22nd North Sea Flow Measurement Workshop 2004

22nd International North Sea Flow MeasurementWorkshopTuesday 26 - Friday 29 October 2004

Ultrasonic Metering Training CourseMeasurement Uncertainty Training CourseMonday 25 October 2004

St Andrews, Fife, UK

Organised byNEL, EAST KILBRIDE, GLASGOW, UKBOOKING HOTLINE: +44 (0) 1355 272858

NORSK FORENING FOR OLJE OG GASSMLINGNORWEGIAN SOCIETY FOR OIL AND GAS MEASUREMENT

Monday 25 OctoberUltrasonic Metering Training Course 09:30 - 16:30

Measurement Uncertainty Training Course 09:30 - 16:30

Main Workshop Technical Programme

Day 1 - Tuesday 26 OctoberRegistration Period 12:00 - 12:30

Buffet Lunch 12:30 - 13:30

Welcome:Michael Valente, Managing Director, NEL 13:30 - 13:35

Keynote Address:Louise Kingham, Chief Executive, Energy Institute 13:35 - 13:50

Session 1 13:50 - 16:10Allocation Chairman: Richard Paton

Application of Phase-Behavior Models in ProductionAllocation SystemsWilliam Donkervoet, Shell Exploration & ProductionCompany, USARobert Webb, BP America Production Company, USA

Field Applications of Model-Based Multiphase FlowComputingge Rasmussen, FMC Kongsberg Subsea, Norway

Quantifying Financial Exposure Due to Meter UncertaintyUsing Monte Carlo SimulationClaudio Giglio, Advantica, UKLol Coughlan, Shell Expro, UK

Equity Exposure in Wet Gas Allocation MeteringDavid Stewart, NEL, UKMark Skelton, BP, UK

Coffee 16:10 - 16:40

Facilitated Discussion Session 16:40 - 17:40

Pre-Dinner Drink 18:45

Dinner 19:15

Day 2 - Wednesday 27 OctoberSession 2 0830 - 10:15Gas Metering 1 Chairman: Andy Jamieson

Tests of the V-Cone Flow Meter at Southwest ResearchInstitute and Utah State University in Accordance with theNew API 5.7 ProtocolBob Peters and Richard Steven, McCrometer Inc, USADarin George, E Bowles and M Nored, Southwest Research Institute, USA

An Assessment of the Impact of Contamination on OrificePlate Metering AccuracyMichael Pritchard and David Marshall, Advantica Ltd, UKJohn Wilson, National Grid Transco, UK

Dirty versus Clean Ultrasonic Flowmeter Results - An UpdateJohn Lansing, Daniel Measurement & Control, USATom Mooney, Daniel Measurement & Control, UK

Coffee 10:15 10:45

Session 3 10:45 - 12:30

Gas Metering 2 Chairman: Lol Coughlan

A Powerful New Diagnostic Tool for Transit Time UltrasonicMetersWilliam Freund, Jr and Klaus Zanker, Daniel Measurement& Control Inc, USA

Evaluation of Flow Conditioners - Ultrasonic MetersCombinationsBruno Delenne, Gaz de France, FranceGerard Mouton, GSO, FranceMichael Pritchard, Advantica, UKMichel Huppertz, Fluxys, BelgiumKristof Ciok, GASTRA, DenmarkAernout van den Heuvel, Gasunie Research, The NetherlandsTrond Folkestad, Norsk Hydro, NorwayDetlef Vieth, Ruhrgas, DenmarkFrancisco Lezuan, ENGAS S.A., SpainGuiseppe Marini, Snam Rete Gas, Italy

Coded Multiple Burst (CMB) Signal Processing Applied to Ultrasonic Flow Meters in Applications with High NoiseLevelsMarcel Vermeulen and Geeuwke de Boer, InstrometUltrasonics, The Netherlands

Lunch 12:30 - 13:30

Session 4 13:30 - 14:40

Liquid Metering Chairman: Douglas Griffin

Installation Effects and Diagnostic Interpretation Using theCaldon Ultrasonic MeterTerry Cousins, H Estrada and D Augenstein, Caldon Inc, USA

ALTOSONIC III - A Dedicated Three-Beam LiquidUltrasonic Flowmeter For Custody Transfer of SingleProductsJankees Hogendoorn, Andr Boer, Dick Laan and Helen Danen, KROHNE, The Netherlands

Session 5 14:45 - 17:35Flow Computing Chairman: Douglas Griffin

Oil & Gas Industry Can Now Benefit From Digital PlantArchitectureStuart Brown, Damon Ellender and Randy Page, EmersonProcess Management, Flow Computer Division, UK

Coffee 15:20 - 15:50

Chairman: Jim Ryan

Flow Measurement Asset Management in the 21st Century -How the Internet Will Play Its PartAndrew Webb, Ambrit Ltd, UK

Flow Computers and Control Systems - Interface orIntegrate?Ben Leach, Swinton Technology Ltd, UK

APIs Microprocessor Based Flowmeter Testing ProgrammeKenneth Elliott, Omni Flow Computers Inc, USA

Comfort Break 17:35 - 17:50

Facilitated Discussion Session 17:50 - 18:25

Dinner 19:30

Day 3 - Thursday 28 OctoberSession 6 08:30 - 10:15Learning Chairman: Alan Downing

Installation Effects on Venturi Tubes of Convergent Angle 10.5Michael Reader-Harris, Ronnie Rushworth and Jeff Gibson, NEL, UK

Determination of Hydrocarbon Dewpoint from On-line GasChromatographic AnalysisDouglas Pettigrew, Daniel Europe Ltd, UK

Overview of CFD Simulation of Orifice Plate FlowmetersNeil Barton, Michael Reader-Harris and Jeff Gibson, NEL, UK

Coffee 10:15 - 10:45

Poster & Manufacturers Sessions 10:45 - 12:30

Lunch 12:30 - 13:30

Session 7 13:30 - 15:15Coriolis Chairman: Eddie Spearman

Coriolis Mass Flow Meter Developments: Increasing TheRange of Applications in Oil & Gas Production andProcessingMichael Tombs and Manus Henry, University of Oxford, UKHoi Yeung, Cranfield University, UKRobbie Lansangan, Invensys Foxboro, UK

Two-Component Coriolis Measurement of Oil and Water atLow VelocitiesOle Andersen, Maersk Olie og Gas, DenmarkGary Miller, NEL, UK

First Tests and Experiences of Application of CoriolisMeters in Natural Gas Custody TransferHenk Riezebos, Gert van Essen and Andre van der Horn,Gasunie Research The Netherlands

Coffee 15:15 - 15:45

Session 8 15:45 - 18:05

Multiphase / Wet Gas Chairman: Trond Folkestad

Validation and Operational Experience of a Dualstream IIWet Gas Meter in a Subsea Allocation Application on theStatoil Mikkel FieldAlan Downing and Paul Daniel, Solartron ISA, UK,Andre Jacobsen and Harald Denstad, Statoil, Norway

Wet Gas Flow Measurements with Mixtures of Natural Gas,Hydrocarbon Liquids and WaterJeff Savidge, CEESI, USA

Operational Experience of Smartvent Wet Gas Metering atBintangRick de Leeuw, Petrotech B.V., The NetherlandsC Nilsson and B Dybdahl, Petrotech ASA, Norway

Characterization of the Turbulence Properties of Wet GasFlow In a V-Cone Meter with Neural NetsHaluk Toral and Shiqian Cai, Petroleum Software Ltd, UK,Bob Peters and Richard Steven, McCrometer, USA

Closing Address:Trond Folkestad, Chairman, Norwegian OrganisingCommittee 18:05 18:20

Cocktail Reception 19:30

Workshop Dinner 20:00

Day 4 - Friday 29 OctoberBuses Leave For Airport 09:00

Optional Golf Tournament 09:30 13:30

Paper 1.1 Application of Phase Behaviour Models in

Production Allocation System

Robert Webb BP Exploration and Production Inc.

William Donkervoet

Shell Exploration & Production Company

North Sea Flow Measurement Workshop 2629 October 2004

1

APPLICATION OF PHASE BEHAVIOR MODELS IN PRODUCTION ALLOCATION SYSTEMS

Mr. Robert A. Webb, BP Exploration and Production Inc.

Mr. William Donkervoet, Shell Exploration & Production Company

1 INTRODUCTION

The current and future scenarios for production in the Gulf of Mexico (GoM) are such that application of phase-behavior models within the allocation system is appropriate. Furthermore, the reasons supporting this application are likely to exist in other major production regions throughout the world. The GoM has most recently ventured into ultra deepwater production. Subsequently, development costs (drilling, topsides facilities, and export infrastructures) have risen in some cases to nearly three billion USD. These investment levels are driving a case to tie-back subsea wells from farther distances to new or existing production facilities. The result is a hub like structure with a minimal number of floating production facilities serving many different tiebacks from the surrounding area.

