■ will not identify leak location or From Pressure Drop
quickly and cost effectively; volume; The underlying principle here
■ fast indication of larger leaks; ■ steady state only; is conservation of mass—material
■ relatively sustainable, due to use of ■ incompressible liquids only; flow leaving and entering the line is
existing infrastructure that needs to ■ only a rough estimate of the true balanced by line pack, unless a leak
be maintained anyway; leaks size; occurs—which is also covered by
■ instrumentation repeatability is ■ not applicable to slack line flow; Geiger (2006).
critical rather than accuracy; ■ will not detect small or pinhole In the upstream oil and gas busi-
■ relatively easy to apply to existing leaks. ness, many lines transporting in-
brown fields pipelines—especially Note that as this method works compressible fluids have sensors
if instrumentation and SCADA only in steady state, it is necessary measuring flow and pressure upstream,
are already in place; to implement an automated means along with just a pressure measure-
■ reliability—low level of false alarms of detecting transient operations and ment downstream. A relationship can
if properly tuned. suspending LD until steady state is be derived from historic sensor data
Disadvantages of this approach achieved. This is normally achieved that facilitates estimation of the flow
include the following: by tracking parameter rate of change. out of the line during steady state
■ LD sensitivity diminishes with dis- This is by far the most common and transient conditions as illustrated
tance from point of measurement; means of pipeline LD due to wide in Figure 5.

FIGURE 4
Traffic light display showing overview LD alarm status.

34 Marine Technology Society Journal


FIGURE 5 ■ only a rough estimate of the true
leak size;
Real-time export pressure and flow measurements showing good fit to Bernoulli (dotted line) at
higher flows (steady state) and deviation from Bernoulli at lower flows due to transient operations.
■ not applicable to slack line flow;
■ will not detect small, pinhole leaks.

Internal LD by Mass
Imbalance for Subsea Well
Multiphase Fluids Using
Virtual Flow Metering
(VFM) or Multiphase
Flow Meters (MFM)
The prevailing subsea multiphase
well flowline LD standard is the pres-
sure safety low (PSL) system. On a
major line rupture, PSL is designed
to render the network safe auto-
matically by tripping the system on
low pressure. This section describes
how PSL sensitivity can be improved
LD is achieved by tracking the im- marking can be achieved by using a by combining with a real-time LDS
balance between flow in as measured commercially available transient phys- based upon flow imbalance. First, we
and flow out as estimated from this ical model to simulate the dynamics of need to consider how to continuously
relationship, once again as a COV to the pipeline system. estimate (or measure) the subsea,
minimize false alarms, Advantages of this approach in- multiphase flows (MF).
clude the following: Virtual flow metering (VFM) is a
b OUT Þ  QIN  Q
COV ¼ ðQIN  Q b OUT ■ Simplicity—based on simple sensor flow calculation based on pressure
measurements and SCADA/DCS, drop and temperature. Pressure and
where Q I N is the metered mea- all of which may preexist, in which temperature transmitters at the well-
surement and where the Q b OUT ¼ case the LDS can be installed quickly head and within the well bore and
f ðPIN  POUT Þ relationship is derived and cost-effectively; tree are used for these calculations.
from Figure 5. ■ fast indication of larger leaks; Multiphase flow meters (MFM) are
Features of this approach are sim- ■ relatively sustainable, due to use of more sophisticated instruments in-
ilar to that of inlet flow and pressure existing infrastructure that needs stalled at the wellhead or in the choke
monitoring, except for an increase in to be maintained anyway; module of the tree. Most MFMs are
LD sensitivity due to the addition of ■ relatively easy to apply to existing capable of deriving flow rates for gas,
outlet flow estimation. Sensitivity can brown field pipelines—especially water, and hydrocarbons separately.
further be increased in the (less com- if instrumentation and SCADA are Traditionally, it has proved dif-
mon) upstream situations where flow already in place; ficult to reliably measure subsea
and pressure are measured at both ■ Reliability—low level of false alarms well flow rates. However, MFMs are
ends of the pipeline. if properly tuned. being installed on newer subsea facility
Once again, it is important to tune Disadvantages of this approach well flowlines ( Jackson et al., 2012),
the COV alarm limits so that LD is include the following: primarily for allocation purposes
most sensitive and the false alarm rate ■ will not identify leak location or but with the potential to be used for
is minimized. It is also important to volume; leak detection. Where MFMs are un-
benchmark the system in terms of ■ steady state only; available, VFM can be used to derive
LD sensitivity. Tuning and bench- ■ incompressible liquids only; transient and steady-state flow rates

