Joanna Mabe and Keefe Murphy, ATMOS International Limited, Manchester, UK;
and Gareth Williams and Andrew welsh, BTC company, Baku, Azerbaijan
Due to stringent environmental requirements, it is essential for a leak detection system to work from the moment that crude oil is introduced into the pipeline. Without any prior operational data and with the pipeline partially filled, it is challenging for the leak detection system to monitor the integrity of the pipeline throughout the whole filling process.
For these reasons, it is useful to describe the process of incremental pipeline filling and the phased commissioning of a real-time leak detection system (LDS) for the 1,768-km (roughly 1,100-mi) BTC crude oil pipeline. A key aspect of this innovative system is that it delivers real benefit at the first point of hydrocarbon introduction, and is close to optimal tuning on day one of full production. The application of the pig tracking software to track the oil front as the crude displaces nitrogen is a key part of this process, and is therefore also discussed.
The BTC pipeline
The BTC crude oil pipeline runs from the Sangachal terminal near Baku in Azerbaijan, via Tbilisi in Georgia, through Turkey to an export terminal located near Ceyhan on the Turkish Mediterranean coast.
BTC committed to partners, lenders, NGOs and others that the highest possible performance of leak detection would be provided from the point when hydrocarbons were introduced into the pipeline. Following successful completion of all hydrotests and commissioning activities, some 700,000 Sm3 of nitrogen were injected into the pipeline up to IPA1, the first intermediate pigging station along the line. This nitrogen remained in the pipeline all the way to Ceyhan. It was contained by closing block valves downstream of the injection point. When crude was pumped into the pipeline, two “line fill” pigs were used to separate the crude from nitrogen. Therefore, there were two challenging tasks for the on-line leak detection and pig tracking system:
- Accurate tracking of the oil front using the pig tracking system. Due to elevation changes, the speed of a pig varies significantly when traveling in an uphill section compared to when traveling in a downhill section. Since crude is much heavier than nitrogen, the changes in pig velocity are much greater than in a normal liquid-filled pipeline.
- Reliable leak detection for the partly filled pipeline section without any historical data. As the crude starts to fill one section, the leak detection system has to monitor the integrity of the section based only on the readings from the inlet flow meter and pressure transducers along the section.
Despite these challenges, leak detection and pig tracking were provided, but these activities required close analysis of the system data in line with the pipeline operations and close cooperation of the engineering and operations teams.
Facilities description
As noted, the pipeline is 1,768 km (roughly 1,100 mi) in length, and is routed through some extremely mountainous terrain, rising to greater than 2,500 m in Georgia, remaining above 1,500 m until near KP 1,650 in Turkey, where it falls steeply towards the Mediterranean Sea. Figure 1 shows the elevation profile of the whole pipeline. Slack flow is inevitable at several high points along the pipeline in Georgia and Turkey. The leak detection system has to continue to work under such slack flow conditions.
The pipeline is 42-in. diameter for most of its length. There is a 46-in. section throughout Georgia and a 34-in. section in Turkey for the last 125 km into the Ceyhan Terminal. There are eight pump stations along the pipeline and 87 block valve stations. Pumping stations are at Sangachal terminal in Azerbaijan, a further station in Azerbaijan, two in Georgia and a further four pumping stations in Turkey. Each of these stations has a pig launcher and receiver, and there are three further intermediate pigging stations.
The pipeline is well instrumented. At each of the pumping stations in Azerbaijan and Georgia there are upstream and downstream ultrasonic flow, pressure and temperature meters available. At the pumping stations in Turkey there are upstream and downstream pressure meters, and downstream ultrasonic flow and temperature meters.
At all block valve stations, there are upstream and downstream pressure meters and downstream temperature meters. Upstream ultrasonic flow meters are available at six of the block valves where slack flow is expected when running at reduced flow rates. Additionally, fiscal positive displacement flow and density meters are available at the first pumping stations in Azerbaijan, Georgia and Turkey, respectively. As an example, Figure 2 provides an overview of the Azerbaijan Section.