The hub scenario introduces an order of magnitude greater complexity. Specific to measurement and allocation, is that the various tiebacks inevitably contain wells with different owners and different tax rates. This will be referred to as ownership disparity. In the past equal ownership and tax rates was more often the case. This equated exposure to only that of the more accurate custody transfer measurement. Ownership disparity exposes each owner to the full measurement error over the entire production measurement system. Also note that tax rate differences equates to an ownership difference for the taxing authority.

Fig. 1 Hub Scenario

The hub scenario makes it more likely that a certain ownership disparity will exist. However, this is only part of the story. With the hub scenario a second disparity is likely to exist, in that fluid types will also likely vary. The more tiebacks that exist, the more likely it is that gas, gas/condensate, and black oil wells will eventually commingle in the same process. This is referred to as fluid dissimilarity. Ownership disparity and fluid dissimilarity taken together, compound the measurement error exposure for all parties. The presence of ownership disparity makes relevant the real metrology problem which is phase behavior differences between dissimilar fluids. Once an ownership disparity exists, the full impact of fluid dissimilarity is encountered by the allocation system. A typical measurement error exposure of one or two percent may increase to as high as five to seven percent due to the phase behavior of the dissimilar commingled fluids. A way to reduce the measurement error is to utilize phase-behavior models within the allocation system to better predict the final quantities that determines the associated revenue.

This paper will explain the basic problem of phase changes in production processes and how they affect the allocation system. The typical past methods for dealing with the problem (non-modeled approach) will be critically examined. A phase-behavior model approach will be presented along with optimization techniques needed to effectively apply the solution. Lastly, advantages, disadvantages and possible future improvements will be highlighted. The

PROCESSINGHUB

GASWELLS

OILWELLS

GAS / CONDENSATEWELLS

EXPORTPIPELINES


2

realities that have heightened phase-behavior as a measurement issue are here to stay in the GoM and will likely become evident in other major production regions throughout the world.

2 PROBLEM

The problem starts with the fact that commingled production streams must be measured prior to commingling in order to perform a proper allocation. However, once commingled, the streams experience a certain production process where the temperature and pressure conditions vary greatly from the original measurement conditions. Within the process the fluids undergo phase changes between gas and liquid phases.

PROCESSVESSEL(S)

STREAM 2SEPARATOR

HIGHERSTAGE

COMPRESSION

DEHYDRATION

LOWERSTAGE

COMPRESSION

LIQUIDMETER

GASMETER

STREAM 1SEPARATOR

LIQUIDMETER

GASMETER

GASEXPORT

INTRA-PROCESSRECYCLE

LIQUIDPUMPS

GASMETER

LIQUIDMETER LIQUID

EXPORT

Fig. 2 Typical Production Process

Therefore, the measured quantities of gas and liquid at the start of the process will not equal the measured quantities of gas and liquid at the end of the process with respective to gas and liquid independently. While the total mass of the gas and liquid combined remains constant, there is an exchange of liquid to gas and gas to liquid during the process. This seems inconsequential given that the total mass is constant. However, since measurement and allocation systems are the basis of revenue assignment between all the affected parties (i.e. working interest owners, taxing authorities, etc.) and the real value of hydrocarbons in the form of gas (natural gas) and liquid (crude oil) are seldom equal, each commingled streams correct portion of the final gas and liquid streams must be known in order to ensure equity in the fiscal assignments.

A system must be put in place to accurately predict and assign the final quantities of gas and liquid independently to the individual commingled streams. This accounts for gas to liquid and liquid to gas phase changes within the process. Complicating the situation even more is the interaction of individual components within the streams. A stream undergoing the process by itself will experience a certain phase change. That same stream undergoing the process commingled with another stream will experience a different phase change. The two streams (assuming dissimilarity in their respective gas and liquid compositions) will interact and influence the phase change of each other.


3

2.1 The Previous Gulf of Mexico Approach

Typically the problem is approached by a sampling and analysis techniques. The quantity (volume) and quality (composition) of each stream is determined at the beginning of the process independently. This is normally accomplished by independent gas and liquid metering and sampling systems situated at the outlets of separation vessels. The relative pressure and temperature must also be acquired. Once the composition is known from a laboratory analysis, an Equation of State (EoS) is used to predict the phase changes from the measured pressure and temperature conditions to the end-of-process conditions. The appropriate corrections to the liquid and gas quantities are then applied.

In the GoM the above approaches has normally entailed performing flash and shrink calculations on the liquid portion of the stream and assuming the gas portion of the stream remains as gas throughout the process. Shrinkage in these terms is the ratio of the liquid quantity at the beginning of the process to the liquid quantity at the end of the process, and the flash is the quantity of gas that is predicted to evolve from the liquid during the process. The EoS approach can also be used for the gas portion of the process but it is normally not used. The reason for this is that the process includes intra-process recycled streams (see Fig. 2). An intra-process recycled stream is where liquids condensing from the gas (normally collected before and after compression) are diverted back to an upstream liquid process point. These liquids are predominately propane, butane, pentane, and hexane mixtures. In the process these liquids are exposed to a lower pressure and normally would return to the gaseous phase. However, when these liquids mix with longer-chain molecules (black oil) a portion of the liquids remain in liquid phase. As one can see it is difficult to make a simple application of an EoS in this iterative fashion. One must not only predict the gas to liquid phase changes, but also predict the residual gas to liquid quantities. That is to say the quantity of liquid from the gas stream that remains as liquid to the end of the process must be calculated, discounting the liquids that re-gasify before the end of the process. The typical approach in the GoM and possibly other areas has been to ignore this issue. Furthermore, the API standard on allocation measurement [1] only mentions an EoS approach for flash calculations of gas evolving from liquid. The liquids condensing from gas are not mentioned as a quantity measurement issue but only as a quality issue. While the liquid corrections using shrink and flash calculations are roughly appropriate, ignoring any gas to liquid phase change carries a large impact. With the ever increasing hub-based production quantities in the GoM and the realities of fluid dissimilarity, a more rigorous approach is warranted.

2.2 Financial Significance

It can be easily seen that the phase changes within the process affect the final answer. And, that due to varying ownership or tax rate for the commingled streams the revenue distribution will also be affected. But the real question concerns the magnitude of the effect. Since no measurement system or technique is perfect, the question becomes how much imperfection is acceptable. Or, more precisely stated, what are the random uncertainties and possible biases associated with the typical approach in the context of the current and expected future fluid compositions? If the errors are small they may be ignored and assumed to be a cost of doing business. However, as this paper will show, the errors in the context of fiscal hydrocarbon measurement are not small. Furthermore, the errors are bias in nature and when applied to the large production quantities on GoM hubs, result in significant financial impact.

In order to demonstrate the significance a comparison is made between the typical approach explained above and a new approach using a phase-behavior model. The model allows the commingled streams to be viewed simultaneously and deciphers the intra-process recycled streams. The comparison is based on a typical GoM deepwater hub accommodating black oil, gas condensate and gas wells. The oil and gas prices are set at US $30 per barrel and US $4 per MMBTU, respectively. Please note if the gas price on a barrel of oil equivalent (BOE) basis goes above the oil price the gas to liquid economics depicted below will be reversed.


4

Table 1 shows the relative theoretical volume of oil and gas for the typical approach, referred to as the non-modeled approach, followed by the same calculation using the model. The theoretical quantity is the quantity in terms of energy for the gas and standard volume for the oil, as measured at the start of the process but in terms of the pressure and temperature at the end of the process. Since the model and the non-model approaches begin with the same input quantities and their differences lie in the proportioning of the final quantities, no greater or less mass is generated. Thus, ultimately the net difference is zero. And, if all ownerships remained equal, no party would suffer financially. However as demonstrated above, this is seldom the case.

Table 1 Non-Modeled versus Modeled Approach

Stream Fluid Type

MeasuredQuantities

TheoreticalQuantities

AllocatedQuantities

Total Value

Typical Approach (Non-Modeled) Oil (Bbls.) 1,800,000 1,706,966 1,733,314 $ 51,999,411Stream A

Black Oil Gas (MMBTU) 737,369 892,454 837,048 3,348,194 Oil (Bbls.) 900,000 733,300 744,619 22,338,561 Stream B

Gas Condensate Gas (MMBTU) 1,769,360 2,364,437 2,217,647 8,870,587 Oil (Bbls.) 120,000 86,095 87,424 2,622,709 Stream C

Gas Gas (MMBTU) 2,076,610 2,267,232 2,126,477 8,505,908

Modeled Approach Oil (Bbls.) 1,800,000 1,710,000 1,711,651 51,349,533 Stream A

Black Oil Gas (MMBTU) 737,369 857,130 856,375 3,425,499 Oil (Bbls.) 900,000 760,651 761,385 22,841,561 Stream B

Gas Condensate Gas (MMBTU) 1,769,360 2,096,955 2,095,108 8,380,433 Oil (Bbls.) 120,000 92,231 92,320 2,769,588 Stream C

Gas Gas (MMBTU) 2,076,610 2,231,655 2,229,689 8,918,756

Difference Modeled minus Non-Modeled Oil (Bbls.) 3,035 -21,663 (649,879)Stream A

Black Oil Gas (MMBTU) -35,324 19,326 77,305 Oil (Bbls.) 27,351 16,767 503,000 Stream B

Gas Condensate Gas (MMBTU) -267,481 -122,538 (490,153)Oil (Bbls.) 6,136 4,896 146,879 Stream C

Gas Gas (MMBTU) -35,578 103,212 412,848

Stream A (572,574)

Stream B 12,847

Net difference in revenue

Stream C 559,727

Total Net difference 0

Refer to the Appendix for a listing of the compositions used within this paper for streams A, B, and C.