January/February 2015 Volume 49 Number 1 35


at subsea wells. VFMs for subsea wells are becoming progressively more common ■ All wells are closed-in → suspend
(Poulisse et al., 2006), and as MFMs continue to be relatively rare, VFM is con- leak detection.
sidered in this paper although the same LD principles apply for MFM. ■ Wells are producing and total
A typical subsea production system is sketched in Figure 6. virtual flow is stable, i.e., the
LD is achieved by tracking the imbalance between inlet flow estimated by variance of the flow is below a
VFM and flow out as measured after fluid separation at the receiving production certain threshold → perform leak
facility. Required real-time data include the following: detection.
■ pressure (P) measurements on wells and MF riser inlet to surface production ■ One or more wells are ramped
facility; down/up (if changes in valves are
■ estimated flow of each well using real-time well pressure measurements available) or closed-in (using the
(by VFM); VFM closed-in detection based on
■ the total virtual flow for each surface production system is the sum of the virtual configured criteria, e.g. flow rate
flows of all wells producing to that facility; below a threshold) → suspend
■ valve status measurements at the wells and platforms (if available); leak detection until wells are
■ flow measurements of separated liquid and vapour flows as measured at surface closed-in or in steady production
 gas 
production facility Q oil
TOTAL and Q TOTAL . and total virtual flow is stable).
With N wells producing to a production facility, the total virtual flow is com- Again it is important to tune COV
puted as follows: alarm thresholds to maximize sensi-
b oil ¼ ∑N b oil
q noil ¼ fnoil ðPn Þ
Total estimated liquid flow: Q n¼1 q n , where b tivity while minimizing false alarms
b gas ¼ ∑N b gas
q ngas ¼ fngas ðPn Þ
Total estimated gas flow: Q n¼1 q n , where b and to benchmark the system in
Flow imbalance for multiphase flow is again performed using COV: terms of LD sensitivity by hydraulic
modeling.
The physical/hydraulic model is
b oil
COVoil ¼ ðQ oil b oιl oιl
TOTAL  QTOTAL Þ  Q TOTAL  QTOTAL configured for the target upstream
b gas  Q gas Þ  Q
COVgas ¼ ðQ b gas  Q gas asset, inclusive of pump curves, piping
TOTAL TOTAL TOTAL TOTAL
configuration and profile, valves, mani-
folds and fittings, etc. The model is
Leak detection for multiphase flow can be performed only during steady state. then used offline to simulate leak tran-
Therefore, transient states should be distinguished from steady states automat- sient conditions, with leaks as “pseudo
ically. For example: valves” at given positions in the lines
as shown in Figure 6. The model out-
put is time series pressure profiles that
FIGURE 6 are input to the LDS as a real-time data
stream to help tune threshold LD sen-
Typical subsea production system showing data acquisition for real-time VFM estimation and LD, sitivities and alarm settings.
as well as simulated leak locations at various positions to benchmark and calibrate LD alarm limits.
Some general features of this
approach include the following:
■ fast indication of large leaks;

■ relatively easy to apply to existing

brown fields pipelines, especially if


well flowline VFM is already in
place;
■ will not identify leak location or

volume—steady state only.


Note that multiphase flow mea-
surement can substitute for VFM
in above LD calculation if a reliable
MFM signal is available.

36 Marine Technology Society Journal


Advantages of this approach in- using the sequential probability ratio corrected flow difference (including
clude the following: test (SPRT). line pack) to calculate a leak probabil-
■ simplicity—based on simple sensor Some general features of this ap- ity. To minimize false alarms, pattern
measurements, virtual flow estima- proach are as follows: recognition techniques are used to
tion, and SCADA/DCS, all of ■ improves LD sensitivity over and identify changes in the relationship
which may preexist, in which case above basic pressure/flow or im- between the pressure and flow when
the LDS can be installed quickly balance techniques; a leak occurs.
and cost-effectively; ■ requires setup and expertise, but no The system was tested using a sim-
■ fast indication of larger leaks; modeling; ulation of the LDS that generates time
■ relatively sustainable, due to use of ■ not directly applicable to leak lo- series pipeline data before, during, and
existing infrastructure that needs to cation or volume. after a simulated leak. The inlet and
be maintained anyway; Further detail of this methodology outlet flows, pressures, and tempera-
■ relatively easy to apply to existing is provided by Zhang (1998). tures logged from the site were im-
brown fields pipelines—especially This LD methodology is widely ap- ported into the model as time profiles
if instrumentation and SCADA plied to pipelines that are equipped and then assigned as constraints.
are already in place; with the required pipeline pressure, Model configuration data included
■ reliability—low level of false alarms flow-in, and flow-out measurements. pipeline elevation profile, ambient
if properly tuned. An extreme case of application of this temperature profile, gas composition,
Disadvantages of this approach LD technique is a 500 km, 24-inch di- pipe diameters and heat transfer coeffi-
include the following: ameter dry gas export pipeline shown cients. An example extract of the corre-
■ will not identify leak location or in Figure 7 and described below. lation between the simulator and the
volume; The LDS acquires flow and pres- actual site data is shown in Figure 8.
■ steady state only; sure data from the SCADA system The LD simulator was validated
■ incompressible liquids only; every 5 s. After automatic data cleans- by comparing results against a stan-
■ only a rough estimate of the true ing and filtering, it applies the sequen- dard pipeline hydraulic modeling tool
leak size; tial probability ratio test (SPRT) to the (SHM) for a simulated full bore
■ not applicable to slack line flow;

