Flexibility need prompts installation of Zeepipe modeling system

Svein Birger Thaule, Willy Postvoll
Statoil
Stavanger

• Conversion of Annual Contractual Quantity to Design Flowrate [64,227 bytes]

Installation by den norske stats oljeselskap A.S. (Statoil) of a powerful pipeline-modeling system on Zeepipe has allowed this major North Sea gas pipeline to meet the growing demands and seasonal variations of the European gas market.

The Troll gas-sales agreement (TGSA) in 1986 called for large volumes of Norwegian gas to begin arriving from the North Sea Sleipner East field in October 1993.

Deliveries will average 45 billion cu m/year (bcm/year), but gas-export commitment from Norway 2000-2009 is 62 bcm/year. Fig. 1 [38,302 bytes] shows committed gas-export volumes from the Norwegians.

Initially, gas deliveries under TGSA were provided by Zeepipe I, the world's largest offshore gas-pipeline system, constructed along with the Sleipner field production facilities.

In May 1996, the giant Troll field, with estimated 1,300 bcm recoverable gas reserves, began producing. New technology was needed to realize this $4.5 billion project.

With the field's daily production capacity of 100 million standard cu m/day (MMscmd; 3.5 bcfd), gas deliveries to the continent have been significantly increased.

Additionally, gas delivered from the various fields differs in compositions. Thus, gas in the transport network exhibits a range in heating value (gross calorific value - GCV) and Wobbe index (WI). These properties must comply with allowable specific ranges upon delivery.

The essentially mature continental market will reflect only seasonal demand changes. TGSA allows for nominations to range 40-110% of daily contractual quantities (DCQ) to account for consumer nominations.

It is important to Statoil to maintain regular gas deliveries from its integrated transport network. In addition, high utilization of transport capacity maximizes profits.

In advance of operations, Statoil realized that state-of-the-art supervisory control and data acquisition (scada) and pipeline-modeling systems (PMS) would be necessary to meet its goals and to remain the most efficient North Sea operator.

Linking Troll and Zeebrugge

Peculiar in Norway's gas-transport network ( Fig. 2 [123,167 bytes]) are long distances of pipelines with no intermediate system control.

Zeepipe I, 813 km (505 miles) long, is operated with no intermediate compression or control because of the relatively deep waters in the North Sea and associated cost of platform installation.

Consequently, the pipe lines are operated at high pressures to achieve capacity. This is illustrated by the two export pipelines for Troll gas which have a design pressure of 191 barg (approximately 2,770 psig) at Kollsnes, the Troll gas onshore terminal. Total distance from Kollsnes to Zeebrugge without recompression is then 1,120 km.

In order to enhance capacity further, an internal pipeline epoxy coating reduces internal-surface roughness, a necessity for long-distance pigging operations.

The 40-in. OD Zeepipe I's coming into service in 1993 achieved the first step towards satisfaction of the TGSA commitment. A characteristic of this 35.5-MMscmd capacity pipeline is the 3-4 day transmission time of the gas from Sleipner to the terminal in Zeebrugge, Belgium.

This interval represents a challenge for the operation, given that buyers' nominations are daily.

After single-stage separation, gas and condensate produced at the Troll platform move together in two 36-in. OD pipelines to the Kollsnes terminal on Norway's western coast.

Specific to operation of these two pipelines is the concern of multiphase flow. Water and condensate precipitate from the gas as a result of ambient cooling by the sea water and pressure drop during transmission to shore. Because the seabed is hilly, liquid slugging has been envisioned.

Two slug-catcher batteries, 175 m long, consisting of four 48-in. OD pipe sections, have been constructed to handle the flow dynamics. After dew point control, the gas is compressed and routed to the pipelines: Zeepipe IIA to the Sleipner riser platform and Zeepipe IIB to the Draupner riser platform.

The world's largest variable-speed electric motors (40 mw) are installed to power the gas compressors.