5

2.3 Sensitivity to the Process

In addition to the problem created by commingling dissimilar fluid types, the configuration of the process itself constitutes another problem. Figure 3 illustrates the process used for calculations in this paper. Based on how the process is configured and where liquids are removed, or where in the process rich and lean gases commingle, a greater variance in the final liquid and gas quantities is encountered. To illustrate this effect a close-up view of one process stream is shown in Figure 4 with the normal separation vessel package. In Figure 5, the above process is changed slightly where one separation vessel is eliminated and the associated gas stream is connected at the discharge of the previous compressor stage.

Fig. 4 Normal Separation Vessel Package (same as Fig. 3)

Fig. 3 Process Flow Diagram


6

Fig. 5 Reduced Separation Vessel Package

The process configuration in Figure 5 shows an alternate design normally used for debottlenecking or to reduce weight (capital cost). Note that there is a significant difference in the quantity of condensed liquids. In the Figure 4 design more condensed liquids ultimately exit the process as liquids. Alternatively, in Figure 5, a significant amount of heavier components that would have become liquid, remain in the gas phase and exit the process as gas. Refer to the Condensed Liquids data box in each figure. The Figure 4 design yields 16,830 barrels while the Figure 5 design yields only 3,360 barrels.

The non-modeled approach would treat both these process configurations exactly the same in that the initial fluid pressure and temperature are equal and they have the same process endpoint pressure. However, as can be seen a significant difference exists due to the configuration of the process itself. Simple EoS (i.e. shrink and flash calculations) cannot differentiate process changes beyond a simple pressure and temperature change. The EoS inability to depict the actual process configuration creates an un-resolvable imbalance. On simple systems with low fluid dissimilarity this imbalance may be relatively small, but on complex processes with greatly differing fluid types, this imbalance becomes critical. Additionally, there is a significant financial impact in that gas and oil prices are seldom equivalent compounded by greater ownership disparity.

2.4 Dissimilar Fluids

While the inability to depict the process complexity is problematic, dissimilarity in the commingled fluid types presents a greater problem. Since the fluid types can change over time and even within an allocation period, the non-modeled approach can give varied and unrepeatable results. The most direct way to examine this effect is by comparing the shrink factors and flash gas quantities using the non-modeled and modeled approaches. The shrink factor is the ratio of the liquid volume at the end of the process (P2,T2) over the volume at the beginning of the process (P1,T1). The volume at P1,T1 is the measured volume at the separator and volume at P2,T2 is the theoretical volume at the end of the process. For the example streams A, B, and C following are the results of the model and non-modeled approaches.

Table 2 - Shrink Factor Comparison

Stream Shrink Factor Non-Modeled

Shrink Factor Modeled

Difference

Stream A 0.9483 0.9500 0.0017

Stream B 0.8148 0.8452 0.0304

Stream C 0.7175 0.7686 0.0511


7

The simple EoS approach assumes more flashing of liquid and thus a lower ultimate liquid volume. Conversely, providing for conservation of mass, the EoS over predicts the flash gas quantity. And, due to an over prediction of heavier components in gaseous phase, the EoS predicts a higher flash gas energy content as shown in Table 3.

Table 3 Flash Gas Comparison

Flash gas Non-Modeled Flash gas Modeled Stream

Quantity Quality (BTU/SCF)

Quantity Quality (BTU/SCF)

Difference (MMBTU)

Stream A 86,113 1,801 80,984 1,479 155,086

Stream B 276,655 2,151 240,902 1,360 595,077

Stream C 76,610 1,594 68,839 1,258 122,154

3 SOLUTION

A solution to the problem is to utilize a phase-behavior model to predict the phase changes throughout the process. Generically these are called Process Simulation Models (PSM) and will be referred to as PSM in this paper. The PSM reduces the uncertainty of the shrinkage predictions, is comprehensive enough to resolve the intra-process recycled streams, and accounts for the interaction between the fluid streams when they are processed simultaneously. The PSM still uses a base EoS. Generally, an EoS such as Peng-Robinson or Soave-Redlich-Kwong (SRK) which are cubic equations capable of modeling vapor liquid equilibrium [2] is used. And, the EoS must still be fit for the particular application in regards to the hydrocarbons and their proximity to critical regions, handling the phase-behavior of polar systems, etc. Different EoS may give rather differing results, especially in terms of the crucial compressibility factor Zc resulting in different liquid densities. However, provided the EoS chosen uses the same alpha function and mixing rules, the ability of one EoS to model the process should be comparable to the next [3]. Therefore, the PSM offers a distinct improvement in reducing the biases created by an individual stream application of an EoS. However, even with this improvement, there are various application problems which must be accounted for. In order to properly apply the results in the allocation system an optimization process should be followed. Otherwise certain biases will remain in the allocation that cannot be resolved.

3.1 Application of the PSM

Whenever a PSM is used in an allocation system, there exists a particular problem. The PSM can only give the correct answer when all inputs are considered simultaneously. However, the theoretical quantity for each input (or commingled stream) is needed on an individual basis in order to perform the allocation. Therefore a particular method of application must be employed to divide the more precise quantity considering all the commingled streams simultaneously into individual quantities relative to each commingled stream.

Three distinctive approaches are used in combination to resolve the problem. First, all the inputs are taking simultaneously to develop the best estimation of the theoretical quantity (note that normally two theoretical quantities are determined namely one for gas and one for liquid). These quantities represent the best estimation of the amount gas and liquid available at the end of the process based on the input quantities, their respective compositional make-up, and the various pressure and temperature changes relative to the process. This calculation is referred to as the combined calculation where all inputs are taken in combination. It is the results of the combined run that eventually are used on a stream-by-stream basis to perform the allocation. Let q stand for the quantity determined by the combined calculation.


8

Second, a theoretical quantity for each input (i) must be determined. The best approach is the by difference method. For every input stream a calculation is performed using all the inputs except the one input in question. Let qxi stand for the quantity determined excluding the ith input. Then:

xii qqq = (1)

This calculation (q double prime) is called a by-difference quantity. The by-difference quantity is normally very close to the proper or optimal quantity for an individual stream. However, when all the by-difference quantities are summed, the summation does not equal the results of the combined calculation. A differential (q) exists as follows:

= n iqqq 1 (2)

The differential represents a small, but noticeable, difference in the PSM calculations due to the interaction of the different fluid types. The dissimilarity of the fluid properties causes a change in the result each time the PSM calculation is performed. Removing one stream from the combination constitutes a change in the overall combined fluid properties. The only exception to this rule is the rare case where the fluid properties from each combined stream are identical. This is never really the case, but often is nearly so. In fact, if the stream compositions are nearly all the same (e.g. all black oil, or all gas condensate, etc.) the need to perform a by-difference calculation or even a combined calculation would not be necessary. However, in todays complex hub arrangement, one can expect a significant dissimilarity in fluid types to exist.

3.2 Determination of the Differential

The differential between the combined calculation and the by-difference calculation must be incorporated back into the allocation system. This is accomplished by determining the projected theoretical quantity when an input stream is considered by itself (i.e. no commingling). Let q represent the individual calculation. Based on the individual calculation (q) in comparison to the by-difference calculations (q) an assignment of the differential is made so that the sum of the quantities assigned to each input will balance to the total quantity (q) determined in the combined calculation. The process of assigning the differential is referred to as an optimization since it optimizes the assigned quantities to fit the combined calculation, which ultimately is the correct answer.

There is more than one approach to this optimization process. A simple approach is to assign the differential proportionally based on the by-difference calculation. While this method distributes the differential, it is not the most optimized. Since some of the input streams changed more than others when considered simultaneously in the combined calculation, it is unfair to assume they all changed proportionately to their respective original measured qualities.

