Jon R. Bose, Marcus K. Olson
Alyeska Pipeline Service Co.
Anchorage
Since start-up of the Trans-Alaska Pipeline System (TAPS), operator Alyeska Pipeline Service Co. (Alyeska) has developed, maintained, and upgraded a highly sophisticated leak-detection system.
The pipeline's length and unique physical and operational characteristics make leak detection a technological challenge for the designer.
Alyeska continues to evaluate new technologies and their possible TAPS applications.
THE SYSTEM
TAPS transports hot crude oil from Alaska's North Slope fields on the Arctic Ocean to a tidewater marine terminal at Valdez, Alas., located on ice-free Prince William Sound.
It consists of slightly more than 800 miles of 48-in. steel pipe, half of which is above-ground and insulated. The remainder is uninsulated and buried conventionally.
Crude-oil temperatures range from 120 F. at Pump Station (PS) No. 1 to 85 F. on arrival at Valdez.
Transported crude oil is a mixture from as many as five major oil-bearing formations, each with its own characteristics and custody-transfer measurement. Although similar, each has distinctive properties which affect pipeline hydraulics. Mixing ratios vary as a result of planned maintenance and unplanned upsets.
Pipeline flows have varied from 600,000 to more than 2 million b/d. The static line fill inventory of the pipeline system is slightly more than 9 million bbl. This volume must be accounted for regardless of flow rate.
Eleven staffed pump stations are distributed along the pipeline (Fig. 1). Each of these stations has a crude-oil breakout tank in a berm-enclosed containment area as well as incoming and outgoing flow metering.
System hydraulics are characterized by extensive use of drag-reducing additives (DRA) and the presence of multiple slack-line regions of varying extent (Fig. 2). As will be discussed presently, each of these conditions presents unique challenges in modeling fluid flows in pipelines.
Crude-oil storage at the Valdez marine terminal consists of 18 cone-roof tanks with a total working inventory of approximately 8.9 million bbl.
Because of the pipeline's construction through wildlife habitat, design of the line paid considerable attention to the potential damage of any oil spill.
Extensive modifications to the original pipeline design sought to minimize any spill. This concern led to installation 142 pipeline valves (Fig. 3) outside of pumping stations solely to limit spill volumes.
Coupled with the design considerations for spill minimization, comprehensive oil spill-response plans, including reconnaissance, containment, and clean up provisions were established and continue to be updated and improved as technology allows.
As are all pipelines, TAPS is regulated by a broad spectrum of state and federal agencies, including U.S. Departments of Transportation and Interior and the U.S. Environmental Protection Agency.
Additionally, TAPS is regulated by the Stipulations and the Agreement and Grant of Right of Way from the governments of the U.S. and Alaska. These documents specifically require a leak-detection system at the time of construction and commissioning of the pipeline.
LEAK DETECTION
Since start-up in 1977, Alyeska has recorded 10 incidents resulting in crude-oil spills of 50 bbl or more in the pipeline, pump station, and terminal systems. Six of these were recorded outside pump stations, 3 within pump stations, and 1 within the terminal.
Each incident was first discovered by visual observation, not leak-detection systems.
In some cases, personnel were on site at or soon after the time of the leak, and subsequent action occurred before detection by leak-detection systems.
In others, the leak rate was less than detection thresholds, or pipeline-operating instabilities prevented the leak-detection system from immediate detection before visual discovery.
Subsequent investigations of some events indicated that the leak-detection system had sensed a change in operations, but visual discovery was made before a system alarm.
Pump station and terminal tankage is located inside lined containment dikes designed to prevent escape of any spilled fluid.
Patrols covers these areas several times daily looking for evidence of any tank or related piping leaks. In addition, any maintenance activities within tank farms include visual damage and leak surveillance.
Tank volumes in the pump stations and terminal are monitored by liquid-level indication. Tanks not actively receiving or delivering oil are monitored for small changes in level.
If small changes not attributable to normal changes in fluid temperature, vapor loss, etc. become evident, further investigation of the tank area is made.
The current setpoint for alarm is approximately 0.10 ft of level: roughly 865 bbl at the terminal or 70 bbl for a typical pump station tank.
TAPS uses five different methods to detect leaks along the pipeline:
- Visual surveillance
- Flow deviation
- Pressure deviation
- Flow difference deviation
- Computer analysis of volume balances.
All of these methods are supported by an extensive data-acquisition system.
Operating conditions and status of critical systems are relayed to the operations control center (Fig. 4) in Valdez via a sophisticated supervisory control and data acquisition (scada) network consisting of redundant microwave and satellite communications paths.
