Expanded Knowledge on Orifice Meter Response to Wet Gas Flows

Transcription

1 32 nd International North Sea Flow Measurement Workshop October 2014 Expanded Knowlede on Orifice Meter Response to Wet Gas Flows Richard Steven, Colorado Enineerin Experiment Station Inc Josh Kinney, Colorado Enineerin Experiment Station Inc Babajide Adejuyibe, BP Exploration Operatin Company, Ltd Kim Lewis, DP Dianostics LLC 1 INTRODUCTION With orifice plate flow meters bein one of the most widely used flow meters in the natural as production industry it is inevitable that they are often used in wet natural as flow applications In 2011 BP, ConocoPhillips (CoP) and CEESI released a comprehensive paper (Steven et al [1]) on the known performance of the orifice meter to wet as flow conditions This paper described the performance of 2 throuh 4 orifice meters across a wide rane of wet natural as flow conditions However, as with all as meter types, there are still sinificant unknowns reards the performance of orifice meters with wet as flow In this paper three of these unknowns are addressed: The public knowlede of orifice meter wet as performance includes an understandin of as with a liht hydrocarbon liquid and / or water In this paper a rare 4, 0683 beta ratio orifice meter data set is discussed that includes the effects of wax in the hydrocarbon liquid and / or MEG bein present The publicly available orifice meter wet as flow data set is larely confined to the nominal diameter (Ф) rane 2 Ф 4 In this paper massed CEESI 8 orifice meter multiphase wet as flow data is discussed Orifice meter wet as correction factors require the liquid flow rate be supplied from an external source Traditionally the liquid flow rate is metered by periodic spot check techniques An unnoticed chane in the liquid flow rate between such checks induces a as flow rate prediction bias A real time qualitative liquid loadin monitorin system is therefore useful In this paper the DP meter dianostic system Pronosis (e Skelton [2]) is analysed with wet as orifice meter data to investiate its on line liquid loadin monitorin capability 2 THE DEFINITION OF WET GAS FLOW PARAMETERS In this paper a wet as flow is defined to be any two-phase (liquid and as) flow where the Lockhart-Martinelli parameter (X LM ) is less or equal to 03, ie X LM 03 This definition covers any combination of aseous and liquid components That is, the liquid can be a liquid hydrocarbon, water or a mix of liquid hydrocarbon and water Other liquids and substances such as MEG / methanol, wax etc may also be present X LM ml, total ρ = --- (1) m ρ l The Lockhart-Martinelli parameter (equation 1) is a non-dimensional expression of the relative amount of liquid with the as Note that ṁ & m l, total and liquid mass flow rates respectively (where m l, total are the as is the sum of the individual 1

2 liquid component flows), while ρ & ρ l are the averae bulk as and liquid densities respectively The as to liquid density ratio ( DR = ρ ρ ) is a non-dimensional expression of pressure The as densiometric Froude numbers ( Fr ), shown as equation 2, is a non-dimensional expressions of the as flow rate Note that is the ravitational constant, D is the meter inlet diameter and A is the meter inlet cross sectional area Fr m = A D l ρ 1 ( ρ ρ ) l ---- (2) With one sinle liquid component a wet as has one liquid density With multiphase wet as flow there is two (or more) liquid densities In this case the liquid density used to calculate the as to liquid density ratio and the as densiometric Froude number is the averae liquid density Water cut is the ratio of the water to total liquid volume flow rates when the fluid is at standard conditions In this paper water to liquid mass ratio (or WLR m ) is defined as the ratio of the water mass flow rate to total liquid mass flow rates As this paper discusses mixtures of water, hydrocarbon liquid and MEG, the WLR m is calculated here by equation 3, where ṁ w, ṁ hcl & ṁ MEG are the mass flow rates of water, hydrocarbon liquid and MEG respectively = m --- (3) m w + m hcl + m MEG WLR w m It is common enineerin assumption to consider the liquid components homoenously mixed The homoenous liquid phase density ( ρ l, hom) is calculated by equation 4, where ρ w, ρhcland ρ MEG are the water, hydrocarbon liquid and MEG densities 1 respectively For multiphase wet as flows it is this liquid mixture density that is used to calculate the wet as flow parameters ml, total ρwρhclρw ρ = --- (4) l,hom ρhclρmeg mw + ρwρmeg mhcl + ρwρhcl mmeg Equation 5 shows the orifice meter sinle phase as mass flow equation E & A t are the velocity of approach and minimum cross sectional area respectively (both eometric constants), C is the dischare coefficient, ε is expansibility and P d is the traditionally read differential pressure (DP) The liquid induced as flow rate prediction error is often called an over-readin, denoted here as OR The traditional DP read when the flow is wet ( P tp ) is different to that which would be read if that as flowed alone ( P ) The result is an erroneous, or apparent, as mass flow rate prediction, ṁ (see Equation 5a) Note that apparent ṁ apparent, tp and C d, tp are the apparent (incorrect) as mass flow rate prediction, the as ε 1 The authors consider wet as flow fluid properties to be information supplied to the flow computer from external means Fluid property information is assumed to be correct Whereas this is standard practice when discussin multiphase / wet as flow meterin, it is reconized that the supply of accurate fluid properties in field applications is a difficult challene for operators 2