Another way to distribute the differential is based on their respective change between their by-difference calculation (q) and their individual calculation (q). The following formula represents this relationship:

( )

( )

=

nin

ii

iii qq

qqqqq

1

1

(3)

3.3 Divergence of Differentials

While this method appears to be more consistent than a simple proportional distribution it has a fatal flaw. Since some of the changes are positive and some are negative the distribution may diverge and result in larger and larger positive and negative numbers that when combined equate to the differential, but in very erroneous ways. Ultimately this method


9

produces an unlimited number of solutions each diverging from a single optimal solution. Table 4 shown below, provides two cases to illustrate the divergence of the differentials.

Case one is a typical case. The sum of the differentials (i) as always equals the difference between the combine quantity (q) and the sum of the by-difference quantities (qi) as per Equation (2). The divergence is evident in that the sum of the absolute values of the differential is greater than difference between the combine quantity (q) and the sum of the by-difference quantities (qi). For case one the divergence is only about 6%.

In the extreme case (Case 2), however, the divergence is over 45 % of the original difference value. It is illogical to expect the magnitude of the differentials as a whole (sum) to be much greater than the original difference. Each assigned differential is supposedly only a portion of that difference. Therefore, in a perfect world the magnitude of the differentials (i.e. their absolute values) will equate to the original difference. The fact that some input streams change in a positive direction while others change in a negative direction keeps this from being the case. But this difference in mathematical sign, positive or negative, is also a prime consideration in the solution.

Table 4 Divergence of Assigned Differentials (all quantity values in barrels)

Input

iq By-Difference

iq Individual

iq Difference

ii qq Differential

iq

CASE1 - Typical (same as streams used in Table 1)

Stream A 1,710,000.19 1,712,668.35 2,668.15 1,264.55 Stream B 760,651.02 763,357.10 2,706.08 1,282.53 Stream C 92,230.54 92,076.96 -153.58 -72.79 Sum 2,562,881.76 2,568,102.41 5,220.65 2,474.29 Combined ( q ) 2,565,356.05 Sum of absolute values 2,619.86 Difference iqq 2,474.29 Divergence ( )

i

ii

qqq

0.059

CASE 2 Extreme (uses only liquid streams from Table 2)

Stream A 1,708,027.46 1,719,914.73 11,887.27 950.76 Stream B 751,729.17 752,818.25 1,089.08 87.11 Stream C 91,485.71 89,086.16 -2,399.55 -191.92 Sum 2,551,242.34 2,561,819.14 10,576.80 845.95 Combined ( q ) 2,552,088.29 Sum of absolute values 1,229.79 Difference iqq 845.95 Divergence ( )

i

ii

qqq

0.454

3.4 Optimization of Differential

The differential created is distinctively either positive or negative. While, incorporated into the differential is any offsetting positive and negative changes, it is fair to assume that the resulting positive or negative differential was created more by those streams changing in the same (positive or negative) direction as the differential. Keep in mind that the differential represents a net positive or negative change. This is similar to distributing the system imbalance within a measurement system utilizing Uncertainty Based Allocation (UBA) [4]. However, unlike UBA where all inputs are assume to have additively created the imbalance, in this case a means to determine if an input contributed positively or negatively is possible.


10

Therefore, a more optimal approach is to divide this net negative or positive differential (q) among only the input stream demonstrating a like positive or negative change between the individual calculation (q) and the by-difference calculation (q). These particular inputs are designated as participating (part) and the other input streams as non-participating (non-part). The participating and non-participating differentials are defined such that:

n

parti qq1

, (4)

and;

0, partnoniq (5)

When considering only the participating input streams and considering the relative change between the individual and by-difference calculations, the following equation emerges as the assigned differential:

( )

( )

=

qqq

qqq partnpart ii

iiparti ,

,1

, (6)

Using the same scenario as above Table 5 shows the results of optimizing the differential according to equation (6).

Table 5 Optimized Differentials (all quaintly values in barrels)

Input By-Difference q

Individual q

Difference [ ]qq

Part. Differential

iq

Stream A 1,708,027 1,719,915 11,887

775

Stream B 751,729 752,818 1,089 71

Stream C 91,486 89,086 -2,400

0

Summation 2,551,242 2,561,819 10,577 846

Combined ( q ) 2,552,088

Difference iqq 846

3.5 Mass versus Volume

A fair question at this point is why a mass based systems is not used. The mass of the streams whether in the form of gas or liquid at the beginning of the process will equal the mass of the combined stream at the end. This is true. Therefore, if the compositions of the streams are known at the beginning of the process, and likewise known at the end of the process, then simple mass balance of each component seems to resolve the problem. However, two realities must be dealt with. First, the rudimentary measurement of the inflow and outflow is based on volume. In the GoM the typical gas meter is an orifice meter and the flow equation can be easily converted to yield mass terms. The liquid meters, however, are typically turbine meters where a mass registration would require integration of the flowing liquid density. Therefore, the conversion to mass, at least on the liquid metering systems, represents another direct multiplier in the uncertainty, i.e. flowing density. However, this is not the main reason for using the volume based system.


11

The main reason mass is not used is that one export stream is inherently based on volume terms. The crude oil business in the US commercially operates on a volume basis. Since ultimately the purpose of the allocation system is to distribute the revenue, eventually a conversion to a volume basis is needed. It is possible to maintain a mass component based system and make the conversion to standard volume (barrels at standard temperature and pressure) at the end of the theoretical quantity calculation. To suggest this approach is a fair debate point, since it may simplify the process. It would be a very strong debate point if the base measurement of the liquid were in the form of mass such as with a direct mass measurement like Coriolis meters.

Overall, however, approaching the problem from a mass perspective will not diminish the need to use a PSM in the allocation process. The differential between the combined and the by-difference calculations will still exist and will need to be distributed among the inflow streams. While the total mass in will always equal the total mass out, the division of the mass outflow between gas and liquid phase will change. That is, some mass will change (predicted via the PSM) from gas to liquid or visa versa, when the streams are viewed individually as compared to when viewed in combination. Considering these issues, however, the mass question is still a valid point and should be studied for future implementation in the US metrology community.

4 ADVANTAGE, DISADVATAGES AND FUTURE IMPROVEMENTS

4.1 ADVANTAGES

Equity Assurance: Overall the primary advantage is equity assurance in the assigned theoretical quantities. This improvement has long-reaching effects. The model approach reduces the measurement loss risk created by second generation tiebacks. Whenever a deepwater hub is installed, there exist a certain number of original tiebacks. The fluid types of the original tie-backs are often the same and thus dissimilar fluid type discrepancies are less of a problem. However, when new tiebacks seek to join an existing hub structure, a dichotomy is created. The new tieback may contain a different set of owners and a different fluid type. Often the second generation tiebacks are gas wells since gas can travel a longer distance. The arrival of a second generation tieback with a dissimilar fluid type will increase the measurement risk, either for itself or the original tiebacks, or possible for both depending on the scenario. The use of a PSM minimizes these effects. This greatly helps in the commercial viability of new production that needs to occupy the ullage created over time on the existing deepwater hubs.

Balance Performance: As highlighted in section 2.4 above, the PSM improves the system balance when viewed as a liquid and gas balance independently. Since the gas and liquid measurement systems within the process operate somewhat independently, monitoring their respective material balance independently is an advantage over a simple mass balance for the entire gas-liquid combined systems. The overall measurement system is better controlled and mis-measurements are more easily detected.

Fuel Allocation: A large secondary advantage offered by a PSM is the capability of predicting energy requirements for compression and pumping equipment. Normally the most detailed portion of the allocation systems is the fuel assignment. Since each stream requires a different amount of fuel for processing, the fuel assignment must be very detailed in order to allocate the consumed gas properly. Calculating a proper fuel allocation is very tedious. The PSM however, when calculating a stream on an individual basis, can predict the energy required for compression (the major fuel consumption need). This becomes a theoretical fuel and is called the primary fuel. The fuel for smaller equipment, crew quarters, light, etc. is called the residual fuel and is taken as the difference between the total measured fuel gas and the sum of the theoretical fuel calculated by the PSM. The residual fuel is normally small compared to the primary fuel and is allocated on a general produced volume basis. This simplifies the fuel assignment a great deal.


12

Sampling: The main advantage here is a general reduction in sampling and a simplification of the analytical process. Previously, non-modeled approaches required a determination of the shrinkage and flash factors. This is extremely difficult and costly to do precisely. The best approach has always been to perform a full PVT (pressure-volume-temperature) analysis which can range up to US $3,000 per test. Short-cuts have also been developed to avoid a full PVT analysis but this normally involves an EoS and detailed compositional analysis. These short-cut tests have traditionally been hampered by high random uncertainty, resulting in degradation in the confidence level in their results. The short-cut test cost about US $1,000 per test. By utilizing a modeled approach, only the composition of the liquid and gas is needed to be known. As a comparison to past practice, the simple composition tests are more easily performed which results in better repeatability and reproducibility and thus instill higher confidence. Composition tests generally cost less the more complex tests.