Contractual issues

To optimize daily operations, operators handle contractual issues together with characteristics of the infrastructure for production and delivery of gas. These issues relate to delivery commitments (swing, gas properties, pressure, and temperature), transport agreements for the various pipeline systems, and field-delivery agreements.

Generally, gas-sales contracts are related to the specific fields. Contractual swing obligations, as stated, range 40-110% DCQ, although other flexibilities have been considered. Simultaneous maximum delivery obligation, however, is normally exercised in short periods during a year. Conversion from annual contractual quantify (ACQ) to DCQ is elaborated in an accompanying box.

The various sales contracts specify the range of GCV and WI together with water, sulfur, CO₂, and H₂S content.

Some fields may deliver gas with properties outside these ranges which may limit operational flexibility. Commingling of gas streams will in those cases be necessary to fulfill contractual commitments.

Operational characteristics of the infrastructure are mainly related to system pressure limitations, system temperature limitations, transport capacities of the pipelines, and bypass flexibilities at Draupner and Sleipner.

Resumption of gas transportation following shutdown at Draupner has been widely studied for optimization purposes. Of concern is to what extent the low-temperature specification of the risers limits operations.

Specific to shut-in situations is the pressure differential that develops across the platform as a result of pressure equalization along the pipelines. During start-up, the gas temperature downstream from the platform will decrease as a result of expansion cooling. This is governed by development of the pressure differential between the pipeline feeding gas to the platform and the pipeline receiving the gas.

Fig. 3 [75,976 bytes] shows the relation between upstream pipeline pressure and the maximum pressure drop across the Draupner platform, based on the minimum allowable riser temperature. This curve is based on an upstream gas temperature of 39° F. Computations are performed with the predictive model (PM).

Supervisory system

Measured data relevant to operation of the transport network are transferred via satellite from remote terminal units (RTUs) to Statoil's transport control center at Bygnes, Norway. These data are gas pressure and temperature at both ends of the pipelines.

In addition, flowrate and gas composition are measured where fiscal metering or metering for allocation is needed. In total, approximately 2,500 data signals are transferred via satellite to the scada master system at the transport control center.

Another 500 parameters are generated by scada from these signals and transmitted together with the data signals to the scada database. The scada master system acts as primary filter for the scada system.

Pressure, temperature, and flowrate are transmitted every 20 sec. Gas compositions from gas chromatographs are updated every 10-15 min.

In 1991, ABB Norway was awarded the contract to deliver a state-of-the-art computerized transport control system to support operations of the new integrated pipeline network under construction.

Modules of the control system are the scada system, automatic message handling system (AMHS), and pipeline modeling system (PMS).

Fig. 4 [174,186] presents a schematic of the computerized control system.

Until recently, the PMS hardware consisted of two parallel DEC 5000/240 computers running under the DEC Ultrix operating system. Although state-of-the-art when installed, the processor capacity was insufficient to support all available applications.

The new hardware, however, DEC Alpha 2100 5/250 servers, possesses 20 times greater processor capacity and a disc capacity of 10 gb.

Operators can employ the various modules simultaneously at their terminals. Operator machine software is distributed to the terminals.

These are supplemented by work stations linked to the office environment that allows information to be analyzed and presented. The various modules are linked in an Ethernet network, and the transport-control system monitors the status of its own computer hardware and external communication lines as well as the gas-transport process.

Scada

The scada module allows the operator to adjust gas deliveries according to buyers' nominations and production plans under established operational conditions.

In addition to display of scada data on a large mimic panel, the operator has available all data via a graphical user interface. The system allows the operator to generate trend curves of all measured data together with computed data such as pipeline inventory and cumulative deliveries.

At normal operation, these features are sufficient to manage operations successfully.

At present, the transport control center daily handles 315 electronic messages via an x400 telecommunications system concerning dispatching data and production availability.

In addition, weekly and monthly delivery and production forecasts are received. This data flow may be significantly increased in case of renominations and altered production and transport availability.