■ will not detect small, pinhole


leaks. FIGURE 7
Statistical flow imbalance LDS as applied to a long dry gas export pipeline.

Internal LD by Statistical
Analysis of Measured Flow
Imbalance and a Long Dry
Gas Pipeline Example
This technology detects leaks by
performing a statistical analysis of
real-time pipeline line pack derived
from flow imbalance calculations
using the flow and pressure measure-
ments along the pipeline.
Given an imbalance R(t), the statis-
tical approach asks the question: Is the
imbalance at this time t likely to be on
average of the old value μ or has it in-
creased to μ + Δμ? This is a statistical
hypothesis question and is approached

January/February 2015 Volume 49 Number 1 37


FIGURE 8 ■ some leak detection capability dur-
ing pipeline transients;
The LD simulation validated by comparing the predicted time series data with actual logged data
■ reliability—low level of false alarms
for the 502-km gas export pipeline.
if properly tuned.
Disadvantages of this approach in-
clude the following:
■ requires accurate in and out flow

and pressure measurement to effec-


tively estimate line pack;
■ some extra cost associated with
license fees and need for ongoing
vendor support;
■ statistical algorithms need to be

tuned and sustained;


■ usually requires a dedicated leak

detection server;
■ only a rough estimate of the true

leaks size;
■ not applicable to slack line flow;
rupture. An example extract (Figure 9) minimum detection size is likely ■ will not detect small, pinhole
shows that both models produce similar within that range. leaks.
results for the flow rate, with an ex- Advantages of this approach in-
tremely high initial leak rate followed clude the following:
by a rapid decay that becomes much ■ based on simple sensor mea-
steadier after 1 h. surements and SCADA/DCS, all Internal LD Case Study—
A number of rupture conditions of which may preexist, in which North Sea Deep Water
were then simulated to determine the case the LDS can be installed and Long Subsea Wet
LDS response time as shown in Table 1. quickly; Gas Flow Pipeline
The LDS gave consistent results ■ fast indication of larger- and North Sea gas field production is
across all the test cases including the medium-sized leaks; achieved by subsea (1,100 m deep)
cases using site data. The LDS demon- ■ relatively easy to apply to existing
wells, commingling into two 110-km,
strated that the minimum detection brown field pipelines—especially 30-inch production pipelines trans-
size was a 0.5-inch hole—0.25 inch if instrumentation and SCADA porting the multiphase gas condensate
was not detected; therefore, the actual are already in place; to large slug catchers at the onshore pro-
cessing plant as shown in Figure 10.
Since the flowing fluid is domi-
FIGURE 9
nated by gas and gas flow measure-
Dynamic response to rupture from LD simulation and hydraulic simulation validation indicating ment subsea is very accurate, only the
good fit. gas rate is used in the flow imbalance
calculation. In the event of a leak in
the pipeline, the gas will escape to the
sea and the onshore gas rate will fall.
There is a regulatory requirement
(DNV-RP-F302, 2010) for leak detec-
tion and location for both the mono-
ethylene glycol (MEG) distribution
lines and the multiphase pipeline.
Consequently an LDS has been installed

38 Marine Technology Society Journal


TABLE 1
Simulated LD results for line ruptures of different sizes at different locations.