Furthermore, use of the model can validate samples. Whenever a liquid or gas sample composition is determined, the model is used along with the pressure and temperature of the associated separation vessel to check the sample. If the model indicates that a liquid sample does not exist fully as a liquid, or visa versa for a gas sample, the samples are disqualified. Normally this indicates that some free gas or free liquids were inadvertently introduced to the sample. Overall, however, the greatest sampling advantage found by using phase-behavior modeling, is that the laboratory analysis is simplified. The laboratories can concentrate on what they do best; compositional analysis. Shrinkage and flash analysis which are by far the highest variables in the measurement system are greatly reduced through the proper application of a PSM.

4.2 DISADVANTAGES

Engineering Oversight: The greatest disadvantage to using a PSM within the allocation process is the depth of engineering oversight needed. Previously, in the GoM measurement data was passed to hydrocarbon accounting personnel for processing. That was generally the end of the story. The use of PSM creates an additional step in the allocation process. The process pressure and temperature, and fluid compositional data must be validated prior to executing the PSM logic. Additionally, abnormal operations must be viewed critically against the PSM. The individuals needed to perform this work are either process engineers or under the direction of process engineers.

Intra-Process Points: There is also a greater dependence on pressure and temperature data from within the process (i.e. not just at the metering points). Often pressure and temperature data not associated with metering systems is less scrutinized. Therefore, errors may be made by faulty instruments that, being not critical to day-to-day operations, may be overlooked.

Individual Routines: To the authors knowledge no known vendor markets allocation software systems where a PSM model is integral to the routine. However, even if this were available, it would likely not be attractive, since it would make future revision to the PSM software very difficult. And due to the individualistic nature of the PSM, it makes it almost impossible to duplicate or standardize an allocation routine. However, it is the opinion of the authors that duplicating allocation routines between various operations is dangerous and should be avoided. Most operations are different enough to warrant individually developed allocation routines.

4.3 FUTURE IMPROVEMENTS

Standardization: One primary future improvement is the standardization of PSM selection and application. Which EoS is chosen, and how the PSM is formulated should follow a set of general principles. Generally little is known or standardized in this area. Furthermore, it is known that each EoS may yield different results, especially in predicting the behavior of long-chain molecules [5].


13

Auditability/Security: Another area for improvement is auditability. While normally the pressure and temperature data is well documented, and the PSM itself is documented, during times of the abnormal operation some critical decisions must be made. Integrating a system to provide and audit trail and thus security on necessary process data changes, will improve the accounting aspects of using a PSM.

5 CONCLUSION

Based on the direction of large oil and gas operations, disparity of ownership together with dissimilarity of fluids is a continuing reality. The consequences when applied to oil and gas valuation make the use of phase-behavior models in production allocation systems advantageous. The advantages out-weigh the disadvantages by an over-whelming amount. Continued use of non-modeled systems introduce biases that over the life of a production facility will likely disadvantage one party over another well beyond the capital cost required to establish and operated the PSM. Improvements within the PSM application, especially improvements in the EoS applications and reproducibility between competing equations, should be a primary goal of the metrology community within the oil and gas industry. In the authors opinion, the industry should move quickly to recognize the practice and standardize the selection and application of the PSM. In the future the PSM should be considered as basic component of most measurement and allocation systems.

6 NOTATION

q stands for a quantity predicted by the PSM considering all input stream simultaneously

qxi stands for a quantity predicted by the PSM considering all input stream simultaneously except the ith stream

qi stands for the by-difference quantity predicted by the PSM (see equation 1)

q stands for the total differential between the PSM predicted quantity considering all stream simultaneously and the sum of the by-difference

qi stands for the portion of the differential assignable to the ith stream

qi,part stands for the portion of the differential assignable to the ith (participating) stream, where a like change between individual and by-difference calculations either positive or negative as compared to the differential either positive or negative, constitutes participation

qi,non-part stands for the portion of the differential assignable to the ith (non-participating) stream, which is always equal to zero.

7 REFERENCES

[1] API Manual of Petroleum Measurement Standards, Chapter 20, Section One, Allocation Measurement, First Edition, September 1993

[2] Suppes, G.J., year unknown. Selecting Thermodynamic Models for Process Simulation of Organic VEL and LLE Systems, Department of Chemical Engineering, The University of Missouri-Columbia. Available via online search at the following source: http://students.aiche.org/pdfs/thermodynamics.pdf

[3] Twu, C.H., Sim, W. D., Tassone, V., Getting a Handle on Advanced Cubic Equations of State, Chemical Engineering Process (CEP) Magazine, Vol. 98, (November, 2002): 58-65.

[4] Webb, R., Letton, W., Basil, M., Determination of Measurement Uncertainty for the Purpose of Wet Gas Hydrocarbon Allocation, North Sea Flow Measurement Workshop, September, 22-25, 2002.

[5] Ghosh, A., Chapman, W.G., French, R. N., Gas Solubility in Hydrocarbons a SAFT-based Approach, Fluid Phase Equilibrium, Vol. 209 (2003): 229-243.


14

8 ACKNOWLEGEMENTS

The following individuals contributed to the technological development of the various techniques described above. The authors acknowledge their contribution and give them many thanks for their assistance.

Mr. Tom Hurstell Innovative Technical Services (New Orleans)

Mr. Chad Simonton - Innovative Technical Services (New Orleans)

Mr. Charles Marth Metco Services (Houston)

Mr. David Escobar - Metco Services (Houston)

Mr. Winsor Letton, PhD. Letton-Hall Group (Houston)

Mrs. Perla Gonzalez Lerma BP Exploration and Production (Houston)

Mr. Roger Leach - BP Exploration and Production (Houston)


15

APPENDIX

Following is the gas and liquid composition of the three streams (A, B, and C) used as an example within this paper.

Table 6 Stream Compositions

Gas Liquid A B C A B C Methane 0.8906 0.8954 0.9647 0.0642 0.2129 0.4346 Ethane 0.0194 0.0435 0.0109 0.0152 0.0299 0.0190 Propane 0.0384 0.0303 0.0112 0.0285 0.0368 0.0360 i-Butane 0.0050 0.0069 0.0021 0.0108 0.0233 0.0248 n-Butane 0.0188 0.0093 0.0060 0.0312 0.0376 0.0293 i-Pentane 0.0161 0.0038 0.0019 0.0171 0.0280 0.0303 n-Pentane 0.0083 0.0042 0.0024 0.0220 0.0370 0.0287 n-Hexane 0.0016 0.0041 0.0003 0.0248 0.0704 0.0373 n-Heptane 0.0004 0.0019 0.0002 0.0311 0.1372 0.0843 n-Octane 0.0004 0.0003 0.0001 0.0496 0.1628 0.0575 n-Nonane 0.0004 0.0002 0.0000 0.0313 0.0699 0.0327 n-Decane 0.0007 0.0002 0.0001 0.0319 0.0579 0.0362 n-C11 0.0000 0.0000 0.0000 0.0383 0.0345 0.0292 n-C12 0.0000 0.0000 0.0000 0.0392 0.0187 0.0218 n-C13 0.0000 0.0000 0.0000 0.0441 0.0130 0.0194 n-C14 0.0000 0.0000 0.0000 0.0456 0.0073 0.0153 n-C15 0.0000 0.0000 0.0000 0.0444 0.0050 0.0133 n-C16 0.0000 0.0000 0.0000 0.0369 0.0026 0.0094 n-C17 0.0000 0.0000 0.0000 0.0366 0.0019 0.0083 n-C18 0.0000 0.0000 0.0000 0.0317 0.0012 0.0069 n-C19 0.0000 0.0000 0.0000 0.0311 0.0008 0.0056 n-C20 0.0000 0.0000 0.0000 0.0264 0.0005 0.0043 n-C21 0.0000 0.0000 0.0000 0.0233 0.0003 0.0035 n-C22 0.0000 0.0000 0.0000 0.0200 0.0002 0.0027 n-C23 0.0000 0.0000 0.0000 0.0191 0.0002 0.0025 n-C24 0.0000 0.0000 0.0000 0.0163 0.0002 0.0020 n-C25 0.0000 0.0000 0.0000 0.0169 0.0016 0.0015 n-C26 0.0000 0.0000 0.0000 0.0143 0.0016 0.0014 n-C27 0.0000 0.0000 0.0000 0.0138 0.0012 0.0001 n-C28 0.0000 0.0000 0.0000 0.0115 0.0010 0.0009 n-C29 0.0000 0.0000 0.0000 0.0110 0.0012 0.0009 n-C30 0.0000 0.0000 0.0000 0.1218 0.0032 0.0001