3 m = EA C ε 2ρ P t d m = EAε C 2ρ P --- (5) apparent t tp d, tp tp --- (5a) m OR = Apparent m = ε tpcd, tp εc d P tp P P tp P --- (6) m Apparent P % 1 *100% tp OR = 1 *100% --- (6a) P m expansibility and the dischare coefficient found respectively when applyin the wet as differential pressure (For many flow conditions Cd, tpε tp Cdε ) The overreadin is expressed either as a ratio (equation 6) or percentae (equation 6a) comparison of the apparent to actual as mass flow rate Correction of this overreadin is the basis for orifice meter wet as correlations 3 ORIFICE METER MULTIPHASE WET GAS FLOW CORRECTION FACTOR By 2011 CEESI, BP and CoP had athered a massed orifice meter wet as flow data set (of the rane 2 D 4 ) from multiple owners, tested at different facilities over many years Fiures 1, 2 & 3 show some of the orifice meters under test Fi 1 2 Orifice Meter at CEESI Fi 2 4 Orifice Meter at TUVNEL 2 multiphase wet as flow facility 4 two phase wet as flow facility Fi 3 4 Orifice Meter at CEESI 4 multiphase wet as flow facility 3

4 m m, apparent = --- (7) 2 1+ CX LM + X LM strat Fr * transition ρ C = ρl ( 02 WLR ) = -- (9) { 01* ( ( ))} A = 04 + exp WLR m m n ρl + ρ # -- (10) = 1 2 Fr # A, transition n -- (11) 2 n --- (8) = 1 2 n = for Fr Fr, transition n strat 2 # A Fr -- (12a) n for Fr > Fr, transition -- (12b) Note that Fr then 1 2 n as required Fi 4 All to 4 orifice meter wet as data with and without new correction Parameter Rane Pressure 67 to 789 bara Gas to liquid density ratio < DR < 0111 Fr rane 022 < Fr < 725 X LM 0 X LM < 035 Inside full bore diameter 194 D 4026 Beta 0341 β 0683 Gas / Liquid phase Gas /Liquid Hydrocarbon/ Water Table 1 The 2011 Multiphase Wet Gas Flow Data Set Flow Rane 4

5 Table 1 shows the rane of the 2011 data set Not all the maximum or minimum parameters were tested toether Analysis of this massed data set of 1656 wet as flow points led to a 2 to 4 orifice meter wet as correction factor reproduced here as equation set 7 throuh 12b Fiure 4 shows both this uncorrected data, and the data corrected for known liquid flow rates For precisely known liquid flow rate inputs this correction factor has a 2% uncertainty to 95% confidence 4 A UNIQUE 4, 0683β ORIFICE METER WET GAS FLOW DATA SET In 2013 BP commissioned CEESI to test equipment installed in 4 pipe with a natural as, liht liquid hydrocarbon (or condensate ), water and MEG flow The condensate had in excess of 22% by weiht of C30+ At ambient conditions this condensate forms wax This wax component could solidify at temperatures less than approximately 97 o F Fiures 5 & 6 show the condensate with wax at ambient conditions before and after mixin Fi 5 Condensate with separated wax Fi 6 Condensate with wax, mixed CEESI built a specialist test facility to conduct such tests The test rane is shown in Table 2 All tests were conducted at approximately 105 o F to assure no wax deposits formed in the pipe network Parameter Rane Pressure 67 < P (Bara) < 74 Gas to liquid density ratio 005 < DR < 0089 Fr rane 20 < Fr < 48 X LM 0 X LM < 0086 Inside full bore diameter 4026 Beta 0683 Gas / Liquid phases Natural Gas/Hydrocarbon Liquid with wax/water/meg Table 2 The 2013 BP Multiphase Wet Gas Flow Data Set Flow Rane In addition to the primary equipment under test CEESI added a 4, schedule 40 flane tapped orifice meter upstream of the main test section This meter installation was ISO 5167 Part 2 compliant includin > 44 pipe diameters of straiht pipe between the upstream sinle ninety deree bend and the meter inlet Fiure 7 shows photoraphs of the test facility From top left oin clockwise, 1) the orifice meter under test, 2) the multiphase wet as flow view port, 3) separation and storae vessels, & 4) the water, liquid hydrocarbon (with > 22% by weiht of C30+) & MEG injection into the natural as flow The test matrix was formed to suit the primary equipment bein tested Relatively hih flow rates meant that a 0683 beta ratio plate was required The test matrix consisted of dry natural as, natural as with water only, natural as with water 5