Case 6 Case 1 Site Data June Data


Detection Detection Detection Detection
Leak Time Location Time Location Time Location Time Location
1″ KP276 4h 268.6 km 3 h 46 m 266 km 3 h 38 386 km 4 h 2m 302 km
1″ KP470 1 h 56 m 335.9 km 1 h 53m 335 km 2 h 11 m 497 km 2h3m 367 km
1″ Riser 2 h 19 m 0 km 3h7m 0 km 2h3m 0 km 1 h 34 m 0–200 km
4″ KP276 35 m 394.3 km 29 m 19 s 476 km 36 m 39 s 492–368 km 44 m 444–334 km
4″ KP470 7 m 19 s 504 km 6 m 59 s 504 km 7 m 19 s 504 km 8 m 40 s 504 km
4″ Riser 2 m 39 s 0 km 2 m 59 s 0 km 2 m 39 s 0 km 2 m 40 s 0 km
FB KP276 22 m 19 s 504 km 20 m 59 s 504 km 21 m 19 s 504 km 20 m 40 s 482–353 km
FB KP470 2 m 59 s 504 km 2 m 59 s 504 km 2 m 39 s 504 km 3m 504 km
FB Riser 2 m 39 s 0 km 2 m 39 s 0 km 4 m 39 s 0 km 4 m 40 s 0 km
.5″ KP276 19 h 40 m 243 km 21 h 19 m 243.1 km 18 h 16 m 0–403 km 7 h 26 m 229 km
.5″ KP470 14 h 21 m 260 km 11 h 55 m 285.5 km 16 h 15 m 316 km 6 h 53 m 230 km
.5″ Riser 16 h 7 m 169 km 28 h 45 m 224.7 km 18 38 m 0–504 km 7h7m 280 km

that in the event of a leak generates detection system is based on compar- infield flowlines) starts at the wellhead
CCR alarms showing leak rate, leak ison of measurements and calculated wet gas flow meters and ends at the
percentage and location, plotted on values. Leak detection for the produc- onshore export gas metering station.
a pipeline profile and map. The leak tion system (production pipelines and Flow rate and pressure measurements
are compared to calculated values
FIGURE 10 from the leak detection system. The
deviations between measured and cal-
LDS for the production system (production pipelines and infield flowlines) from wellhead wet gas
flow meters to the onshore export gas metering station. culated values are detected and used
to identify and report the leak rate
and location.
In order to reduce number of false
alarms, the discrepancies from the
mass balance estimator are analyzed
over three different time spans, and
the average over each time span is con-
tinually calculated. Each average from
the three different time spans is com-
pared against three different leak alarm
thresholds. The shortest time span has
a high leak rate threshold to be able to
detect large leaks in a short time; the
longest time span has a low leak rate
threshold to be able to detect small
leaks after a longer time. If the leak de-
tection has been disabled or turned off,
the filters will clear the stored values

January/February 2015 Volume 49 Number 1 39


and will not start to store measured data running in real time. Pipeline pressure ■ leak detection during periods of
until LD is re-enabled and turned on. and flow profiles are calculated based pipeline transient operation.
Additionally, the LDS is based on a on line inlet and outlet measurements. Disadvantages of this approach in-
normal line configuration both subsea The actual, measured, and calculated clude the following:
and onshore. In the event of an alter- profiles are compared, and if there is ■ expensive, complex—numerous
ation in the normal line-up, the LDS a discrepancy, the location and size instruments, extensive controller
will turn off some of the functionality of the leak are computed. The model training, and system maintenance;
or will delay the report of an alarm. An can be tuned to distinguish between ■ simulated transients can be dif-

alteration of the line-up could be caused instrument errors, normal transients, ficult to verify, e.g., water cut
by an equipment maintenance plan. and leaks. Commercially available changes, GOR changes during
An offline simulation exercise was RTTM packages can simulate pipeline well start-up as these values are not
used to benchmark wet gas flowline flow conditions using advanced fluid measured—consequently, there
LDS capabilities as shown in Table 2. mechanics, hydraulic modeling, con- may be tuning difficulties and it
An offline simulation exercise was servation of momentum calculations, may be hard to sustain the model
used to benchmark the MEG pipeline conservation of energy calculations, over time;
LDS capabilities as shown in Table 3. and numerous flow equations. ■ difficult to keep the model well

Further detail of this application Advantages of this approach in- enough tuned to eliminate false
is given by Lunde et al. (2009) and clude the following: alarms;
Angelo et al. (2013). ■ increased LD sensitivity—models ■ rarely applicable for multi-
dynamic fluid characteristics (flow, phase flow—real-time model-
pressure, temperature), inclusive of ing of multiphase flow is very
Internal LD Using pipeline length, diameter, thickness, complex;
Real-Time Transient as well as product characteristics ■ extra expense associated with
Modeling (RTTM) (density, viscosity, etc.); software license fees, complex
RTTM consists of a complete de- ■ model can be tuned to distinguish model maintenance, dedicated IT
tailed transient and steady-state physical/ between instrument errors, normal servers, and accurate in and out
hydraulic model of the pipeline system transient, and leaks; flow and pressure measurement;

TABLE 2
LDS benchmark for 30″ 110 km wet gas flowline.