1.0000 1.0000 1.0000 1.0000 1.0000 1.0000

Paper 1.2 Field Applications of Model-Based

Multiphase Flow Computing

ge Rasmussen FMC Kongsberg Subsea

North Sea Flow Measurement Workshop 26-29 October 2004

1

FIELD APPLICATIONS OF MODEL-BASED MULTIPHASE FLOW COMPUTING

ge Rasmussen, Flow Management Systems, FMC Kongsberg Subsea

1 ABSTRACT A virtual flow meter is an alternative or supplement to the traditional multiphase flow meters. The technology makes use of all existing instrumentation, combined with models of the production system, to estimate the flow in wells and flow lines. The advantage of using virtual flow meters is that the model provides additional information of the flowing conditions in the entire production system, as opposed to at the meter positions only. Thus it is well suited as basis for a Flow Management System (FMS). A FMS typically provides a set of tools to extract important information from available data, give access to additional information using process models, and include automatic controllers which adjust the control variables in an optimal way. Vast improvement can be achieved in processes where quality measurements are a commodity in short supply and the natural control loops are interacting. The need for a FMS is even more conspicuous when there is a need for measurement redundancy, and constraint handling and profit maximization are important parts of the control purpose. The basis of such a production support system is a mathematical model of the underlying processes. In short, the combination of multiphase flow models and available measurements has numerous field applications with a considerable economic potential compared to using the measurements alone, also when multiphase flow meters are available. This paper demonstrates some of the capabilities of FlowManager applied to subsea wells and pipeline networks. The concepts will be illustrated by field applications. Future possibilities using the technology will also be addressed.

2 INTRODUCTION Metering of pressure and temperature through the entire production system has been common practise for years. The measurements are used to estimate the contribution of the different wells, and to supervise the condition of the production system. Every production system design today is based, to a certain extent, on numerical flow correlations. The correlations are already adequately proven to base the field design on. The idea behind virtual metering is to combine the numerical correlations that were used to design the field, with the metering that is already in place. Combined these two pieces of information can give a full field online production diagnostic. In this sense virtual metering is nothing new; it is only a way of refining the information that is already available. The virtual meter is inherently a link between the available measurements, and a numerical model. Prediction of any event that has an accurate model is possible within this framework. Such a system easily lends itself to production optimization on a well by well basis, or de-bottlenecking of the full production system.


2

3 PRINCIPLES Virtual metering can be performed online or offline. The only prerequisite is that all sensor information is gathered within the same time frame to facilitate the steady state correlations used. The basic idea is to combine the sensory input with the numerical flow models. The virtual meter calculates a flow field through the entire production system. The flow field will inherently contain the calculated sensor responses at all sensor locations. The discrepancies between measured and calculated properties are quantified within an object function, and the object function is minimized through an iterative search. The procedure is described in its simplest form in Figure 3.1. The figure shows how the analytically solvable problem of one phase flow through a venturi can be solved iteratively by imposing an object function.

Figure 3.1 The concept of the Object function. The one dimensional problem lends it self easier to illustration than the three dimensional problem consisting of three flow rates, oil, gas and water, flowing through an oil well with multiple sensors. The principle however, is the same.


3

Q water

Store

Initiate @ P 1 ,T 1

Calculate object function

Guess phase rates

Simulate Well or System Performance

Accept?

Search for new rates

No Yes

Qgas

Q oil

P,T

P,T

P,T

Valve Position

Figure 3.2 Oil well with standard instrumentation Figure 3.2 shows the operating principles of a virtual meter. All required information is fed into the model either on an on- or offline basis. The model makes an initial guess of the rates and calculates the equivalent sensor responses. The discrepancy between the measured and the calculated system responses are quantified through the object function and evaluated. The procedure is repeated iteratively until the value of the object function reaches one of a set of cut off criteria. In order to obtain a well-posed mathematical problem, there needs to be at least one measurement pr unknown, the unknowns being the three flow rates. The previous figures demonstrate the procedure with one and three unknowns. The system can be expanded to a full converging network of wells and lines. Recent implementations of the software have included the production separator measurements and/or the fiscal export measurements. By increasing the number of available measurements the calculations will become over determined, and the virtual meter will become increasingly redundant towards errors and failures in the individual sensors. To enable the software to use the separator measurements, or the results from a multiphase meter, it is necessary to have a solution that allows for a controlled difference between the measured and the calculated quantity. The FlowManager software contains a weighted object function. Assigning different weights to the different measurements allows the user to accentuate or discard certain measurements. In an over determined system this feature gives an operator the possibility to use the system for sensor evaluation. Any fluid property that can be both measured and modelled is available to be included in the virtual meter. In this sense the meter can be used as both an addition to, and/or a substitute for traditional multiphase meters. The virtual meter can incorporate the individual sensor readings that a multiphase flow meter consists of, or it may use the calculated rates directly.


4

The readings from the multiphase meters are incorporated with all other measurements within the production system, to give a best fit solution for the whole production system. In many fields were multiphase meters are employed the uncertainty caused by discrepancies between the sum of all meters, and the fiscal measurements, causes uncertainty in all measurements. The virtual meter is not limited to giving a spot reading at the sampling position. Hence, it may be used as a supervision tool to establish the most likely source of error.

4 EXPERIENCE

4.1 Closed loop control One of greatest success stories so far in using virtual meters is to link the output from the rate calculations directly with a closed loop choke controller. This functionality is currently applied on the TrollB and TrollC fields in the North Sea. The reservoirs are characterized by their high mobility and large gas cap. Consequently, the reservoir has very rapid transients following the onset of a gas breakthrough. The optimum production point is a constant level of gas breakthrough to provide the wells with additional lift. On the other hand, if a choke is opened to much, the well oil production will die temporarily due to the high mobility of the gas phase. Figure 4.1 shows a simplified schematic of the anticipated behaviour of a Troll well.

Figure 4.1 Simplified production curve for Troll Oil well Because of the high mobility and rapid transients, in addition to the fact that there are more than 110 wells for the two fields combined, it would be virtually impossible for the operators to control this situation if operating all wells by hand.


5

Figure 4.2 Wellhead pressures of two wells, before and after applying Production Control. Figure 4.2 shows the wellhead pressure of two wells before and after the closed loop control was applied. In addition to stabilizing the two wells, the sum of the oil production from the two wells was increased by 300 Sm3/day (1890 bbl/day). The increase in wellhead pressure correlates with a higher oil production due to the increased gas production (Figure 4.1).

4.1.1 Deduction Testing The previous paragraph shows a scenario where the financial benefits of closed loop control are quite obvious. For other fields with less permeable reservoirs, the benefits might not be as pronounced. However, there are several scenarios were test separator availability is low. Due to the increased cost focus in developing new fields, fields are being equipped with single manifold risers, and maybe even without a test separator. This kind of production system is cost effective to develop, but will most likely suffer in both recovery speed and overall recovery if the contribution of the individual wells can not be established. Other commonly encountered scenarios are in the latter stages of a field life when gas and water production exceed the test separator capacity, prohibiting testing at full production, or the test separator may be tied up to producing low pressure wells. Any of these scenarios will leave the operator dependant on deduction testing to establish the contribution of the individual wells. A virtual meter may be used to calculate the contribution of each well, but a software based solution will be dependant on the available calibration data to provide correct flow rates. Without the calibration data the results from the virtual meter will be as uncertain as the model and measurement uncertainties combined. Deduction of the measured rate on the production separator is a valuable tool to increase the accuracy of the well allocation, but this tool will only give information if the remaining wells are kept at constant production. The effect of closing or reducing the flow from one well to enable


6

deduction testing might significantly alter the flow rates of other wells. To facilitate deduction testing one is dependant on having a tool that will ensure consistent flow rates from all wells in spite of the varying backpressures. By applying a tool such as closed loop rate control, one can ensure that all other wells than the one being tested are held at a constant rate. The closed loop control can be set to maintain the upstream choke pressure at a constant level for all wells except the one being tested. Even without significant calibration of the remaining wells, the choke controller is able to maintain the pressure based on standard regulation principles. By running the wells continuously in the control loop, all controlled choke changes will inherently function as a differential test.

4.2 Chameleon Case Due to time limits there was not enough time to clear the release of the data with the operator of the field. However, we were allowed to use the data as long as all presented data would be rescaled and made anonymous. For presentation purposes we will address the field as the Chameleon field. Initially a three well subsea development with pressure and temperature readings bottom hole and upstream and downstream the production choke of each well (Figure 4.3). The bottom hole gauges on well B-2 failed prior to first oil from the well. In addition something got caught in the production choke of the well, which rendered the choke calculations useless. This rendered the virtual meter with no useful information for the well, and in turn unable to calculate rates, even with fixed ratios.