LDS description
In order to achieve the best possible leak detection performance as committed by BTC, redundancy is provided at several levels. These are described below.
Redundancy by LDS technology. By the time the pipeline becomes fully operational, there will be two independent leak detection systems running in parallel:
- Statistical leak dection – detects leaks by performing statistical analysis of flow imbalance calculations using the flow and pressure measurements along the pipeline under running conditions. Under static conditions (pipeline shut-in), the system relies on pressure and temperature measurements to perform leak detection.
- Hydraulic model-based leak detection – detects leaks by modeling the pipeline behaviour and then comparing the predicted flow and pressure with flow and pressure measurements from the pipeline instruments. Then it checks the discrepancy to see if it indicates a leak.
The hydraulic model-based leak detection system cannot become operational until the pipeline is full and running in steady state. Therefore, the focus here is solely on the application of the statistical system for the incremental line fill, leaving the traditional system to be commissioned when the line is full during production.
Redundancy of LDS hardware. The statistical leak detection system runs on a stand-alone PC at both the Sangachal Control Room (SCR) and Ceyhan Control Room (CCR). The LDS PC at SCR interfaces with the ICSS at SCR, and the PC at CCR interfaces directly with the ICSS at CCR (Figure 3). Communication is by means of OPC over an Ethernet network using TCP/IP. The system at CCR is a copy of that at SCR and they run independently of each other.
Redundancy within the LDS Architecture. The pipeline is considered in discrete sections, as shown in Figure 4. These include:
- Seventeen sub-sections where flow readings are available
- Three main “country” sections – Azerbaijan, Georgia and Turkey
- One further section – monitoring the entire pipeline.
This configuration allows the LDS to achieve the best possible leak detection sensitivity and response time, and provides a high level of redundancy.
In the event of losing a sub-section, for instance due to a critical instrument fault, leak detection is still provided by the “country” section. It also allows slack flow areas to be isolated so that the sensitivity and performance of other sections do not have to be compromised. Figure 5 shows the summary screen of the above leak detection sections.
Each of the redundant leak detection systems rely on flow, pressure and temperature measurements along the pipeline to determine if a pipeline leak has occurred. Therefore, to have the leak detection capability available, it is essential that the following conditions are satisfied for the pipeline section to be filled prior to start-up:
- The pipeline instrumentation commissioned and in good working order
- Communications from the field and ICSS verified
- The interface between the ICSS and LDS tested and commissioned.
Typically, commissioning of a leak detection system starts when the pipeline is already full, the above conditions have already been satisfied, and the following conditions have been met:
- Most model-based systems rely on the pipeline being in a steady state before the tuning process can begin. Thus leak detection during the filling operation is not possible.
- Some mass balance and neural network systems require leak trials to be carried out before they can be tuned.
- Statistical systems require “normal operation” including transients caused by pumping operations, pump trips and changeovers.
In the case of the statistical leak detection system for the BTC pipeline, since it had been sectionalized, it could be commissioned as soon as its sections were filled. The system could therefore provide leak detection from the moment crude oil filled the first section. As more pipeline sections were filled, more pig tracking and leak detection systems were activated. To provide full support to BTC, one project engineer from ATMOSi was on site for each of the incremental pipeline fill periods. In addition to the daily support to the commissioning team, the engineer was also responsible for tuning the leak detection systems for the liquid-filled sections so that their performance was optimized.
In addition, some parameters of the standard algorithm were temporarily modified during the line fill phase to be able to provide leak detection in the partially filled pipe section. This meant that leak detection was possible as soon as oil first entered the pipeline. There is also an application to track pigs in the pipeline, both cleaning and line fill pigs. The location of the line fill pigs defined the location of the oil front as the pipeline was being filled. This proved to be particularly valuable during line fill, where traditional on-the-ground pig tracking techniques suffered as a consequence of vapour pockets and steep downhill land profiles.