Leak Size (Shell DEP) Flow, MSm3/d Detection Time, s LDS Gas, vol% Company Specification Gas, vol%
Small leak Not applicable 1–5
Medium leak 5.25 14,400 13 5–25
Large leak 10.5 7,200 26 25–50
Major leak 17.5 3,600 44 50–100

TABLE 3
LDS benchmark for MEG lines.

Leak Size Flow, kg/h Detection Time, s LDS Flow, Mass % Company Specification, Mass %
Small leak 1,667 7,200 5 1–5
Medium leak 3,355 1,800 10 5–25
Large leak 6,711 600 20 25–50
Major leak 50–100

40 Marine Technology Society Journal


■ will not detect small, pinhole FIGURE 11
leaks.
Typical acoustic wave installation showing pipeline pressure gauges and high-speed telemetry to
Further detail regarding this ap- transmit signals back to host server.
proach to leak detection can be found
in API 1130 (2007).

External LD Using
Acoustic Pressure Wave
(Also Known as Negative
Pressure Wave)
This LD technology is designed to
detect the (acoustic) pressure wave
(PW) that is generated at the onset of
a leak. The wave travels in all directions
from the point of the leak at the speed
of sound in the particular fluid. Pres-
sure sensors installed on the pipeline
detect/report the acoustic wave as it
travels past the point of measurement. ■ may be prone to false alarms due to be more difficult to cost-justify for
Each sensor utilizes GPS technology spurious sound effects like mech- brown fields operations.
to accurately time-stamp the sensor anical noise, ocean currents; also, Further detail regarding this ap-
response. By comparing wave arrival small variations in pressure can proach to leak detection can be found
times at a sensor upstream of a leak lead to false alarms; in Scott and Barrufet (2003).
and downstream of the leak, the speed ■ inappropriate for facilities that
of sound in the pipeline medium interfere with sound transfer, like
can be used to automatically compute complex pipeline networks, pigging External LD Using
the location of the leak between the operations; Fiber Optic Cable (FOC)
sensors. A typical PW installation is ■ it is only able to detect leaks Distributed Sensing Systems
shown in Figure 11. reliably in relatively steady state Fiber-optic leak detection systems
Advantages of this approach in- conditions; utilize “standard,” communications
clude the following: ■ mainly for liquid pipelines, as pres- grade, fiber-optic cables installed
■ fast leak detection (speed of sound sure waves are quickly attenuated along the outside of the pipeline. The
in the medium); in gas pipelines; systems operate based on the principle
■ fairly precise indication of leak ■ relatively little evidence of success- that light transmission through the
location. ful, sustained field operation; fiber-optic cable will be altered by the
Disadvantages of this approach in- ■ will not detect small, pinhole presence of leaked product absorbed
clude the following: leaks. into the cable coating. The coatings
■ sensitive only to the singular sound The main concerns with this on the fiber-optic cable react with
wave produced by the breach— methodology are its relative imma- leaked hydrocarbons to change the
if sensor resolution and telemetry turity, tendency to false alarms and refractive properties of a segment of
speed is not fast enough, the “one- problems anticipated with tuning/ the cable. High-frequency pulsed
shot” signal could be missed; sustaining in a pipeline environment laser is used to locate the position of
■ needs accurate speed of sound that is subject to changing fluid com- this change in refractive index and
m ea s ur e m e n t i n th e fl o w i n g positions, pigging, and transients. has potential to detect small leaks (see
medium—this could be a problem May be more appropriate for green Figure 12). FOS can be used effectively
for multiphase flow; fields operations as pressure measure- on liquids, multiphase fluids, and nat-
■ no indication of leak volume; ment and high speed telemetry may ural gas systems.

January/February 2015 Volume 49 Number 1 41


FIGURE 12 ■ Rayleigh banding is sensitive to
the acoustic signature generated
Typical FOS installation.
by leaking fluid for distributed
acoustic sensing (DAS). It acts as
a pseudohydrophone without
the need to contact leaking fluids.
Advantages of this approach in-
clude the following:
■ potential for fast small leak detec-

tion and location estimation with


high sensitivity;
■ potential for continuous LD for

longer pipelines, e.g., Arctic,


subsea;
■ minimal external power require-

ments and immune to electro-


magnetic interference;
■ LD for all pipeline conditions
(start-up, shutdown, valve closure,
transients) and for all pipeline
Anomalies along the fiber are de- ■ Raman/Brillouin band scatter- fluids (single phase oil, gas, and
tected by the changes in the properties ing is used for distributed tem- multiphase);
■ complex electronics not exposed
of scattered light backscattered to the perature sensing (DTS) to detect
source. Different light scattering pro- leakage, which causes significant to operational environment—
cesses respond to different physical temperature change, as in gas located in protected, air con-
property changes at differing locations pipelines where Joule-Thomson ditioned location at one end of
down the fiber-run. Two of these are sig- effects may result in significant the pipeline; intrinsically safe
nificant for pipeline LD (see Figure 13). cooling. installation;
Disadvantages of this approach in-
clude the following:
FIGURE 13 ■ no indication of leak volume;

