Você está na página 1de 11

Testing of Small Differential Pressure Cone Meters on Wet Gas

Philip A Lawrence Director of Business Development Cameron Valve & Measurement Systems & Chairman of ISO TC193Sc3 (First Published March 2011)

Abstract Wet Gas is becoming more prevalent in modern oil and gas production systems due to the decline in the world wide hydrocarbon reservoir slate. Measurement Techniques and Methodology also have to be modified to compensate for this change. Whereas single phase differential pressure measurement instruments are accurate where the gas fluid measured is dry (pipeline quality), when liquid is entrained in the gas the common phenomena of gas fluid over-reading first mentioned by Murdoch many years ago is seen to occur. A differential pressure meter measuring a 2-phase flow of gas and liquid experiences this excess differential pressure due to the liquid, which causes the meter to over-register. Correction methods have been developed over the years by Murdock, James, Chisholm, DeLeeuw, Steven, and others. A 2008 paper described an improved correction method based on Chisholm's algorithm, modified by Steven (ref [1]), applicable to orifice meters and D.P. cone meters. This Chisholm-Steven 2008 (herein the C-S 2008) correction is the current state of the art. The C-S 2008 correction was originally developed for 4 inch and 6 inch cone meters of a 0.75 beta ratio. However, a meter size most commonly used in the industry in the USA is nominally 2 inch, for which little experimental data exists. The object of this research and this subsequent paper was to determine to what extent the C-S 2008 equation is valid for 2 inch 0.75 beta generic cone meters. To this end, a generic D.P. Cone Meter of beta ratio 0.75, and a bore of 1.969 inch, serial number 200902-0042, was tested on mixtures of natural gas and Exxsol 80 at pressures between 185 to 790 PSIA, with gas, liquid flow-rates, and density ratios chosen to result in Lockhart-Martinelli parameters (a measure of wetness XL.M.) covering a range from 0 to 0.3 XLM. Testing was carried out at the Colorado Engineering Experimental Station, Inc (CEESI), Nunn, Colorado between 23-Feb-2009 and 26-Feb-2009 and also further testing was done at the same th th facility between the 14 of December and the 16 of December 2010. The concept of using single phase meters in multiphase environments is not new and was first discussed and postulated by J.W. Murdoch in 1962 an American engineer working on orifice plate D.P. over reading effects due to liquid loading in pipes. The subsequent technical document was entitled ;"Two-Phase Flow Measurement with Orifices", Journal of Basic Engineering, pp.419-433, 1962. Other wet gas researchers mentioned in the abstract have contributed to the development of the subject and a plethora of data and correlations exist to suit different metering types concepts and

installations, the major metering devices being used are of a differential pressure types due to the robustness of the design. The subject is hard to deal with because of the lack of test data available to the market place, this sometimes results in data being kept in house and confidential, also the inability to produce a coherent test condition in the laboratory that will match the in field location is also a big issue. Multiphase and wet gas Laboratory data shows that even with a well managed laboratory offering various multi-product fluids, at differing pressures it may be impossible to match the in-field condition which means that any meter correlation or correction algorithm formed outside of the laboratory conditions from which a wet gas meter is tested may have uncertainty issues This is not all doom and gloom it is possible to work with data sets that are not exactly ideal, but caution must be taken and the expected metering system uncertainty may need to be relaxed to allow a sensible non-stressed operation in the field. What is Wet Gas? The term is used to denote a natural gas flow containing a relatively small amount of free liquid by volume, usually this may be limited up to about 10%. There are presently few techniques or methods available which can measure this type of fluid regime to a reasonable degree of accuracy. Wet Gas is considered to be a subset of general multiphase metering The phenomenon of wet gas may occur in several ways. For example:

a) Over time as dry natural gas wells age, changes in flow conditions including a reduction in
line pressure may result in the heavier hydrocarbon gases condensing in flow-lines and transportation pipelines.

b) Production wells for gas condensate fields usually may have wet gas flow. c) The quantity of lift gas injected to increase production from many oil wells can bring them to
flow conditions can be termed as a wet gas stream.

Many gas wells worldwide are now approaching the latter stages of their production life making wet gas metering more common and driving meter manufacturers and users to new ideas and methodology. ISO TC 193 Wet Gas Definition. An ISO DIS - 14532 Standard (terminology Wg 4) sights the following wet gas definition: Wet Gas is defined as gas with inclusion of desirable or undesirable components like water vapour, free water and / or liquid hydrocarbons in significantly greater amounts than those quoted for pipe line quality natural gas. Typically wet gas may consist of unprocessed, (well head) or partially processed natural gases and may also contain condensed hydrocarbon, traces of carbonyl sulphide and, process fluid vapour such as methanol and glycol.