6 Fi 7 The CEESI Sinle Pass Multiphase Wet Gas Flow Facility and MEG only, and natural as with water, condensate and MEG In total 274 flow points were recorded from the orifice meter Fiure 8 shows the three dry as flow results recorded More as flow data would have been ideal, but the main equipment under test did not require it The root mean square of the orifice meter ISO dischare coefficient uncertainty (approximately 065%) and the 6 reference turbine meter uncertainty (075%) is approximately 1% The orifice meter is shown to be serviceable The CEESI multiphase wet as view port allowed filmin of the flow pattern The flow pattern, ie the distribution of the liquid in the as flow, influences the orifice meter s as flow rate over-readin The flow pattern is influenced by liquid properties The introduction of MEG, and condensate with > 22% by weiht of C30+, meant that this wet as data set had some different liquid properties to the existin massed orifice meter wet as flow data sets There was therefore a possibility that the flow pattern, and hence the over-readin, would be somewhat different than the existin massed orifice meter wet as flow data sets Fiures 9 & 10 show stills taken from multiphase wet as flow video footae recorded durin these tests There was no visual indication of the presence of wax or MEG The flow patterns looked similar to those that CEESI would expect from 15 years of experience Nevertheless, visual inspection would not show small chanes in the liquid dispersion Evidence of that could only be found in data 6

7 Fi 8 Sinle Phase Gas Flow Performance of 4, Sch 80, 0683β Orifice Meter Fi 9 4, Stratified to Annular Mist Fi 10 4, Annular Mist Flow As the test matrix was not chosen for the orifice meter research the data did not allow the effects of different liquid component ratios on the orifice meter wet as over-readin to be isolated The 2011 orifice meter wet as correlation (ie equation set 7 throuh 12b) was fitted from wet as flow data with no MEG and no condensate with sinificant C30+ content This correlation is bein extrapolated when applied to this data set Fiure 11 shows all the recorded data from these tests Both uncorrected data and data after the 2011 wet as flow correction factor has been applied, with known liquid flow rates, are shown The flow ranes were: 78 as flow (k/s) condensate flow (k/s) 0 26 water flow (k/s) MEG flow (k/s) 0 The Lockhart Martinelli parameter rane (X LM < 0086) was achieved by various combinations of condensate, water and MEG An increasin Lockhart Martinelli parameter causes an increasin over-readin Extrapolatin the available correlation for known liquid flow rates corrected the as flow rate prediction to within 2% of the systems as reference system That is, the 2011 correlation corrected this 4, 0683β orifice meter wet as data set to within the published uncertainty of 2% to 95% confidence The presence of MEG & condensate with sinificant C30+ content had no noticeable adverse effects Fiure 11 shows that at low Lockhart Martinelli parameters the meter tended to under-read the as flow This is typical of orifice meters, and this is a welldocumented phenomenon (e Tin [3]) 7

8 Fi 11 All Multiphase Wet Gas Flow 4, 0683β Orifice Meter Data Uncorrected, and Corrected with 2011 Wet Gas Flow Correction Factor It can be concluded that the presence of the MEG & condensate with sinificant C30+ content had no sinificant effect on the orifice meters reaction to the wet as flow This suests that the correlation is relatively robust and not too sensitive to certain real world practical problems However, it is very likely that at thermodynamic conditions where wax forms there will be sinificant adverse affects 5 CEESI 8, 0689β ORIFICE METER WET GAS FLOW DATA SETS In 2011 CEESI commissioned a nominally 8 multiphase wet as flow facility An 8, flane tapped dual chamber orifice meter was permanently installed downstream of the test sections This meter was ISO 5167 Part 2 compliant, inclusive of 44 pipe diameters of straiht pipe between the upstream sinle ninety deree bend and the meter inlet A beta orifice plate was selected Fiures 12 & 13 show photoraphs of this orifice meter and the facilities view port system respectively Fiures 14 & 15 show sample stills taken from 8 wet as flow pattern footae This orifice meter has remained in place since the initial commissionin of the 8 facility It has athered a massed 8 orifice meter multiphase wet as flow data set from multiple test projects Table 3 shows the data set rane This flow facilities as mass flow rate reference is an 8 as turbine meter with as chromatoraph / PVT packae The root mean square of the test meter ISO dischare coefficient uncertainty (approximately 065%) and the reference turbine meter uncertainty (075%) is approximately 1% Fiure 16 shows sample sinle phase flow data from the 8, beta orifice meter The orifice meter was fully serviceable durin all wet as flow tests 8

9 Fi 12 8 Orifice Meter Fi 13 8 View Port System Fi 14 8, Stratified to Annular Mist Fi 15 8, Annular Mist Flow Parameter Rane Pressure 14 Pressure (bar a) 77 Gas to liquid density ratio 0011 < DR < 0083 Fr rane 055 < Fr < 34 X LM 0 Xlm < 0275 Inside full bore diameter m ( 7981 inch) Beta Gas / Liquid phase Natural as / Exxsol D80 / Water WLRm 0 WLR 1 Table 3 Orifice meter, CEESI, 8 inch, beta Fi 16 Sample Dry Gas Flow Data 9