Filling operation
The line was filled with crude oil at a fixed rate of approximately 150 mbd. The Oil/N2 interface was maintained using two high-seal, bi-directional pigs known as the “line fill” pigs which were spaced approximately 400 m apart. Ahead of the “line fill” pigs was a large pressurized nitrogen blanket approximately 120-km long. The nitrogen was used to inert the pipeline and maintain a backpressure to help control the speed of the pigs in downhill sections.
Selected block valves in the section being filled were used to control the nitrogen pressure ahead of the “line fill” pigs. These specific valves were left closed and the nitrogen throttled around them using their bypass valves as the oil front approached. Once the pressure had equalized at either side of a closed block valve, the bypass valve was closed and the block valve was fully opened. Nitrogen pressure then built up against the next closed block valve, and the procedure was repeated. This ensured the pressure in front of the pigs was maintained as much as possible throughout the Oil/N2 displacement.
Because a partially filled section behaves so differently from a full section, leak detection under these conditions is normally considered impossible. However, with the high level of instrumentation on the BTC pipeline, the ATMOS Statistical Pipeline Leak Detection (SPLD) system was applied to each partially filled section as soon as crude entered the section. Due to the uncertainties in the pig behavior and potential faults in the newly installed instrumentation system, all leak alarms were diagnosed in conjunction with the pig tracking system.
System optimization
Once a section got filled, the actual pipeline data for this section was used to tune and optimize the leak detection system. As more sections were filled, more subsystems were activated to monitor more of the pipeline. The performance of each sub-system improved as more actual pipeline data became available. Each section was tuned to full sensitivity during steady-state operations. However, transients (such as unexpected pump trips) during this time were not expected to be representative of normal operations, so the system was tuned to a reduced sensitivity during line fill transients. It should be noted that the final fine tuning of the system was slightly different from the initial tuning carried out during incremental line fill, and required the entire pipeline to be in operation.
Once a section of the pipeline was full, the standard leak detection algorithm was applied as the minimum instrumentation requirements were fulfilled: flow and pressure measurements at both the inlet and the outlet of the pipeline section. The principle of the standard system is illustrated in Figure 6. Flow and pressure measurements along the line are the significant variables feeding the algorithm after data validation checks are carried out. Then, statistical analysis is performed on the flow imbalance calculation to determine whether a leak condition is present. If a leak is detected, then an alarm is generated and an estimate of the leak size and location is reported.
The system also adapts itself to changes that occur in the pipeline. The system accommodates inventory changes by monitoring pressure variations in the pipeline and normal metering errors by continuously estimating the flow difference between the inlet and the outlet. The minimum the system requires to be tuned is a sample of “normal operation” including transients caused by pumping operations and throughput changes. Once a filled section of the pipeline had been tuned, then the settings for this tuned section were used as a starting point for subsequent newly filled and un-tuned sections. This allowed newly activated sub-sections to start off at a more realistic sensitivity, thereby minimizing the number of false alarms being generated.
Line fill system
When the crude oil first entered a pipeline section, the reliability and performance of the overall instrumentation system were uncertain, particularly for the part where no crude was present. During a line fill operation, the hydraulics in the pipeline were significantly different from those in a filled section. The pressure downstream of the “line fill” pigs was governed by nitrogen being moved along the pipeline. This gas-filled section provided a compressible cushion for the “line fill” pigs to push against, and, as such, did not behave as a liquid would.
In addition, the minimum instrument requirements were not met: the outlet flow measurement needed to be disregarded, as that section of pipe was effectively empty. Also, the number of pressure measurements increased as the section was gradually filled. Due to the above challenges, standard leak detection technologies cannot perform leak detection effectively.