A pair of Data General Eclipse MV/20000 computers operating in on-line/standby mode receive, route, and process this information. The control system monitors and processes more than 3,000 data points at a nominal scan frequency of approximately 4 sec.
These data are presented in both tabular form for historical purposes and graphically for use by the controller personnel. Software in the control system makes validity checks on the data before they are used or recorded.
Physical surveillance of the pipeline by driving, walking, and overflight is one of the primary means of detecting the occurrence of any leak, especially very small ones.
TAPS has procedures and schedules for physical pipeline surveillance by land and air in all seasons. Part 195 of the U.S. Code of Federal Regulations (CFR) Part 49 requires visual surveillance 26 times/year at no greater than 3-week intervals. In addition, any maintenance activities conducted along the right-of-way include visual surveillance.
FLOW, PRESSURE ALARMS
Deviation alarms based on rapid changes in flow or pressure data indicate possible leaks in the range of 1% of throughput and typically respond within several minutes.
Station suction or discharge pressure variations of more than 1% of scale (15 psi) generate a pressure-deviation leak alarm.
These alarms are inhibited for a period of 10 min following normal changes in operational parameters, such as control pressure setpoint changes.
This inhibition ensures that such changes will not trigger an alarm. In this manner, the deviation alarm system would detect a sharp transient generated between stations by a line break.
If the flow rate into or out of a station varies by more than 700 bbl/hr, a flow-deviation alarm appears to the controller. Again, the 10-min interval for alarm inhibition is established so that normal operational changes in flow will not trigger an alarm.
A flow-difference deviation alarm is generated if the difference between an upstream and downstream flow measurement changes by more than an established tolerance (typically 700 bbl/hr). This method allows detection of a leak which does not necessarily generate a sharp pressure or flow transient by looking for a segment's flow-balance difference.
Each of these systems provides detection abilities for large leaks which generate sharp pressure or flow transients. Typically, the sensitivity of the deviation alarm systems is approximately 1% of maximum flow, or about 20,000 b/d.
Note that while the leak rates detected by such a system are sizeable, they are detected quickly, resulting in a relatively small potential spill volume.
VOLUME BALANCE ALERT
The TAPS line-volume-balance (LVB) leak analysis system has been developed to detect the presence of small persistent leaks which may have evaded detection by the pressure and flow deviation methods.
An LVB system was designed for original construction in 1974 and underwent major revisions in 1979, 1982, and 1988.
Line-volume balancing involves a computation, performed every 30 min, of the quantity of oil gained or lost by the pipeline over that period. The sum of all pipeline receipts for a 30-min period is subtracted from the sum of all pipeline deliveries over the same period.
The difference in metered volumes is adjusted by inventory changes in station tankage and each pipeline segment to produce a more realistic estimate of the system's gains or loss.
Accumulation of these gain-loss results for a succession of 30-min periods allows the cumulative imbalance for the total period to be used to compute an apparent average leak rate for the period.
The long averaging period permits the method to detect a small leak after a delay during which the cumulative imbalance, or total spilled volume, is progressively increasing.
The LVB system searches for the most stable periods of time for which to estimate a leak rate. This search is accomplished through selection of periods in which the standard deviation of the 30-min results is a minimum.
The magnitude of this standard deviation allows establishment of an alarm limit which varies automatically according to the stability of pipeline operation.
After removal of a long-term trend value, the resulting gain-loss estimate compared with the alarm limit indicates whether a leak alert should be generated.
ADVANCES
Improvements in TAPS's LVB system have paralleled upgrades in computing power and capacity in the pipeline control system.
When first installed at pipeline start-up, the computing system data base had no provision for correction of corrupted or missing data. This version of the software also had rudimentary line-pack calculations restricted in complexity by the limited capacity of the computing hardware.
In its original implementation, the LVB had a fixed alarm limit of 3,000 b/d.
Subsequent increases in computing power enabled more-sophisticated algorithms to be used in the calculation of segment volumes in the LVB methodology. Expanded survey profiles of the pipeline topography, improved data validation, and inclusion of a variable alarm limit based on pipeline-operating stability were enabled by the more powerful computers.
In addition, integration of the LVB system with the historical data base of operating conditions, enabled improved statistical methodologies to be used in trend analysis and alarm-limit calculations. These improvements significantly reduced the number of false alarms.
Integration of the latest pipeline control-system hardware included individual-segment alarm analysis, improved long-term trending analysis, and graphics capabilities for presentation of information to the pipeline controller.
While the LVB system is based essentially on metered custody-transfer volumes at either end of the system, intermediate metered deliveries and receipts, and tank-level changes, the intervening volume of all interstation line segments may undergo significant changes because of a variety of operating conditions. These changes must be compensated for.