FOC optical scattering patterns. ■ fiber installation is an issue—e.g.,

offshore may need to pass over lay


vessel rollers/trenching equipment,
onshore need to assess impact of
trench and soil conditions;
■ s u s t a i n a b i l i t y u n c l e a r — very
small operational installed pipe-
line LDSs, hence uncertain mini-
mum LD thresholds of detection
and false alarm issues, especially
subsea;
■ fiber not so good for complex pipe-

line networks.
Further detail regarding this ap-
proach to leak detection can be found
in Minto (2013), Lowell (2013), and
Dutoit (2013).

42 Marine Technology Society Journal


External Subsea LD for FIGURE 14
Short Pipeline Lengths, Typical subsea manifold acoustic sensor LD installation.
e.g., Manifolds, Well
Heads Using Passive
Acoustic “Hydrophone”
The principle of operation of
all these systems is that any leak
causes a sound that lies within a spe-
cific frequency range depending
on the leaking fluid. The sound in-
tensity of the leak is mostly depen-
dent on the pressure difference over
the leak orifice and the distance from
the leakage point to the acoustic sen-
sor. LD acoustic sensors incorporate
filtering to remove all sound that
does not fall within the frequency
range and filter out all random bust and maintenance/calibration during an ROV inspection. The
(white) noise. The performance of free; thrusters in the ROV generated
these systems depends critically upon ■ needs to be resistant to biofouling; noise and was interpreted by the
the signal processing algorithms that ■ uncertain minimum LD thresh- system as a leak.
reject extraneous noise, identify very olds; Further detail regarding this ap-
faint leak sounds, and locate the ■ tends to be very expensive com- proach to leak detection can be found
leak. The ALD system contains three pared to real-time computer-based in Smith (2013).
hydrophones per detector, which gives LDS;
a possibility to detect the leakage point ■ false alarms can be an issue—one
by triangulation. operator experienced problems Combined External
Acoustic systems can be used effec- and Internal LDSs
tively on both liquids and natural gas For Greenfield assets, it is best to
systems. FIGURE 15 include multiple independent means
Figures 14 and 15 illustrate subsea of detecting leaks, including both in-
Acoustic sensor LD subsea assembly.
sensor installations. ternal and external LDS. This is both
Advantages of this approach in- an accepted engineering best practice
clude the following: (Geiger, 2006, 2012; DNV-RP-
■ rapid leak detection at leak source; F302, 2010), and in some situations
■ detects liquid and gas leaks; it is part of the regulations (TRFL,
■ potential for relatively large subsea 2003). This strategy can be very robust
detection area (ca. 100 m). in detecting small to big leaks, from
Disadvantages of this approach in- focused equipment to very long pipe-
clude the following: lines. In the event of a leak, the com-
■ subsea applications only—usually plete system would generate the same
only in a given, higher risk area, alarm, which would mean that the
e.g., well head, manifold, riser; leak alarm is more accurate and reli-
■ unsuitable for long pipeline leak able. For example, a medium leak in
detection; a tree would be detected both by
■ inaccessible for routine mainte- acoustic leak detection and flow im-
nance, hence needs to be highly ro- balance LDS (Figure 16).

January/February 2015 Volume 49 Number 1 43


FIGURE 16 Findings Synopsis
Multiple leak detection systems installed in a subsea asset. The following table attempts to
summarize the findings of this paper
and perhaps provide some indications
of LD system applicability with the
following cautions (Table 4):
■ LD systems rarely are “one size fits

all” and will vary with individual


pipeline circumstance.
■ LD systems considered are
all “continuous” and hence ap-
plicable to continuous pipeline
operations with a view to fast leak
detection.
■ The list of LD systems considered

is not comprehensive and tends


to reflect the author’s experiences
and opinions.

TABLE 4
Overall evaluation of leak detection systems considered.