10 As the orifice meter wet as flow correlation released in 2011 [1] is specifically for 2 Ф 4 this 8 orifice meter wet as data set is outside of it s rane With no published correlation for Ф > 4 operators may chose (or be forced) to extrapolate the existin correlation Therefore, CEESI investiated the effect of extrapolatin the 2011 correlation to this massed 8 data set The results are shown in Fiure 17 Fi 17 CEESI 8 Orifice Meter Data, Uncorrected & Corrected with 2011 Correlation (for Known Liquid Flow Rate) Extrapolatin the 2011 orifice meter wet as correlation for use with the 8, 0869 beta orifice meter wet as flow data produces results out with the correlations stated uncertainty of 2% at 95% confidence The corrected data show a sliht positive bias There were 67 out of the 702 wet as points (ie X LM > 0) outside this uncertainty The spread of the corrected data relative to the as mass flow reference meter flow rate was -23% to +51% With no available alternative it is still beneficial to extrapolate the correlation, althouh the uncertainty has increased There are 35 out of 702 points (ie 5% of points) where the difference between the test and reference as meter as flow prediction is in excess of the rane -2% to +3% Therefore, when applyin the 4 orifice meter wet as correlation to an beta orifice meter the uncertainty is -2% to +3% to 95% confidence All DP meters exhibit a wet as flow over-readin dependency on beta For a iven wet as flow condition the hiher the DP meter s beta the lower the overreadin Orifice meters do exhibit this eneric wet as flow response but compared to other DP meter desins the orifice meter is rather resistant to the effect Steven et al [1] showed the beta influence on 2 Ф 4 orifice meter wet as flow over-readins small enouh to be practically neliible This 8 orifice meter wet as flow data is all from a hih beta of Therefore, if the beta effect was sinificant, this hih beta should produce a lower over-readin, causin the correlation to over-correct the as flow rate prediction However, this result shows a positive as flow rate prediction bias Therefore, the positive bias on the 8 orifice meters correction by the correlation can not be accounted for by a beta effect Orifice meters are often used to meter low liquid loadin wet as flows It is common for the effect of such small liquid loadins to be inored on the assumption that the flow prediction error induced is of the same manitude as the correlations uncertainty 10

11 Fi 18 CEESI 8 Orifice Meter X LM < 002 Data, Uncorrected and Corrected with Existin Correlation (for Known Liquid Flow Rate) Fiure 18 shows 8 orifice meter data for X LM < 002 It is shown that at X LM < 002 the over-readin can still be as hih as 5% Applyin the Ф 4 orifice meter correlation corrects 306 of these 309 points to 2% uncertainty (relative to the as mass flow reference value) Hence, at X LM 002 the existin correlation corrects the 8 orifice data to 2% uncertainty at 95% confidence In may be worth applyin such a correction even for low liquid loadins The existin knowlede on DP meters (includin orifice meters) states that the wet as over-readin: increases as the Lockhart-Martinelli parameter increases reduces as the as to liquid density ratio increases increases as the as densiometric Froude number increases reduces as the WLR m increases For otherwise set wet as flow conditions, Fiure 19 shows a sample 8 orifice meter natural as with liht liquid hydrocarbon (ie Exxsol D80) data set for different as to liquid density ratios For otherwise set wet as flow conditions, Fiure 20 shows a sample 8 orifice meter natural as with water data set for different as densiometric Froude numbers The expected trends are evident As the data was not recorded primarily to investiate the orifice meter multiphase wet as flow response there was no data sets available where the WLR m effect could be effectively isolated It is therefore only assumed likely here that the 8 orifice meter exhibits the same WLR m as other sized orifice meters, and eneric DP meters There is no obvious pattern to the points that fell outside the -2% to +3% uncertainty ratin The Lockhart Martinelli parameter, as to liquid density ratio and as densiometric Froude number ranes all fell within the Ф 4 correlations rane of applicability, as do the fluid types The only extrapolation is the meter size In 2008 Britton et al [4] showed that for set Lockhart Martinelli parameter, as to liquid density ratio and as densiometric Froude number values, as the pipe diameter considered is increased, the as and liquid superficial velocities are chaned This in turn potentially chanes the flow pattern, and by association, the size of meter s over-readin It is postulated that the increase in size from the 2 Ф 4 diameter rane to an 8 meter was enouh to cause this effect to be 11

12 noticeable If this is the case the Ф 4 correlations as flow prediction bias will increase as the meter size increases Fi 19 Sample 8 Orifice Meter Data Confirmin OR% to DR Relationship Fi 20 Sample 8 Orifice Meter Data Confirmin OR% to Fr Relationship 6 WET GAS FLOW AND AN ORIFICE METER DIAGNOSTIC SYSTEM In 1997 de Leeuw [5] released a seminal paper describin a conceptual wet as Venturi meter desin that included a downstream pressure tap For the specific case of a Venturi meter with the specific problem of wet as flow, de Leeuw showed that a Venturi meter s Pressure Loss Ratio, ie the ratio of the permanent pressure loss to the Venturi meter s traditional DP, could indicate chanes in Lockhart Martinelli parameter The DP meter dianostic system Pronosis is a comprehensive eneric DP meter dianostic system (ie it is applicable to all DP meter desins) that also utilises a downstream pressure tap, but to monitor the eneral health of a eneric sinle phase DP meter in multiple ways In this paper the performance of Pronosis when specifically applied to orifice meters in wet as flow applications is reviewed 6a DP METER DIAGNOSTICS In 2009 Swinton Technoloy partnered with DP Dianostics to produce the eneric DP meter dianostic suite Pronosis (see Steven [6,7]) A brief overview 12