To enhance accuracy and effectiveness, the AMHS was implemented. Written reports of gas orders received at the transport-control center via Telex or electronically are processed by the AMHS before gas-delivery instructions are transmitted to the production units in accordance with procedures and agreements.

Further, the system includes an operational gas account for checking gas volumes supplied and established swap agreements.

Pipeline modeling

Scientific Software Intercomp Inc., Houston, has developed and installed the pipeline-monitoring system (PMS) at the transport control center under the ABB contract.

PMS is the module which includes the fluid-flow simulation models: real-time model (RTM), look-ahead model (LAM), and predictive model (PM)

All PMS control functions are handled through "Sammi," a Windows-oriented graphical user environment that is used to enter and display PMS data.

A numerical model of the system which exhibits the generic properties of the transport network has been established. This model includes regulators, valves, bypass arrangements, pipeline topography, and ambient conditions including seabed temperatures.

The governing equations include conservation of momentum, energy, and mass expressed on a time-dependent form, one-dimensional in space. Closure of the equations is provided by the Boors equation of state.

Radial heat transfer is generally computed quasitransient by means of a constant overall heat-transfer coefficient. This is normally sufficient for slow transients.

A fully transient computation of the radial heat transfer, however, is an optional selection. This option facilitates calculations of blowdown, shutin, and start-ups.

The discretized equations are solved for each pipe segment employing an implicit finite-difference approach which is a variant of the socalled "collocation method" or "box scheme."

A noniterative direct matrix solution is used for a single pipe segment. For the network, a solution is obtained by solving the pipe connections and equipment device relationships in linearized form via a direct solution of a sparse matrix. The solution technique is stable and fast relative to other conventional approaches.

A generalized PVT (pressure-volume-temperature) package is provided as part of the PMS, based on the Boors equation of state. This equation employs interacting coefficients for the gas components that are tuned to experimental PVT data for adequate gas compositions.

Pressure drop in the pipeline segments is attributed to pipeline friction. For fully turbulent flow, the friction factor is calculated from the Colebrook equation.

A basic parameter in this equation is the internal surface roughness (e in the Colebrook equation) of the pipeline. In real operation, however, this roughness parameter is not invoked as an inherent property of the pipeline surface alone.

In the PMS models, the roughness parameter represents a calibration factor for the model and thus accommodates inaccuracies caused by ambient conditions, gas properties, instrument readings, and mechanical properties described in the transport system model.

In the RTM, the roughness parameter is tuned continuously to achieve satisfactory correlation between the measured and computed delivery pressures. In this manner, the model remains accurate and adapts itself to the actual conditions of the transport network.

PMS regularity has been improved significantly since installation in 1993. Fig. 5 [65,872 bytes] shows a regularity of approximately 98% in the first half of 1996.

The system's reliability can be attributed to improvement of quality tracking, data validation, bypass modeling of Draupner, and communication with "Sammi."

It should be recognized that successful operation of a sophisticated computer-based modeling system, such as PMS, requires close cooperation of operator personnel with the vendor through all phases of the contract and after installation.

Real-time model

In the present operation, 512 of the data signals received by the scada system are employed by the RTM which computes the operational condition in the network at positions between those measured.

Features generally read by the RTM include current operating state of the network, leak detection, quality tracking, instrument validation, scraper-pig tracking, survival-time analysis, and inventory analysis.

Experience has taught that RTM availability is primarily related to instrument accuracy and regularity. In case one of the basic signals is temporarily discarded or not received, a preset value is normally employed.

In case a coldstart of the model is required, the new system condition is predicted from the latest recorded system condition. The computed operational condition of the network will then gradually move towards the real operational condition.

Look-ahead model

Current RTM profiles are always used by the look-ahead model to project automatically the transient results at a frequency and duration specified by the user.

Display of such alarms as over/under pressure, insufficient line pack, or violation of gas-quality limits are provided through "Sammi." Under normal conditions, the operator will thus be unaware of the LAM operation. The most recently computed trends, however, may be retrieved for display and plotting.