An approximate guideline regarding quantities of liquid per known standard volumes is indicated in figure 1.0 below;Application Dry gas Gas from separator Gas from slug catcher Wet gas production Liquid / Gas production Bbl/MMSCF 0 0-1 0-5 0-20 >20 Mass Ratio 0 .75% 3.7% 13% >13%

Figure 1.0 Some Typical Liquid/Natural Gas Loading Ratios. International Standards Meter performance requirements in the wet gas arena are not covered fully in current standards but an API recommended practice (RP 85) describes the use of wet gas meters in an allocation system. Representation of the fluid velocities, meter types, measured volumes, and mass have also not been exactly defined and various regions of the world use different terminology to obtain a measurement result. This can add some confusion and sometimes many discussions between interested parties ensue world-wide. Current trends indicate approximate ranges of liquid/gas ratios found in most producing gas fields as having GVF > 90 % or a more defined method is to use the Lockhart-Martinelli parameter to indicate a wetness ratio in a wet gas stream usually to a maximum of around 0.3. This measure of gas wetness using the Lockhart-Martinelli Parameter Xlm , is shown mainly in the form of equation # 1.0 below.

Xlm =

qmliq qmgas

gas liq

) ..eq 1.0

Flow pattern maps and charts may also be used to try to predict the expected performance of a specific D.P. flow meter based on the type of regime present Figure 2 shows the expected regimes based on GVF (Gas Void Fraction)

Figure 2 - Flow Pattern/Regime Map (courtesy ISO-TC193 WG1).

Response of a Single Phase Meter on a Wet Gas Regime The overall response of a differential pressure meter illustrated by the Murdock Plot (mentioned earlier) originally developed for the orifice plate D.P Meter - Figure 3.

Figure 3 - The J.W Murdoch Orifice Plot (Over-reading / X LM)

Over-reading and Corrective Algorithms Similar plots can be developed for other types of meters. DP meters as a class exhibit the same overall response; that is the error curve is linear though the slope of the line varies with meter type (Orifice, Venturi, Cone, etc). Work originally by Chisholm and later improved by Steven has resulted in the Chisholm-Steven (CS) correction algorithm [2, 3]. The main drawback of the CS method is that the gas/liquid ratio (liquid load) must be known in advance. Although the liquid load can be found by means of test separators or dye trace methods, it is valid only as long as the gas-liquid ratio doesn't change. A true multiphase meter would need to adapt automatically to changes in the liquid loading ! In a 2007 presentation to the members of the multiphase measurement Joint Industry Project in Estes Park Colorado USA (JIP), Dr Richard Steven described progress to date, and also noted an observation by DeLeuw that the Lockhart-Martinelli parameter XLM ought to be predictable from a second P measurement taken downstream of the flow meter at around 4-5 Ds(see fig 4). Steven then went on to describe how a differential pressure flow meter could take advantage of this phenomenon to become, in effect, a true multiphase flow meter, capable of adapting to varying liquid load. Most multiphase testing to date has been done mainly with 3 and 4 inch meters; there is very little published data for 2 inch meters (the JIP presentation mentions only a single data set). However a majority of differential pressure flow meters in actual service world-wide are a nominal 2 inch. In Feb 2009, Cameron conducted an initial multiphase test of a 2 inch, 0.75 beta cone meter on the CEESI wet gas test loop.

The predominant fluid was natural gas and the liquid phase was Exsol-80, a light hydrocarbon liquid. A 0.75 beta ratio was chosen based on the established performance of larger cone meters; it appears to be close to optimum for this meter type. Data from the Feb 2009 test was used in a multi dimensional regression exercise , the result of which is simple equation that predicts Xlm from the Pressure Loss Ratio (PLR) and Gas/Liquid Density Ratio (DR) with an overall accuracy of about 5%. This preliminary XLM prediction equation, combined with the Chisholm-Steven wet correction algorithm, was implemented in a Scanner 2000 type flow computer firmware version 4.04. We indicated in a previous paper at that time there may have been some limitation to the prediction, as it was based on the initial data from a single meter, and the prediction repeatability from meter-tometer was under investigation. To further the investigation and get more data another 2 inch meter was constructed to a different pipe schedule and thus a different internal diameter and a decision made to go back to CEESI in 2010 The 2009 low velocity (20 ft/sec) data was found to be very inconsistent, and had to be discarded; all we could say with any confidence was that the minimum flow velocity for 2 inch meters lay somewhere between 20 and 50 ft/sec (6 and 15 m/sec).