Figure 4.3 Schematic of the Chameleon Field Wells B-1 and B-2 are completed in the same reservoir zone, at approximately the same depth. As an immediate solution, the bottom hole pressure reading from B-1 was used on B-2 to calculate rates. This scheme would only work provided that the wells had fairly similar or very high Productivity Indexes, or were produced at approximately the same rates. To allow for a greater variation in the production from B-1, a network solution was applied. The obvious danger in using a network solver in a case like this is that one is no longer able to determine whether there are errors in the calculations of the two remaining fully instrumented wells, because all production not accounted for will automatically be attributed to the well without


7

any functioning measurement elements. Hence, an error in the calculations of B-1 or B-3 would most likely be overseen. To circumvent this problem we opted for an intermediate solution; Running single well calculations on all three wells, calculating the contribution of well B-2 by using the bottom hole sensor of B-1, and continuously verifying the sum of the three wells against the production separator. One of the critical issues in this paper is to show how a virtual meter will enable an operator to establish individual sensor errors through modelling the full field. In July of this year a discrepancy between the field flow measurements, and the FlowManager calculations arose. The reason was found to be an error in the separator oil measurement. The sensors would no longer detect oil flow rates below a certain threshold value.

0.50

0.60

0.70

0.80

0.90

1.00

1.10

1.20

1.30

1.40

1.50

30.06.2004 00:00

02.07.2004 00:00

04.07.2004 00:00

06.07.2004 00:00

08.07.2004 00:00

10.07.2004 00:00

12.07.2004 00:00

'FieldFactor.Gas'

'FieldFactor.Oil'

Figure 4.4 Field Factor development. Figure 4.4 shows the development the Field factors over a period of two weeks. The field factor is defined as the calculated production divided by the measured production. On July 6th there is a pronounced increase in deviation between the measure and the calculated rates.


8

Separator Oil Measurement

0

50

100

150

200

250

30.06.200400:00

02.07.200400:00

04.07.200400:00

06.07.200400:00

08.07.200400:00

10.07.200400:00

12.07.200400:00

14.07.200400:00

Figure 4.5 Separator oil measurements Figure 4.5 shows the measured oil production over the same time frame. After July 6th there is not one single measurement beneath the threshold value of 136,67 Sm3/day. The sudden increase in production does not correlate with changes in any of the three wells. The discrepancy was initially seen as an error in the virtual meter, but after examining the system closely, the origin was found to be the production separator. The error was first detected, not by the operating personnel, but by the virtual meter. Without continuously monitoring of the production system response, the error might have gone on virtually undetected.

4.3 Standard Instrumentation The more wells contained within a production system, the less extra information is gained by solving the full production system as a network. In some scenarios it might even be beneficial to split the system in to smaller units to gain an overview of the calculations. The capability for using network solution is a fairly new development within the FlowManager framework. Hence, most of our experience has been gained in fields were the calculations are run on a well by well basis. In this scenario the most common problem is fallout of down hole sensors. For all fields where FlowManager is currently employed, the pressure drop between the down hole and the upstream choke pressure sensors is one of the most important measurements. The reason being that it is often the largest pressure drop in the system, and is therefore least affected by the individual measurement errors. When the down hole sensors fail, which they quite often do, the program is limited to using the choke Dp and well bore temperature drop as its primary measurements. The choke Dp is governed by the choke geometry, in other words choke calculations are determined by the choke Cv-curve supplied by the choke vendor. There are several different difficulties encountered when using the choke Dp as a measurement. The most obvious obstacle is low pressure drops at large choke openings. We have no experience with using chokes with dedicated Dp-cells. Hence, the choke Dp reading is a product of two sensors with individual errors. For very low pressure drops, the sum of the errors may be larger than the actual pressure drop. Secondly, the transition zone between sub-critical and critical flow through the choke is a difficult area. The Speed of sound of each of the individual phases may be established through laboratory testing, or through numerical pvt relations. The speed of sound of the three mixed phases is established through the Woods equation. The accuracy of the individual


9

phase speed of sounds are imperative to the accuracy of the calculations. Laboratory measurements of the phase speeds are very accurate, but also very costly. Hence, we have no experience with measured phase speeds for the fields we operate. The transitional area is therefore a somewhat grey area for determining the flow rates based on the pressure drop. Another commonly encountered problem is the step actuators. Our input is usually in the form of a step percentage, which we convert into a percentage of the maximum area. After a choke has been run for a while, it is very common that the reported step percentage starts to deviate from the actual. Over time this deviation may grow quite large. The reason for the deviation is that the choke does not always move the exact amount of instructed steps. Over time the deviation from the measured step percentage will grow. The error will be reset every time a choke goes to full open or full closed. Overall the choke pressure drop is an excellent measurement for determining the mass flow rate provided that it is sufficiently large, and that the flow rate is either higher or lower than the transitional flow rate between critical and sub-critical conditions. Temperature measurements are somewhat less accurate than the pressure readings. The primary reason is that the sensors rarely give an accurate indication to the average temperature of the flowing liquids. In most cases temperature sensors are non intrusive, and to a certain degree influenced by both the flowing temperature and the ambient temperature. For down hole sensors this does not affect the measurements to a disturbing degree. For measurements in areas where the ambient fluid is seawater, the deviation might be substantial. Ambient temperatures are subject to seasonal changes, this can to a certain degree be accounted for by monitoring the seawater temperature, or even using previously established seabed temperature statistics. But even more importantly is the convective heat loss that arises as soon as the water surrounding the sensor area starts moving. Underwater currents may substantially affect the temperature readings. This effect is even more prominent if one is attempting to model heat losses through a seabed to surface riser. In this situation it is not the actual temperature measurement, but the modelling of the U-value of the riser that is the problem. Another difficulty with using temperature as a primary measurement is the long transient times compared to the pressure responses. The time to reach temperature equilibrium is further increased by the fact that the measurements are conducted in the surrounding metallic casing, and not within the actual flow. For allocation purposes this effect might be marginal since most chokes are held at constant positions for time intervals that are much longer than the temperature transients. However, in a situation were closed loop control (Chapter 4.1) is applied; the temperature responses are too slow to be used as a primary measurement. Increasing use of isolation for hydrate prevention purposes is another obstacle. Better isolation of risers and flow lines may reduce the information available from temperature sensors, or it may accentuate the need for high quality intrusive temperature sensors. As stated in chapter 3, any flow related quantity that can be measured and modelled, may be used as input to a virtual meter. We have very good experience with using Venturi pressure drop readings in addition to standard sensory input. Other possibilities are densitometers, water cut meters, speed of sound meters and cross correlation based flow meters. In sum a virtual meter may easily link up with the individual components of a multiphase meter, and combine this information with all other available information in the production system to establish a best fit scenario for the whole production system, not limited to spot readings at the position of the multiphase meters.


10

5 FLOW MANAGEMENT One of the greatest added benefits of using a virtual meter, compared to other forms of single point meters, is that it automatically gives a tuned model of the full production system. For de-bottlenecking of a production system an operator is constantly dependant on being able to model the process. Usually this is done by tuning the individual wells to the latest well tests. The frequency of testing is very variable. Especially in the later stages of a field life we frequently see test separators being tied up to producing low pressure wells, or unable to test wells at full capacity due to gas and water limitations. Hence, the predictions and the optimizations suffer from having inaccurate model input. The fact that one already has a fully functioning, and calibrated model of the full production system allows for quick and easy assessments of both short and long term production predictions. For long term predictions the FlowManager software has built in functionality for communicating directly with the Eclipse reservoir simulator. For more short term predictions one can use the calculated rates from previous days and run them through an optimization loop. This loop starts with the actual rates from previous days, and then optimizes and maximizes production based on a set of constraints given by the operator. The constraints naturally include the PI and formation pressure of each well. Gas and water production can be controlled as a function of time and/or flowing bottom hole pressure. All major naturally occurring variations in the production system can be expressed mathematically as constraints, to accurately predict how the system will respond to different production scenarios. Linking the model of the production system gives the benefit of accurate production estimates based on the reservoir models. Vice versa the increased accuracy in the model of the production system will actually increase the accuracy of the reservoir predictions. The reservoir models have more accurate input with regards to how much has been recovered from where. The virtual meter is inherently a link between the measurements, and a numerical model. Prediction of any event that has an accurate model is possible within this framework. Possible future model expansions include slug, wax and scale predictions, along with erosion estimates and inherently all other predictions needed to optimize the production from a hydrocarbon reservoir.

PUBLICATIONS AND PRESENTATIONS

The following publications and presentations highlight the methodology and field examples of the Idun allocation application.

[1] Berg, K. and Davalath, J., Field Applications of Idun Production Measurement

System, OTC, Houston, TX, May 2002.

[2] Andfossen, P.O. Idun Multiphase Metering: Subsea Processing in Deepwater OTC, Houston, TX, May 2000

[3] Friedemann, J.D. Data Acquisition, Cost Benefit Evaluation Challenges of Multiphase Flow in Horizontal Wells Workshop, Porsgrunn, Norway, June 1999.