Operators at the transport-control center see the following forecasting features of the LAM: gas composition in the system, pipeline pressure, gas inventory, and GCV, WI, H₂S, water, and hydrocarbon dew points.

Predictive model

Start conditions of the predictive model may either be the current RTM-generated profiles or a user-specified steadystate condition. Application area of the PM has proven to be versatile. So far, the PM has mainly been employed for engineering purposes including the following tasks:

• Preparation of operational procedures
• Design of new pipeline segments in the network
• Prediction of deliverability concerning possible new gas sales
• Analyses of test data to determine pipeline transport capacity
• Personnel training.

Accurate determination of pipeline transport capacity is important with respect to commitments of new gas sales and the associated transport requirements. Prediction of Zeepipe I's transport capacity was performed soon after it was put into service, even though the flow rate was less than half of the design capacity.

High-precision pressure transmitters were installed for permanent use. Analyses of these accurate data-pressure readings by use of the PM resulted in an upgraded transport capacity of 8%. This additional capacity has now been employed as a basis for new transport commitments which can be arranged earlier.

Prediction of "survival time" following shutdown of a terminal or a platform is important as a basis for optimization of operations. Survival time denotes the available time to continue the operations, for example, depacking of a pipeline after upstream shutdown, before reaching the minimum pressure limitation. This kind of transient behavior is well-suited for predictive modeling.

Fig. 6 [40,338 bytes] displays survival time of the deliveries in Zeebrugge following a shutdown at Sleipner. Each curve relates delivery rate and survival time to the initial pipeline packing before the shutdown. Provided initial pressure in Zeebrugge is 100 barg, the gas deliveries may continue undisturbed for another 22 hr at 15 MMscmd.

Model performance

PMS accuracy is as important as availability. Moreover, accuracy is governed by mathematical models, numerical methods, and system boundary conditions. Uncertainty of the latter relates to seasonal profiles of seabed temperatures and accuracy of measured data values.

Testing of PMS performance is fundamental to identifying and narrowing the uncertainties of predictions. Continuous computational analyses of system operations improve the operator's knowledge of the characteristics of the transport network.

Moreover, these analyses provide a means for PMS accuracy assessment. The established R&D program related to this field is to a large extent governed by the results of these analyses.

Analyses of pigging in Zeepipe I is one of these activities. Since flow into the pipeline and gas deliveries in Zeebrugge are literally constant, prediction of arrival time of a pig several days after launching at Sleipner is important.

These calculations are performed by the RTM based on current velocity profile. Experience so far is that arrival time for a pig is predicted within an accuracy of 2 hr when the pig has traveled for approximately 4 days. A slip factor of 1 is then employed by the model.

Changes in gas composition throughout the pipeline are tracked by the PMS. This allows the operator to forecast the feed stream composition to a terminal which will have sufficient time to prepare for the changes.

Further, contractual commitments require that the operator inform the gas buyer of significant changes of the gas properties (for example, composition, GCV, and WI).

Measured and computed variations of methane and ethane content in Zeebrugge for a 2-day period are shown in Fig. 7 [45,497 bytes]. It should be noticed that the computed gas composition in Zeepipe is based on gas chromatograph readings at Karstø Kollsnes, and Sleipner.

A relatively large variation in the CH₄ and C₂H₆ content appears as the flow from Kollsnes is shut down. It should be noted that this is a snapshot from the commissioning period of the Troll-Kollsnes production facilities.

Further, it can be learned from this figure that the composition gradients (fronts) are transported almost unattached through the pipelines. Neither numerical or physical diffusion of the gradients seem to be of significance.

Finally, only minor differences between the measured contents of CH₄ and C₂H₆ are displayed. These differences are chiefly attributed to the accuracy of the various chromatographs involved and coordination of sampling intervals.

Acknowledgment

The authors wish to thank Statoil for permission to publish this article.

The Authors