For the “line fill” section, some parameters in the standard algorithm needed to be adapted. There was no outlet flow measurement, so a raw flow difference was not available. The whole inventory was continually changing, so the inventory correction was not as it would be for a filled section. The algorithm was modified to use all of the available data. i.e. the inlet flow reading and all available pressure measurements.
As the pipeline section was filled, data was analyzed so that the rate of change of pressure in the line could be monitored for leaks alongside the inlet flow measurement. The inlet flow of the section was also monitored to establish whether the pipeline was under shut-in or running conditions. The performance of the “line fill” section gradually improved as the oil front passed more block valves and more valid pressure measurements were incorporated into the calculations. This system relied on maintaining the pressure along the line and having a functional inlet flow measurement at all times as these were the inputs into the statistical algorithm. This adaptation was only required for partially filled sections, once full this method was disregarded and the standard algorithm was applied.
Fine tuning the final system
After the crude filled the whole pipeline, the leak detection system was fine-tuned and all the sub-systems, whole country and whole pipeline systems were activated. The system performance was optimized with minimum false alarms and highest sensitivity.
Pig tracking
The Pig Tracking Module (PTM) integrates with the ATMOSi leak detection software, and adds the capability to track the position of a pig that is inserted into the pipeline. The PTM provides the estimated position of the pig within the pipeline, the velocity of the pig, and the estimated time of arrival (ETA) at the pig receiver, as well as the ETA at every intermediate block valve along the way. The PTM allows for multiple pigs within the same section of the pipeline, and allows a library of default characteristics to be stored for a number of different pig types, including cleaning and in-line inspection (intelligent) pigs.
When a pig is launched into the pipeline, the PTM receives its signal to start tracking from the pig launcher. The pig velocity, location and ETA are updated based on previously saved data for the “pig friction factor.” When the pig passes a block valve, the “pig friction factor” is updated and corrected. The pig velocity and ETA are also updated and corrected at this time. On the first pig run, this “pig friction factor” is a “best guess” based on an operator entered slippage/bypass figure, but it is corrected as soon as the pig passes the first block valve. It is then re-corrected on passage through each block valve, since slippage may change during the pig run due to pig wear, wax build up, and other reasons.
Normally, the PTM uses the upstream and downstream pressure signals at every available location to monitor, correct and update the pig position and “pig friction factor,” which is used in the calculation of pig velocity. Experience suggests that this method is more reliable than using pig signals at each block valve. If pig signals at the block valves are received, then they are acknowledged and used. However, the system is not dependent on receiving these pig signals. If the “pig in line” signals from the pig launchers fail, then the operator has the ability to enter/update the “pig launched time.”
During the line fill operation, the PTM was used to track the “line fill” pigs, which were representative of the oil front moving along the pipeline, replacing the nitrogen ahead. This function proved to be valuable to BTC in terms of tracking the oil front as well as diagnosing leak alarms when a pig moving downhill generated an increased pressure drop upstream.
Line fill examples
In the event of a leak occurring, the expected behavior of a line fill section was that the line pressure in the filled section would drop, and there would be a consequential increase in the inlet flow. If the terrain was flat, the line pressure would rise continuously as the line was filled. However, given the elevation profile, the average line pressure would sometimes fall depending on where the oil front (“line fill” pig) was with respect to the elevation profile; i.e., on downhill sections. This was due to the “line fill” pigs and the oil front behind them free falling against the pressurized nitrogen ahead. Hence, the LDS was expected to generate a leak alarm when the “line fill” pigs briefly free fell on the downhill sections, since this behavior mimics the flow and pressure pattern expected during a leak.
Because the PTM was predicting the location of the “line fill” pigs, it could be seen that the pigs were in a downhill section and that this alarm was most likely triggered by the pigs and oil front travelling downhill, and was not a leak.