[4] Nerby, G., Sther, G, and O.J. Velle, A Cost Effective Technique for Production Well Testing OTC, Houston, TX, May 1995

[5] Sther, G. Software Determines Multiphase Flow without Meters, Petroleum Engineer International, Dec. 1998

Paper 1.3 Quantifying Financial Exposure Due to Meter

Uncertainty Using Monte Carlo Simulation

Claudio Ciglio Advantica

Lawrence Couglan

Shell Exploration & Production Europe


1

QUANTIFYING FINANCIAL EXPOSURE DUE TO METER UNCERTAINTY USING MONTE CARLO SIMULATION

Mr. Claud Giglio, Advantica Mr. Lawrence Coughlan, Shell Exploration & Production Europe

1 INTRODUCTION This paper details the steps undertaken and the thought processes followed in the design and delivery of a user friendly exposure software tool. The developed tool was intended to give the user an appreciation of exposure in financial and quantity based terms as a consequence of physical and contractual conditions applicable to both the complete system and also at each node within the system.

The ability to easily and reliably quantify financial exposure (potential lost revenue) due to oil and gas flow meter uncertainties within gas and oil production fields is regarded as a valuable asset by many shippers. This is especially true for shippers who regularly deal with high oil and gas flow rates through key fiscal meters, which are in turn used as the basis for billing and accounts.

An example of a generic exposure model, representing a typical fiscal meter and associated flow-metered pipelines is detailed. An expression for financial exposure is derived from this model in terms of key domain and market attributes, including metered flow rates, meter biases, random uncertainties, equity shares (percentage take of production), taxation rates and other variables. The model has also been expanded to support the modelling of virtual meters to accommodate the scenario in which there is no telemetry for a particular meter.

2 BUSINESS BENEFITS A dual requirement for support tools based on project and operating requirements inspired the initial concept for the exposure quantification software. The project requirement encompassed the need to be able to determine how much to spend on a new system based on defined deliverables and life cycle returns. The idea was to enable the user to manipulate input parameters and obtain an estimate of life cycle costing thus allowing the project team to optimise the design of the metering system to meet the required specifications whilst determining the best-cost solution. The second requirement was based on the need to help managers understand the consequences of decisions in the short and longer term as a function of product quantity and value. Also to help system managers understand the consequences of third party system performance and equity difference implications.

As each system evolved it became evident that the exposure model elements were similar and could be combined into the one package. After engagement of the key stakeholders on the concept and expectation, it became evident that the input requirements would be at a detail that would potentially need engineering knowledge; conversely the output requirements were simple in terms of visual representation of geographical area with quantification of exposure.

The full value of the tool can be realised by moving away from a stand alone package to one that is loaded on the network so any manager can access to see how their systems are performing and which particular nodes have most exposure. Thus allowing decisions on resource usage to be made with sound supporting information.


2

3 TOOL OVERVIEW The prototype exposure calculator tool (implemented using Microsofts Visual Basic for Applications language - VBA) allows the uncertainty of interrelated meters to be expressed in terms of normal or triangular random error distributions and configurable bias functions thus allowing utilisation of Monte Carlo Simulation to generate likely monetary and non-monetary exposures for a shipper. The output exposure figures can be used to help identify under performing site meters.

As well as reporting detailed statistics from the simulation, the tool allows the financial exposures to be graphically displayed on an interactive map in the form of a configurable green, amber, red traffic light status scheme. This satisfies both operator and managerial requirements by allowing financial exposure results to be viewed at different levels of detail.

Key features of the exposure calculator are listed below.

Calculation of financial exposure due to meter uncertainty at installation and field levels.

Non-monetary and monetary exposure results output (with user defined confidence level).

Configurable meter accuracy and precision parameters via a dedicated meter error definition form.

Easy switching between default and custom meter parameters allowing what if scenarios while preserving protected default information.

Modelling of flow rate dependent bias and flow rate dependent random uncertainties.

Intelligent modelling of virtual meters.

Adoption of a Monte Carlo Simulation Engine enabling generic processing of a variety of meter error distributions (Normal and triangular are currently supported).

Intuitive interactive map with colour coded exposure status results.

User definable traffic light limits, where the user can define what monetary exposure value constitute a high, medium or low risk for each installation.

Built-in utility to facilitate adding new installation labels to the main map (eliminating the need for coding).

Easy modification of global oil and gas unit prices.

Easy cycling through installation Group or installation site details (in alphabetical order) via the Main GUI.

Optional export of results and associated user inputs for reliable auditing.

Editable General information forms for each installation and field to hold non-simulation dependant text.

Built-in (expandable) user help.

Hidden and restricted data areas accessible by user defined passwords.

Comprehensive and configurable data validation and connectivity diagnostic capabilities.

Modularisation of business logic in hidden and protected VBA modules.


3

3.1 Tool Layout

In summary, there are two main user interfaces; these are the interactive map and the main Guest/Global User interface (GUI).

3.1.1 Interactive Map

The Map sheet displays an interactive schematic of the financial exposure for each installation via a three level colour coding hierarchy. The user can easily toggle between the financial exposure status for oil and for gas by checking the appropriate check box.

Note: The diagrams shown are based on fictitious or unpopulated data and do not represent actual conditions.

Clicking on a particular installation label of interest launches the Main GUI interface where the user can view or modify the low-level installation details.

Table 1 gives a description of the exposure status key, which can be seen toward the upper left of Fig. 1.

Table 1 Map Financial Exposure Status Legend

Exposure Status

Colour Code Description

LOW Green Installation is considered to have an acceptable financial exposure value.

MEDIUM Amber Installation is considered to have a financial exposure value that is above normal.

HIGH Red Installation is considered to have an unacceptable financial exposure. Assuming that the traffic light limits have been sensibly set, then the High status can be attributed to unusually high meter biases and/or high random uncertainty in the installation Fiscal meter and/or one of its child field meters.

NO RESULT Blue A No Result status can be returned due to a number of factors; however the most likely causes are, invalid mapping of the actual label to its data source or undefined traffic light limits.

Fig. 1 Partial screen shot of the main interactive map (central fields only)


4

3.1.2 Main GUI

The main GUI shown in Fig. 2 provides the user with an interface for modifying and viewing detailed installation and field related information. This includes modification and viewing of associated fiscal and field meter attributes via detected forms.

The information on the main GUI represents the information for one single installation and its associated child fields.

Alternatively the user can scroll by installation group. The (blue) command buttons to the left of each field name, when clicked, launches a meter form where the user can modify various parameters that define a given meters accuracy and precision. Also the (yellow) command buttons at the far right of the same field table allows the user to view general descriptive text about a given field. These types of buttons are also present for each gas and oil installation.

Fig. 2 Screen shot of the main GUI


5

4 EXPOSURE MODEL OVERVIEW Fig. 3 shows a partial domain model highlighting the important items. This figure gives a visual representation of the real-world objects and concepts of interest. Also see Fig. 6 (page 8) that also illustrates the connectivity for a typical installation and its associated fields.

Note, since we are not concerned with the actual transportation effects of oil or gas then the real-world pipes have been excluded from Fig. 3. Readers familiar with UML1 can see that the model also describes the multiplicity between various domain objects. For example, an installation must have at least one child field, which in turn can have either one flow meter or none.

4.1 Meter uncertainty

There are two basic elements in defining the overall uncertainty for a generic meter, namely its accuracy and its precision.

4.2 Meter accuracy

The accuracy of a meter is usually expressed in terms of a bias value, which may have a constant (offset) component and or a flow proportional component (allowing the metered value to drift from the true value as the flow rate changes). A positive bias means that the meter is overstating (over reading) the true value; conversely a negative bias means that the meter is under reading. A meter may be calibrated against some standard so that any bias is reduced to zero (or very near to zero for a given operating range) thus giving an accurate meter. In practise there will always be some random uncertainty associated with the calibration process itself partially due to the fact that the true value (for this standard) can never be precisely known. This random uncertainty in the calibration process is not modelled.

1 The UML (Unified Modelling Language) is a standard diagramming notation.

Fig. 3 Partial Domain Model

Installation Group

ID Name

Installation

ID Name

Flow Meter

ID Metered Value Bias Rand Uncertainty

1 has 1

Production Field

ID Name Equity Share Tax

1 is supplied by 1.*

1 has 0.1

1 contains 1.*


6

4.3 Meter precision

The precision of a meter is a measure of the random uncertainty associated with the meter result (or reading). If a meter were to be read many times under the same conditions2 and the readings were to be varied, then the meter would be classed as being imprecise, conversely if the readings were very similar then the meter would be regarded as precise. In reality there are many factors that contribute to the random uncertainty for a particular meter measurement and there are a number of ways that this uncertainty can be expressed. A common format is to state the measured value M together with a quoted uncertainty k as

kM . The exposure tool allows an additional flow rate dependent uncertaint

22nd North Sea Flow Measurement Workshop 2004

Documents

Transcript of 22nd North Sea Flow Measurement Workshop 2004