Recall that the temperature-corrected line fill of the TAPS system, regardless of flow, is slightly more than 9 million bbl.
Changes in bulk temperatures and pressures can alter this volume by many thousands of barrels in a single line segment. Cumulative errors in the calculation of volumes due to pressure and temperature can, therefore, strongly affect leak-detection abilities.
The differential volume of a given line segment as a result of pressure or pressure pack is based on the wall thickness of the pipeline, which varies in each segment, the bulk modulus of elasticity of oil in steel, and constraint of the pipe as a result of buried or above-ground construction.
Temperature pack is determined from a calculated average segment temperature and the thermal coefficient of volume expansion in steel.
SLACK-LINE CALCULATIONS
Both of these measurements of line-pack volume changes are fairly straight-forward, and with improved pressure and temperature modeling techniques can be quite accurate. Unfortunately, the existence of slack-line regions in many of the pipeline segments requires another line-pack calculation, which is less straightforward.
Slack line, or open-channel flow occurs whenever the pipeline operating head gradient profile intercepts the pipeline elevation profile. At the onset of slack-line flow, the pipeline cross-sectional area is no longer completely filled, and estimates of the local line volumes become difficult to calculate accurately.
Several slack-line regions along the pipeline exist under essentially all operating conditions and are more easily estimated. There exist many potential slack regions that may appear and disappear relatively quickly; some slack lines may exist for only a few minutes under certain conditions.
The exact timing of the onset of slack flow, or the annihilation of a slack region is never precisely known, which leads to inaccuracies in segment volumes over a short period of time. Over longer time periods these inaccuracies tend to cancel and yield reasonable detection limits under steady operating conditions.
Because of the difficulty of estimating slack volumes, two differing methods calculate the estimated volume change. These, combined with a third choice (that of no change in the volume), bound the estimate of the line-pack change resulting from slack operation.
All remaining regions not operating under slack-line conditions follow classical fluid-flow relationships for pressure, temperature, flow velocities, etc.
Leak analysis is based on the most recent 24 hr of gain or loss of volume from the pipeline inventory data and the long-term gain-loss trend beginning 48 hr previous and extending 7 days back.
Gain-loss is derived from the total pipeline deliveries, total receipts, tankage changes, and line pack compensations. This scheme prevents a leak from entering the long-term trend too quickly and being masked.
Leak alerts are generated when the most recent 24-hr gain/loss results significantly diverge from the long-term trend.
LEAK-ALERT TESTING
Testing for a leak alert involves the following three steps:
- Current leak-rate estimation
- Current alarm-limit calculation
- Trend analysis.
The average of the past 48 half-hourly gain/loss values provides an estimate of the current leak rate. This time interval is searched to identify two statistically independent stable periods of operation over which to estimate the leak rate.
The interval must include the current gain/loss result, but the two periods may not overlap. The search is structured to ignore periods of operational instabilities.
The current alarm limit is established from standard statistical methods to determine how closely the estimated leak rate approximates the overall 24-hr average gain/loss. A probability distribution is used to determine confidence limits for the leak-rate estimate. The confidence-limit value then becomes the alarm limit.
The long-term gain/loss trend is then removed from the estimated leak rate. If the resulting value exceeds the alarm limit, a leak alarm is generated. The generation of an alarm indicates that a maintained shift has occurred in the past 24-hr period as compared to the average established in the previous week's operation.
The gain/loss history and the operating stability of the pipeline determine in large part the alarm limits for the LVB system. In practice, the alarm limits may range from several hundred barrels a day under extremely stable conditions to more than 12,000 b/d under extreme transient conditions.
Typically, alarm limits fall within a 1,000-6,000 b/d range more than 70% of the time.
The LVB calculations being made once per 1/2 hr show that if the pipeline begins experiencing transient operation immediately before the data are gathered, the LVB system's hydraulic calculations must make assumptions regarding the previous 29 min and presume that the current pressure and temperature conditions have prevailed since the last calculation.
LIMITS, CHALLENGES
While the TAPS LVB system is highly sophisticated and provides in most instances levels of leak detection far less than 1% of the pipeline flow rate, there are limitations in its lower detection thresholds.
As with any leak-detection system, a tradeoff exists between leak-alarm threshold limits and the occurrence of false alarms. Low detection limits are desirable but come at a cost of an unacceptable number of false alarms.
False alarms tend to desensitize operators over time to the extent that they may ignore a genuine alarm. The challenge to designers of such a system is to maximize sensitivity while simultaneously minimizing spurious false alarms.
TAPS LVB limitations primarily result from instrumentation accuracies, line-pack uncertainties, and transient operation effects.