Continuous Leak Detection Type


Pressure/ Flow Acoustic Passive
Evaluation Criteria Flow Imbalance VFM Statistical RTTM Wave Fiber Acoustic
Cost Low Low Low Med High Med High Very high
Maturity and Very high Very high Low High Low Very low Very low Very low
deployment
Rapid deployment Yes Yes Yes Some No No No No
capability
Sustainability High High High Med Low No data No data No data
Reliability (minimal High High No data Med Low Low No data Med
false alarms)
Leak size >10% >5% >15% <5% >1% No data >1% No data
Leak detection time Minutes Minutes Minutes Minutes Minutes Seconds Seconds Seconds
Leak location No No No Some Some Good Good No
Leak volume No No No Some Some No No No
Transient LD No No No Some Yes No data Yes No data
Incompressible Yes Yes Yes Yes Yes Yes Yes Yes
Fluid LD
Gas LD Some Some Some Yes Yes Some Yes Yes
Multiphase LD No No Yes No No No Yes Yes
Dedicated hardware No No No Yes Yes Yes Yes Yes
needed

44 Marine Technology Society Journal


Conclusions for total clarity on the underlying LD GOR—Gas-oil ratio
There is a wide scope to apply inter- business requirements, inclusive of LD GUI—Graphical user interface
nal LDS technology to older, brown requirements for sensitivity, false HSSE—Health, safety, security, and
field subsea and surface pipelines, alarms, operator training, and a cost/ environment
which may be more prone to rupture. benefit analysis to ensure that user JT—Joule-Thompson
In upstream operations, we have a and asset expectations are met and KPI—Key performance indicator
plethora of pipelines transporting multi- managed effectively. LD—Leak detection
phase fluids—precious few of these Setting and maintaining realistic LDS—Leak detection system
lines have LDS mainly because there alarm limits—that maximize sensitiv- MF—Multiphase flow
is no flow measurement. However, ity, while minimizing false alarms— MFM—Multiphase flow meter
the increasing use of VFM for sur- is critical to the long-term value of an P—Pressure
veillance and allocation purposes raises LDS; not so wide that leaks are missed PW—Pressure wave
the opportunity of extensions to LD. and not so narrow such that the oper- PMS—Pipeline management system
If fast rupture detection is required, ator is plagued with false alarms. The PSL—Pressure safety low
simple ROC and/or mass balance may use of an offline hydraulic model to Q—Flow
suffice. In many cases, utilization of ex- simulate leak conditions and thus RTTM—Real-time transient model
isting SCADA and flow pressure in- tune workable alarm limits is recom- SC—Slug catcher
strumentation lends itself to fast and mended. It is important to re-tune SCADA—Supervisory control and
cost effective deployment. The tech- alarm limits periodically to reflect data acquisition system
nology is for the most part mature changing pipeline conditions such as SHM—Standard hydraulic modeling
and well proven. The objective of an increasing water cut, GOR, etc. LDS tool
internal LDS is to spot larger leaks, performance will be different for each SPRT—Sequential probability ratio
quickly—not for detecting small leaks. pipeline and, with most upstream ap- test
External LDSs are generally more plications, will change with time. TCI—Technical custodian
appropriate to green field subsea and It is particularly advantageous to THP—Well head tubing head
surface pipelines, which ironically, have multiple independent and cor- pressure
being “new,” are mostly less prone to roborative leak detection systems, par- VFM—Virtual flow measurement
aging phenomena that can result in ticularly for long subsea tiebacks, to WAN—Wide area network
leaks through, for example, corrosion. maximize the robustness and reliability
The technology is for the most part less of any leak alarms.
mature and continually evolving, with References
Angelo, P., Rudrum, G., Holmås, K., Gahr
as yet fewer operational installations.
Lunde, G., & Setyadi, G. 2013. Challenges of
The promise of external systems is Glossary Flow Assurance System in Ormen Lange.
that they are more suitable for multi- ALVD—Acoustic leak and vibration Presentated at 16th International Conference
phase flow LD and the detection of detector on Multiphase Production Technology,
smaller leaks. COV—Change of value (instanta- Cannes, France, June 12.
A major key to the successful im- neous difference between mean and
API 1149. 1993. Pipeline Variable Uncer-
plementation of internal and external latest value)
tainties and Their Effects on Leak Detectability.
LDS is effective use and ongoing CCR—Central control room
www.utsi.com/uploads/kcfinder/files/Nagala_
system maintenance. For this, these CPM—Computational pipeline
Verret_2002.pdf
systems need to be part of the main- monitoring
stream operations—quite the anti- DAS—Distributed acoustic sensing API 1130. 2007. Computational Pipeline
thesis of the pipeline disappearing DCS—Distributed control system Monitoring for Liquid Pipelines. 2nd Edition
overboard and becoming out of sight, DP—Pressure drop (November, 2002). American Petroleum
Institute.
out of mind! DTS—Distributed temperature
Every LD application is different— sensing Brown, M. 2013. Yellowstone River Oil
one size does not fit all. There is a need FOS—Fiber optic systems Spill: Delayed Response Made Matters

