Page 184 - Pipeline Risk Management Manual Ideas, Techniques, and Resources

Adjustments to scores 7/161
detection. These methods may have sensitivity problems The method ofleak detection chosen depends on a variety of
because they must not give leak indications in cases where nor- factors including the type of product, flow rates, pressures, the
mal pipeline transients (unsteady flows or pressures, some- amount of instrumentation available, the instrumentation char-
times temporary as the system stabilizes after some change is acteristics, the communications network, the topography. the
made) are causing pressure swings and flow rate changes. soil type, and economics. As previously mentioned when
Generally the operator must work around the inevitable trade- highly sophisticated instruments are required a trade-off often
off between many false alarms and low sensitivity to actual takes place between the sensitivity and the number of false
leaks. Because pipeline leaks are, fortunately, rare-occurrence alarms, especially in “noisy” systems with high levels of
events, the latter is often chosen. transients.
SCADA-based capabilities are commonly enhanced by
computational techniques that use SCADA data in conjunction Evaluation of leak detection cupubi1itfe.v
with mathematical algorithms to analyze pipeline flows and
pressures on a real-time basis. Some use only relatively simple The evaluator should assess the nature of leak detection abili-
mass-balance calculations. perhaps with corrections for line- ties in the pipeline section he is evaluating. The assessment
fill. More robust versions add conservation of momentum cal- should include
culations. conservation of energy calculations, fluid properties,
instrument performance, and a host of sophisticated equations What size leak can be reliably detected
to characterize flows, including transient flow analyses. The How long before a leak is positively detected
nature of the operations will impact leak detection capabilities, How accurately can the leak location be determined.
with more less steady flows and more compressible fluids
reducing the capabilities. Note that many leak detection systems perform best for only
The more instruments (and the more optimized the instru- a certain range of leak sizes, However. many overlapping
ment locations) that are accurately transmitting data into the leak detection capabilities are often present in a pipeline. A leak
SCADA-based leak detection model, the higher the accuracy detection capability can be defined as the relationship between
of the model and the confidence level of leak indications. leak rate and time to detect. This relationship encompasses both
Ideally, the model would receive data on flows, temperatures, volume-dependent and leak-rate-dependent scenarios. The for-
pressures, densities, viscosities, etc., along the entire pipeline mer is the dominant consideration as product containment size
length. By tuning the computer model to simulate mathemati- increases (larger diameter pipe at higher pressures). but the
cally all flowing conditions along the entire pipeline and latter becomes dominant as smaller leaks continue for long
then continuously comparing this simulation to actual data, the periods.
model tries to distinguish between instrument errors, normal As shown in Figure 7.7. this relationship can be displayed as
transients, and leaks. Reportedly, depending on the system, rel- a curve with axes of “Time to Detect Leak’ versus “Leak Size.”
atively small leaks can often be accurately located in a timely The areannder such a curve represents the worst case spill vol-
fashion. How small a leak and how swift a detection is specific ume, prior to detection. The shape of this curve is logically
to the situation, given the large numbers of variables to con- asymptotic to each axis because some leak rate level is never
sider. References 131 and 141 discuss these leak detection sys- detectable and an instant release of large volumes approciches
tems and methodologies for evaluating their capabilities. an infinite leak rate.
Another computer-based method is designed to detect pres- A leak detection capability curve can be developed by esti-
sure waves. A leak will cause a negative pressure wave at the mating the leak detection capabilities of each available method
leak site. This wave will travel in both directions from the leak for a variety of leak rates. A table of leak rates IS first selected
at high speed through the pipeline product (much faster in liq- as illustrated in Table 7.13. For each leak rate, each system’s
uids than in gases). By simply detecting this wave, leak size and time to detect is estimated.
location can be estimated. A technique called pressure point In assessing leak detection capabilities, all opportunities to
cznalvsic. (PPA) detects this wave and also statistically analyzes detect should be considered. Therefore, all leak detection sys-
all changes at a single pressure or flow monitoring point. tems available should be evaluated in terms of their respective
By statistically analyzing all of these data, the technique can abilities to detect various leak rates. A matrix such as that
reportedly, with a higher degree of confidence, distinguish shown inTable 7.14 can be used for this.
between leaks and many normal transients as well as identify References [3] and [4] discuss SCADA-based leak detection
instrument drift and reading errors. systems and offer methodologies for evaluating their capabili-
A final method of leak detection involves various methods of ties. Other techniques will likely have to be estimated based on
direct detection of leaks immediately adjacent to a pipeline. time between observations and the time for visual. olfactory. or
One variation of this method is the installation of a secondary auditory indications to appear. The latter will be situation
conduit along the entire pipeline length. This secondary con- dependent and include considerations for spill migration and
duit is designed to sense leaks originating fiom the pipeline. evidence (soil penetration, dead vegetation. sheen on water,
The secondary conduit may take the form of a small-diameter etc.). The total leak time will involve detection, reaction. and
perforated tube, installed parallel to the pipeline, which allows isolation time.
air samples to be drawn into a sensor that can detect the product As a further evaluation step, an additional column can be
leaks. The conduit ]nay also totally enclose the product pipeline added to Table 7.14 for estimates of reaction time for each
(pipe-in-pipe design) and allow the annular space to be tested detection system. This assumes that there are diferences in reac-
for leaks. Obviously these systems can cause a host of logistical tions, depending on the source ofthe leak indication. A series of
problems and ai-e usually not employedexcept on short lines. SCADA alarms will perhaps generate more nnmediate reaction