13 Fi 21 Orifice meter with instrumentation sketch and pressure fluctuation raph of these pressure field monitorin dianostics is now iven For further technical detail the reader should refer to the full descriptions iven in by Steven [7,8], Skelton et al [2] & Rabone et al [8] Fiure 21 shows a sketch of an orifice eneric DP meter (in this case an orifice meter) and its associated pressure field The DP meter has a third pressure tap downstream of the two traditional pressure ports This allows three DPs to be read, ie the traditional ( P t ), recovered ( P r ) and permanent pressure loss ( P PPL ) DPs These DPs are related by equation 13 The percentae difference between the inferred traditional DP (ie the sum of the recovered & PPL DPs) and the read DP is δ%, while the maximum allowed difference is θ% DP Summation: P = P + P, uncertainty ± θ% --- (13) t r PPL Traditional flow calculation: mtrad = f ( P ) Expansion flow calculation: m exp = ( ) t t f r P r PPL flow calculation: m PPL = f ( P ) PPL, uncertainty ± x% --- (14), uncertainty ± y% --- (15) PPL, uncertainty ± z% --- (16) Each DP can be individually used to meter the flow rate, as shown in equations 14, 15 & 16 Here ṁ trad, m exp & ṁ PPL are the mass flow rate predictions of the traditional, expansion & PPL flow rate calculations Symbols f t, f r & f PPL represent the traditional, expansion & PPL flow rate calculations respectively, and, x %, y %& z % represent the uncertainties of each of these flow rate predictions respectively Inter-comparison of these flow rate predictions produces three dianostic checks The percentae difference of the PPL to traditional flow rate calculations is denoted as ψ % The allowable difference is the root mean square of the traditional & PPL meter uncertainties, φ % The percentae difference of the expansion to traditional flow rate calculations is denoted as λ % The allowable difference is the root mean square of the traditional & expansion meter uncertainties, ξ % The percentae difference of the expansion to PPL flow rate calculations is denoted as χ % The allowable difference is the root mean square of the PPL & expansion meter uncertainties, ν % Readin these three DPs produces three DP ratios, the PLR (ie the PPL to traditional DP ratio), the PRR (ie the recovered to traditional DP ratio), the RPR (ie the recovered to PPL DP ratio) DP meters have predictable DP ratios Therefore, comparison of each read to expected DP ratio produces three dianostic checks The percentae difference of the read to expected PLR is denoted as α % The allowable difference is the expected PLR uncertainty, a % The percentae difference of the read to expected PRR is denoted as γ % The allowable difference is the expected PRR uncertainty, b % The percentae 13

14 difference of the read to expected RPR is denoted as η % The allowable difference is the expected RPR uncertainty, c % Fi 22 Normalized Dianostic Box (NDB) with dianostic results These seven dianostic results can be shown on the operator interface as plots on a raph That is, we can plot (Fiure 22) the followin four coordinates to represent the seven dianostic checks: % φ%, α% a% λ ξ γ, ( ψ ), ( % %, % b% ) ( χ % ν %, η% c% ) & ( δ % θ%,0) For simplicity we can refer to these points as (x 1,y 1 ), (x 2,y 2 ), (x 3,y 3 ) & (x 4,0) The act of dividin the seven raw dianostic outputs by their respective uncertainties is called normalisation A Normalised Dianostics Box (or NDB ) of corner coordinates (1,1), (1,-1), (-1,-1) & (-1,1) can be plotted on the same raph (see Fiure 22) This is the standard user interface with the dianostic system Pronosis All four dianostic points inside the NDB indicate a serviceable DP meter However, for the special case of monitorin the severity of a known problem, such as the liquid loadin of a wet as flow, the reference with which to compare the found performance is arbitrary When monitorin chanes in wet as liquid loadin it is just as valid to use the meter performance at a known finite liquid loadin as the reference as the dry as flow meter performance In this scenario there is no need to normalise the dianostic results, as the uncertainty of the baseline dianostic parameters is not relevant to the task at hand How the dianostic points move relative to chanes in the known problem bein monitored is what is important, not chanes relative to the correctly operatin meter baseline Without normalised data the NDB can be removed We can chose to ψ, monitor raw, ie un-normalised, dianostic points by plottin ( %,α%) ( λ %,γ %), ( χ %,η%) & ( δ %,0) In this case the reference values with which to compare the results are an arbitrary choice In the followin examples the arbitrary liquid loadin value chosen as the baseline was dry as flow 6b PREDICTING THE PRESSURE LOSS RATIO OF AN ORIFICE METER Orifice meters are not typically calibrated For the eometry stated by ISO 5167 Part 2, ISO states the dischare coefficient (via the Reader Harris-Gallaher equation) and the PLR (via the Urner [9] equation) This information allows the flow rate to be predicted (via the RHG equation) and for the meters permanent pressure loss to be predicted (via the Urner equation) Furthermore, combinin the RHG & Urner equations allows all the orifice meter dianostics parameters to be derived { β ( 1 Cd )} Cdβ 4 2 { β ( 1 C 2 )} + C β 1 PLR = --- (17) 1 In order to predict both the orifice meter permanent pressure loss and the dianostic parameter baselines as accurately as possible it is necessary for the PLR prediction equation to be as precise as possible The Urner equation is theoretical It states the PLR is related to the orifice meter beta (β) and dischare coefficient (C d ) The dischare coefficient effect is a second order compared to the beta effect The Urner PLR prediction (equation 17) has been shown to be d d 14