This behavior was exploited to verify that the LDS was behaving as expected and was operational given that this event generated a leak alarm. The alarm generated triggered the leak location algorithm, which reported a location that corresponded to the BV immediately upstream of the oil front/pig. This information could then be correlated with the data from the PTM. The leak detection system for the partially filled section was sensitive. However, some alarms were generated and the system needed to be interrogated intelligently and in line with the prevailing operational scenario to decide if the alarm was genuine.
Conclusions
Traditionally, leak detection is one of the last systems to be commissioned and come on line in a pipeline project. This is sometimes due to a lack of commitment or a lack of belief in leak detection technology, leading to a lack of urgency. Inevitably, this often results in the production taking priority over LDS commissioning until some time after the pipeline is in full flow.
However, it is also widely accepted that leak detection and pig-tracking systems cannot be commissioned before the pipeline is fully operational. This is largely true in that the final LDS commissioning cannot take place before the pipeline is fully operational.
BTC and ATMOSi have cooperatively striven to break this paradigm and develop a leak detection system that delivers real benefit at the first point of hydrocarbon introduction and is close to optimal tuning on day one of full production. Working together, the two companies have demonstrated that leak detection is possible in a partially filled pipeline and during line-fill; i.e., at all times from the moment fluid first enters the pipeline.
The results from the LDS and the PTM during line-fill must be used with a high degree of understanding, and cannot be used in isolation from other systems. The operator must have knowledge of where the oil front is with respect to the elevation profile, knowledge of the current state of commissioning of all the associated systems (pumps, stations, status and accuracy of instrumentation, block-valves), and in-depth knowledge of how the LDS and pipeline operations interact with each other.
The benefits of commissioning the real-time ATMOS Statistical Leak Detection System on the BTC pipeline during line fill are clear:
- The LDS had already been tested over 12 months by the time oil reached the terminal at Ceyhan.
- The LDS was successfully detecting illegal taps and thefts soon after the pipeline became operational. Typically, at this time in a pipeline project, the leak detection system commissioning would only be commencing.
- The LDS had been tested many times by commissioning processes during line fill, while commissioning pump stations, filling pump loops and commissioning surge relief tanks. These effectively served as real leak trials which the LDS responded to leading to a high degree of acceptance and operator confidence.
Acknowledgments
The authors wish to express their thanks for the support provided by both BTC and ATMOSi for this study. Based on a paper presented at the ASME’s 6th International Pipeline Conference, Calgary, Alberta, Canada, September 25-29, 2006.
Recommended Reading
Solaris to Acquire Mobile Energy Rentals, Rename to Solaris Energy Infrastructure
2024-07-10 - Following the closing of its deal to acquire Mobile Energy Rentals, Solaris Oilfield Infrastructure will also be rebranding to Solaris Energy Infrastructure to more closely represent its expanded solutions offerings.
Come Together: California Resources, Aera Merge for Scale, Drilling Runway
2024-07-10 - California Resources Corp. closed an acquisition of Aera Energy to become California’s top oil and gas producer. Now, CRC President and CEO Francisco Leon wants to grow from a one-rig to an eight-rig drilling program—but faces stiff pushback from regulators and environmental advocates in the Golden State.
Blackstone Buys Enagás’ Tallgrass Stake for $1.1 Billion
2024-07-11 - Spain’s Enagás is selling its Tallgrass Energy interests to Blackstone Infrastructure Partners in exchange for some needed capital.
APA Closes Midland Basin, Eagle Ford Divestitures for $660 Million
2024-07-11 - APA Corp. and subsidiary Apache sold non-core assets in the Permian Basin and Eagle Ford Shale sooner than expected, and for less, as the company looks to reduce debt from its Callon Petroleum deal earlier this year.
Firms Blast ‘Conflict-ridden’ Martin Midstream Deal, Launch Counteroffer
2024-07-11 - Two New York-based capital firms say a May proposal by Martin Resource Management to buy Martin Midstream for $100 million represents a “below market and conflict-ridden proposal,” while the firm’s own offer has been rebuffed.