January/February 2015 Volume 49 Number 1 45


Worse. Huffington Post, 1/2/13. Available at MTS Leak Detection Symposium, Houston,
at http://www.huffingtonpost.com/2013/ TX, 11/19/13.
01/02/yellowstone-river-oil-spill-response_n_
Minto, C. 2013. Innovative DAS Based
2397209.html.
Multimode Leak Detection System. Presented
DNV-RP-F302. 2010. Selection and Use at MTS Leak Detection Symposium, Houston,
of Subsea Leak Detection Systems. https:// TX, 11/19/13.
exchange.dnv.com/publishing/codes/download.
National Transportation Safety Board.
asp?url.../rp-f302
2010. Pipeline Accident Report. 7/25/10.
Dutoit, D. 2013. Fiber Optic Monitoring of http://www.epa.gov/enbridgespill/ar/
Subsea Flow Lines. Presented at MTS Leak enbridge-AR-0003.pdf
Detection Symposium, Houston, TX, 11/19/13.
Poulisse, H., van Overschee, P., Briers, J., &
Furchtgott-Roth, D. 2013a. Pipelines Are Goh, K.C. 2006. Continuous Well Pro-
Safest For Transportation of Oil and Gas. duction Flow Monitoring and Surveillance.
Manhattan Institute for Policy Research, Issue Presented at Intelligent Energy Conference,
Brief Number 23, http://www.manhattan- Amsterdam, The Netherlands, April 2006.
institute.org/html/ib_23.htm http://dx.doi.org/10.2118/99963-MS.

Furchtgott-Roth, D. 2013b. Quebec tragedy Scott, S., & Barrufet, M. 2003. Worldwide
reminds us pipelines are safest way to trans- Assessment of Industry Leak Detection Ca-
port oil. Toronto Globe and Mail, 7/10/13. pabilities for Single & Multiphase Pipelines.
http://www.theglobeandmail.com/globe-debate/ 2003 MMS Report, OTRC Library Number:
quebec-tragedy-shows-that-pipelines-are-safest- 8/03A120.
way-to-transport-oil/article13054202/
Smith, S. 2013. Subsea Leak Detection. Pre-
Geiger, G. 2006. State-of-the-Art in Leak sented at MTS Leak Detection Symposium,
Detection and Localization. Pipeline Simula- Houston, TX, 11/19/13.
tion Interest Group Paper 0301. University of
Swift, A. 2011. The first Keystone tar sands
Applied Sciences, Gelsenkirchen, Germany,
pipeline spills again—Providing twelve reasons
http://www.pipeline-conference.com/sites/
not to fast-track the Keystone XL tar sands
default/files/papers/321%20Geiger.pdf
pipeline. June 1 2011. Available at http://
Geiger, G. 2012. Principles of Leak Detec- switchboard.nrdc.org.
tion. Westphalia Energy Institute Report.
Thodi, P., Paulin, M., DeGeer, D., &
http://krohne.com/fileadmin/files-2/PipePatrol/
Lanan, G. 2012. Offshore Pipeline LD Tech-
KROHNE_Gerhard_Geiger_Principles_of_
nologies. Presented at SALDS Leak Detection
Leak_Detection_2012.pdf
Conference, Houston, TX, November, 2012.
Jackson, M., Napalowski, R., Paris, N., &
TRFL. 2003. Technische Regel für
Moksnes, P. 2012. Operational experience
Fernleitungen (Technical Rule for Pipelines).
with subsea multiphase flow meters. Presented
at Asia Oil and Gas Conference, Perth, Aus- Zhang, J. 1998. Implementing a Reliable
tralia, SPE 158518, 10/22/12. Leak Detection Systemon a Crude Oil
Pipeline. Presented at Advances in Pipeline
Lunde, G., Vannes, K., McClimans, O.,
Technology Conference, Dubai, Advances
Burns, C., & Wittmeyer, K. 2009. Advanced
in Pipeline Technology Conference, Dubai.
Flow Assurance System for the Ormen Lange
Subsea Gas Development. Presented at Off-
shore Technology Conference, Houston, May
2009. http://dx.doi.org/10.4043/20084-MS.

Lowell, M. 2013. Distributed Fiber Optic


Sensing and Monitoring in O&G. Presented

46 Marine Technology Society Journal

You might also like