15 accurate up to β 055 (e Steven [10]) However, at β > 055 the Urner equation shows a sliht neative bias In 2012 Steven [10] showed a massed orifice meter PLR vs beta data set Alon with showin the sliht bias in the theoretical Urner prediction a PLR vs beta data fit was presented More orifice meter hih beta data has now been obtained and that data fit can now be updated and improved Equation 18 shows the latest PLR vs beta data fit which has an uncertainty < 3% at 95% confidence, and no bias at β > 055 Fiure 23 shows the data with both the Urner & equation 18 PLR vs β predictions 15 ( * ) PLR = β --- (18) Fi 23 ISO/Urner & New Data Fit PLR vs β Equations Superimposed on Data Set 6c PROGNOSIS AND ORIFICE METERS WITH WET GAS FLOW The BP 4, 0683 beta orifice meter wet as flow data discussed in section 4, and the CEESI 8, 0689 beta orifice meter data discussed in section 5, are now used as case studies to show Pronosis in operation with wet as flow 6c1 BP 4, 0683β ORIFICE METER WET GAS PROGNOSIS EXAMPLES The BP multiphase wet as orifice meter performance tests (see Fiure 11) were conducted primarily to test other equipment The orifice meter and Pronosis performance were secondary considerations The wet as flow data set with which the orifice meter was tested had many varyin parameters In order to clearly show the Pronosis response to varyin wet as liquid loadin it was desirable to find a set of data where only the liquid loadin had a sinificant variation while the as flow rate and pressure remained relatively constant Fiure 24 shows such a data set The averae line conditions were 72 Bar(a) at 40 0 C, with a as flow of 39 k/s (ie approximately 15 MMSCFD) In this data set the liquid was a varyin WLR mix of water and liquid hydrocarbon No MEG was injected The system was tested at Lockhart Martinelli parameters of 001, 002, 004 & 008 As the liquid loadin increases the Pronosis results divere from the oriin, and as the liquid loadin decreases the Pronosis results convere towards the oriin Fiure 24 shows a trendin application type plot of Pronosis data (ie without an NDB or normalised data) This presentation is not the only way the dianostic output could be presented The six dianostics ( ψ %,α%), ( λ %,γ %) & ( χ %,η%) may also be plotted relative to the parameter bein trended, ie in this case the Lockhart Martinelli parameter Fiure 25 shows this BP / CEESI data for 72 Bar(a) at 40 0 C & 15 MMSCFD plotted in this way The data shown in Fiure 24 is included in the larer data set shown in Fiure 25 15

16 Fi 24 Sample Pronosis Data Plot at 72 Bar(a) at 40 0 C & 15 MMSCFD Fiure 25 shows the Pronosis system s sensitivity to wet as flow Each of the six dianostic parameters are sensitive to chanes in the liquid loadin As the Lockhart Martinelli parameter increases so does each dianostic parameter Fiure 25 also indicates that the six dianostic parameter vs Lockhart Martinelli parameter relationships have different radients That is, the six dianostics have different sensitivities to liquid loadin chanes Fi 25 Alternative Pronosis Data Plot at 72 Bar(a) at 40 0 C & 15 MMSCFD 16

17 There are dedicated wet as meter desins that use a Venturi meter in particular, coupled with the PLR vs Lockhart Martinelli parameter in particular to make a wet as liquid loadin monitor The PLR vs Lockhart Martinelli parameter relationship for this orifice meter is shown in Pronosis via α % For the case of applyin Pronosis to orifice meters, whereas the parameter α % is clearly sensitive to liquid loadin, it is not the most sensitive, and therefore not the most useful of the orifice meter dianostic parameters for monitorin liquid loadin The most sensitive, and therefore most useful of the orifice meter dianostic parameters for monitorin liquid loadin are the DP ratios related to the recovered DP, ie γ % & η % These two parameters are not only more sensitive to small chanes in orifice meter liquid loadin than the other parameters (includin α %, ie the PLR), but also continue to see chanes in liquid loadin until hiher Lockhart Martinelli parameters, when the other parameters have radually become insensitive to the increasin amounts of liquid 6c2 CEESI 8, 0683β ORIFICE METER WET GAS PROGNOSIS EXAMPLES The 8 CEESI multiphase wet as flow facility has a resident orifice meter (see Fiure 12) downstream of the meter under test locations This meter has a massed blinded data set obtained by loin data for each and every multiphase wet as meter test conducted over the last three years Due to the lare number of channels required to lo the varyin multiple test parameters from test equipment this orifice meter only has the traditional and PPL DPs direct read When applyin the Pronosis system here the recovered DP is inferred by equation 13 Althouh Pronosis can operate with the third DP bein inferred, more powerful dianostics are available if the third DP is directly read Two sample Pronosis wet as data sets are shown here Fiures 26 & 27 shows two different representations of a as & liquid hydrocarbon data set from the meter at 172 Bar(a) at 40 0 C, with a as flow of 67 k/s (ie approximately 265 MMSCFD) The system was tested at Lockhart Martinelli parameters of 001, 005, 010 & 014 Fiures 28 & 29 shows two different representations of a as & liquid Fi 26 Sample Pronosis Data Plot at 172 Bar(a) at 40 0 C & 265 MMSCFD 17