Pipeline instrumentation for pressures, temperatures, and flows has typical tolerances of 0.25% of scale. Significant improvements in this accuracy cannot realistically be expected. Recalling, however, that LVB relies on statistically derived methods and routines based on differential mathematics, it is the repeatability and error rate of change which are most critical, not absolute accuracy.
TESTING PROGRAM
As part of Alyeska's ongoing efforts in leak detection, the company has conducted a series of tests to measure the ability of the LVB system to detect actual controlled removal of oil from the pipeline. Four such tests have been conducted.
The test method was based on removal of oil to pump station tankage by means of a metered and controlled-flow apparatus. Local instrumentation for pump station flow and tank level was disabled to convince the LVB system that a leak had occurred along the pipeline right-of-way.
The LVB system has successfully detected controlled leaks of 4,500, 3,500, and 2,800 b/d in times of 13, 6, and 8 hr, respectively. A fourth test at 800 b/d went undetected under the prevailing operating conditions of the test.
Under the pipeline flow conditions of approximately 1.8 million b/d during the tests, the detected leaks represent a range of 0.156% to 0.25% of pipeline throughput.
In 1991, Alyeska commissioned two independent studies of TAPS's leak-detection systems to evaluate its capabilities compared to new technologies.
Both studies agreed that TAPS's systems were complete and more sophisticated than most installed systems. But each noted that improvements in leak-detection times were available, although absolute sensitivity thresholds probably could not be significantly improved.
Each of the two studies recommended real-time modeling (RTM) as a means for satisfactorily dealing with transient conditions in TAPS. With the ability to recognize transient hydraulic behavior, the suggestion was made that a model-based system would be able to distinguish between normal changes in pipeline operating conditions and the appearance of a leak.
In addition to the recommended RTM solution, other technologies are available in the marketplace. Alyeska has evaluated the possible application of several of these to TAPS.
Hydrocarbon-detection systems such as ribbons and cables are deemed impractical for retrofit to TAPS because of the system's length, depth of burial of the pipeline, and maintenance costs.
Coupled with this impracticality is limited location ability, limited usefulness after contamination, and the inability to identify the source and nature of the hydrocarbons being detected (for example, a third-party spill).
Software-based leak detection systems based solely on pipeline pressure data are seen to have possible utility in short line segments such as river crossings but lack proven ability to function effectively in regions of potential slack-line conditions.
In addition, the nature of these systems is to detect the actual leak event at the moment it first occurs. If that event is missed, the leak may go undetected.
Other proposed technologies such as satellite imagery, ground-penetrating radar, and high altitude infrared photography are experimental, especially in a high latitude environment.
RTM PROGRAM
In 1992, Alyeska contracted to fund development of a real time model-based leak-detection system applicable to the unique operation of TAPS. This model is a transient-based volume balance, or TVB model that can respond to transient conditions which must be compensated for in order effectively to predict TAPS's hydraulic behavior.
Perhaps the most important of these considerations is a rapidly varying throughput rate in the pipeline.
Because there is little storage capacity at Prudhoe Bay the pipeline is effectively tightened to the North Slop producers. Changes in production are quickly felt downstream in TAPS.
Pump starts and stops, relief events, changes in mid-point deliveries and receipts, and related flow changes must not affect leak-detection accuracy.
Currently, TAPS has two line segments which are drag-reduced under all normal operating conditions. Since DRA changes the local friction factor of the pipeline, any system modeling transient hydraulics must be able reliably to track DRA history over time and calculate local drag reductions and friction factors.
TAPS operates with multiple slack-line regions at all times, and certain operating conditions can induce short-lived slack conditions where they are not normally encountered.
A model-based system must be able to recognize slack conditions accurately and calculate slack volumes. If DRA passes through a slack region, its effectiveness must be recalculated and tracked downstream.
Finally, the new model must also be flexible in its ability to be modified as conditions warrant. North Slope production has begun to decline. Future opportunities for the shutdown of existing facilities will present themselves and must be anticipated in the design of the leak-detection system.
Design goals for the new TVB system include response to rapidly changing operating conditions, effective modeling of hydraulic performance, and flexibility for the future.
The system will also provide leak-location abilities through calculation of individual segment volume balances and deviation analyses.
Perhaps most important, the TVB system is expected to provide faster detection than provided by LVB.
Although leak-rate sensitivities are not expected to be significantly lower than possible with the LVB, faster sampling and real-time modeling will reduce response times from leak occurrence to leak detection. In the event of a leak event, this means less damage and subsequent clean up.
Copyright 1993 Oil & Gas Journal. All Rights Reserved.