18 Fi Bar(a), 265 MMSCFD, Alternative Plot of Wet Gas Pronosis Data hydrocarbon data set from the meter at 355 Bar(a) at 40 0 C, with a as flow of 47 k/s (ie approximately 186MMSCFD) The system was tested at Lockhart Martinelli parameters of 002, 005, 010 & 015 As the liquid loadin increases the Pronosis results divere from the oriin, and as the liquid loadin decreases the Pronosis results convere towards the oriin As with the BP 4 orifice meter wet as data % are the two most useful dianostics for monitorin wet γ & η% as flow throuh an orifice meter Fi Bar, 186 MMSCFD, Wet Gas Pronosis Data 18

19 Fi Bar, 186 MMSCFD, Alternative Plot of Wet Gas Pronosis Data 7 PROGNOSIS SENSITIVITY TO WET GAS FLOWS All six dianostic parameters considered in section 6 show a relationship with liquid loadin Hence, the DP meter dianostic system Pronosis can monitor the liquid loadin of a wet as flow As these six dianostic parameters are related, for simplicity, in this discussion the dianostic reaction to wet as will be discussed considerin only one parameter, ie the most commonly known PLR The other five dianostic parameters have associated reactions to wet as flow (of varyin sensitivity) Fi 30 Orifice Meter & Wet Gas Flow PLR vs X LM relationship As the wet as flow s liquid loadin increases the PLR increases This is shown by the parameter α % in Fiures 24 throuh 29 However, the sensitivity of the dianostics to wet as flow is dependent on DP meter desin A eneric DP meter s capability to monitor liquid loadin variations is dictated by its sinle phase PLR performance A hypothetical ideal DP meter with no PPL has a dry as PLR value of zero A DP meter where no pressure recovery takes place, ie none of the standard DP is recovered, has a PPL equal to the standard DP Therefore, this meter s dry as PLR is unity Hence, all real DP meter performances are somewhere between these boundary conditions, ie 0 PLR 1 This statement is independent of the flow bein wet or dry flow 19

20 As wet as flow causes DP meter PLR values to increase, the closer a DP meter s dry as PLR value is to the maximum value of unity, the less resolution there is between wet and dry as PLR values The hiher a DP meter s dry as PLR value, the poorer the DP meter s ability to monitor wet as liquid loadin Fiure 31 indicates the problem Two hypothetical meters are represented, one with a relatively low dry as PLR value (left raph), the other with a relatively hih dry as PLR value (riht raph) As the Lockhart Martinelli parameter increases the PLR s of both meters increase towards unity However, the meter with the hiher dry as PLR has a sinificantly shallower radient When real world error bands are included to represent the uncertainty in the PLR measurement it becomes clear that the meter with the lower dry as PLR value is a more capable wet as flow monitorin system Fi 31 Relative Sensitivity of Low & Hih PLR DP Meters to Liquid Loadin In 2006 CEESI were allowed to release the results of a Joint Industry Project, or JIP, (in which BP were members) where various as meters had been tested with wet as flow All DP meters were initially tested with dry as flow, and the dry as PLR values found Each meter was subsequently tested with wet as These tests included a 4, orifice meter with 034β, 04β, 05β & 068β plates The dry as baseline PLR value of an orifice meter is dependent on the beta used (see Fiure 23) Equations set 17 & 18 show the orifice meter PLR prediction Fiures 32 throuh 35 show CEESI JIP orifice meter data The dry as baseline data fits the predicted baseline performance well The wet as flow tests then showed that for β < 05 the PLR (bein too close to unity ) and the other associated dianostic parameters were insensitive to the liquid loadin For β 05 the PLR (bein further from unity) and the other associated dianostic parameters were reasonably sensitive to the liquid loadin Hence, althouh the dianostic system Pronosis is very sensitive to most orifice meter malfunction issues (β < 05 inclusive), for the particular case of monitorin wet as liquid loadin it is stronly advisable to use an orifice meter with β 05 Fi β Orifice Meter PLR vs X LM Fi β Orifice Meter PLR vs X LM 20

21 Fi β Orifice Meter PLR vs X LM Fi β Orifice Meter PLR vs X LM For β 055: { β ( 1 Cd )} Cdβ 4 2 { β ( 1 C 2 )} + C β 1 PLR = --- (17) 1 15 For β > 055: = ( * β ) d PLR --- (18) This phenomenon is not restricted to orifice meters The ability of any eneric DP meter fitted with the dianostic system Pronosis to monitor a wet as flow liquid loadin is stronly dependent on the meter s dry as PLR value Fiure 36 shows the result of analysin the CEESI JIP data to investiate which DP meters could potentially be used with the dianostic system Pronosis to create a wet as liquid loadin monitor system It was found that: All DP meters with a dry as PLR 055 have reasonable wet as liquid loadin monitorin performance (for X LM 015) All DP meters with a dry as PLR > 075 were not useable as wet as liquid loadin monitorin systems DP meters with dry as PLR s between 055 < PLR 075 may be useable as wet as liquid loadin monitorin systems, dependent on meter desin d Fi 36 DP Meter Dry Gas PLR & Wet Gas Monitorin Capability 21

22 8 CONCLUSIONS The wet as flow performance of orifice meters is of importance to the natural as production industry In 2011 Steven et al [1] released a comprehensive horizontal wet as flow orifice meter correction factor for a nominal diameter rane of 2 Ф 4 This correlation was for natural as with water and / or liquid hydrocarbon The effect of extrapolatin this correlation in terms of liquid properties or meter size was not stated It had been shown in 2011 that fluid properties influence the flow pattern, which in turn influences the orifice meter wet as over-readin Therefore, by alterin liquid properties by flowin liquid hydrocarbon with heavier components (ie waxes ), and / or injectin MEG, it could be expected that the existin as / oil / water correlation may produce a as flow rate prediction bias However, it was found that it did not The additional fluids of MEG and liquid hydrocarbon with heavier components (ie waxes ), where under thermodynamic conditions the fluid had no solids (ie wax contamination), did not induce a sinificant bias on the published orifice meter correction factor output It was shown by Britton [4] for set wet as flow parameters, ie set Lockhart Martinelli parameter, as to liquid density ratio and as densiometric Froude number values, the larer the pipe size the faster the superficial as and liquid velocities Gas and liquid velocities influence the flow pattern The faster the phase superficial velocities the more mixed the phases may be, and the hiher the DP meter over-readin will be (as n ½) In 2011 this phenomenon was shown to be neliible between 2 and 4 orifice meters Prior to analysis it was expected that this phenomenon would still be neliible as the orifice meter size bein considered was raised to 8 However, the results of analysin the CEESI 8 orifice meter massed wet as data suests that for set Lockhart Martinelli parameter, as to liquid density ratio and as densiometric Froude number values an 8 orifice meter does have a slihtly hiher over-readin The 2 to 4 orifice meter wet as correlation therefore produces a sliht positive as flow rate prediction bias Finally, it has been shown that the DP meter dianostic system Pronosis can monitor the wet as flow liquid loadin throuh DP meters However, the effectiveness is directly lined to the DP meters dry as PLR The lower the DP meter s dry as PLR value the more sensitive to wet as flow Pronosis is Due to this phenomenon an orifice meter with Pronosis must have β 05 for wet as flow liquid loadin monitorin to be practical REFERENCES 1 Steven R, Stobie G, Hall A & Priddy R, Horizontally Installed Orifice Plate Meter Response to Wet Gas Flows North Sea Flow Measurement Workshop Oct 2011, Tonsber, Norway 2 Skelton M et al, Dianostics for Lare Hih Volume Flow Orifice Plate Meters, North Sea Flow Measurement Workshop October 2010, St Andrews, Scotland 3 Tin VC et al Effect of Liquid Entrainment on the Accuracy of orifice Meters for Gas Flow Measurement, Int Gas Research Conference, Britton C et al, Liquid Property and Diameter Effects on DP Meter Wet Gas Over-Readins, North Sea Flow Measurement Workshop, St Andrews, UK, October De Leeuw R, "Liquid Correction of Venturi Meter Readins in Wet Gas Flow," North Sea Workshop Steven, R Dianostic Methodoloies for Generic Differential Pressure Flow Meters, North Sea Flow Measurement Workshop October 2008, St Andrews, Scotland, UK 22

23 7 Steven, R Sinificantly Improved Capabilities of DP Meter Dianostic Methodoloies, North Sea Flow Measurement Workshop October 2009, Tonsber, Norway 8 Rabone, Operator Experience with DP Meter Dianostics, North Sea Flow Measurement Workshop Oct 2012, St Andrews, Scotland 9 Urner, G, Pressure loss of orifice plates accordin to ISO 5167, Flow Measurement and Instrumentation, 8, March 1997, pp Steven R et al, Differential Pressure Meters A Cabinet of Curiosities (and Some Alternative Views on Accepted DP Meter Axioms), North Sea Flow Measurement Workshop Oct 2012, St Andrews, Scotland 23