Figure 4 - A second dP measurement taken 5 Ds downstream of the flow meter. Initial CEESI Testing 2009 The over-reading / L.M. parameter plot for the 2009 testing is shown next in Figure 3 the test used a standard scanner flow computer as a data collection tool after which manual number crunching was used to derive the over reading curves/plots. See Figure 5

Figure 5- Initial 2009 Over-read / XLM data set plots Uncorrected & Corrected to within 5% CEESI 2010 Testing The ultimate purpose of the test conducted in Dec 2010 was to ;(a) Extend the data set to a second 2 inch meter. (b) Investigate the effect, if any, of a slightly different meter tube diameter (ie ; sch 40 vs sch 80,) (The difference being about 0.13 inch, or about 6% of the diameter). (c) Obtain new data to further help refine the XLM prediction algorithm. (d) Narrow down/improve on the boundary of the minimum flow velocity. The ultimate goal of Camerons development effort is to offer a small, inexpensive, multiphase measurement package consisting of a nominal 2 inch (50 mm), 0.75 beta ratio Cone meter and flow computer, which together provide measurement having an overall uncertainty of 5% or better, over a specified range of operating conditions. (The regression coefficients derived from the test will not be shown
as they are propriety)

Test Meter A Cone meter S/N 02-API-0.75; 600 ANSI flange, schedule 40 meter tube, with a Meter tube diameter = 2.067 inch; Cone diameter = 1.367 inch; Beta ratio = 0.75 Scanner 2000 flow computer; dP range 0 to 400 inwc; P range 0 to 1500 psi; Sensor S/N MV06E1027B; Standard platinum RTD [12] Recovery Transmitter: Fuji FKCT35V5AEAYYAA; D.P range 0 to 400 inwc; S/N AAL4551F CEESI Test Equipment Figure 6 & 7 - Stacked Rosemount dP transmitters, 0-300 inwc. and 0-30 inwc - Transmitter stack on both the meter DP and recovery DP - Rosemount pressure Transmitter - Data acquisition system that recorded readings from the CEESI transducers, and also from the analog output of the Scanner 2000 flow computer. The analog signal was assigned to the corrected volume flow rate, so that it could be compared in real time to the known flow rate.

Figure 6 The CEESI Wet Gas Test Loop.

Figure 7 Data Communications and Interface Layout

Fluids Used for the 2010 test Gas phase: natural gas composition analysis was given and inputted into the scanner 2000 AGA detail method equation of state. Liquid phase: Exxsol 80. Note the test loop can also use water and mixed hydrocarbon-water, but only during the summer months. CEESI Laboratory Test Activities 13-Dec-10 During the first day we spent the first day on mechanical set up: Connecting the meter, transducers, leak testing, electrical connection & testing, etc,(See Figures8-9). The S2000 flow computer configuration included the following set up elements :- Reynolds Number/Discharge coefficient table: ten points taken from a gas calibration of Oct-2010. - Gas Composition - data supplied by CEESI. - Cone meter diameters and other parameters. - Enabling the pressure recovery calculation via Modbus-Registers.

Figure 8 - Cone Meter with Scanner 2000.

Figure 9- Transducers + Scanner.

14-Dec-10 : The Scanner expansion boards analog input refused to read the recovery DP signal from the Fuji transmitter , it took a last-minute scramble to reconfigure the Scanner 2000 and by reconfiguring and the scanner expansion board's second analog input we were ready for the first test runs in the early afternoon. The initial test matrix as used: Velocity 20, 50, and 80 ft/sec (modified to 30, 50, and 80 ft/sec, see below) Pressure 200, 500, and 800 psi Liquid loading, ,XLM = 0, 0.3, 0.25, 0.2, 0.15, 0.1, 0.05, 0.02...for a total of 72 data points For each point, data was collected for 5 consecutive minutes. Each point began and ended on a minute boundary. With each change in flow velocity, pressure, or liquid load, the loop was run until flowing conditions stabilized, typically after 3 to 5 minutes. All dP transducers were zeroed each time the pressure was increased.

We completed the first set of test points, at 200 psi, 80 ft/sec, at all liquid loads. Results of the previous (Feb 2009) tests showed that the 2 inch meter exhibits some inconsistent behavior at a normal minimum velocity of 20 ft/sec (6 m/sec). We requested that for this test the minimum velocity be increased to 30 ft/sec (9.1 m/sec) so that we could get some idea of the meter's "bottom end", ie: the minimum velocity at which the pressure recovery method has satisfactory performance. 15-Dec-10 & 16-Dec-10 Final test flow rates at different pressures were taken and are shown in the chart Figure 10

15-Dec-10 200 psi, 50 ft/sec 200 psi, 30 ft/sec 500 psi, 80 ft/sec
Figure 10 Observations:

16-Dec-10 500 psi, 50 ft/sec 500 psi, 30 ft/sec 800 psi, 80 ft/sec 800 psi, 50 ft/sec 800 psi, 30 ft/sec

The intent of the testing was to gather regression data, so the tests were conducted over the widest possible range. Since the Scanner 2000 does not currently support DP "stacking", the dP range was chosen to cover the highest expected D.P. over all tests. This dictated the use of a 0-400 inwc transducer, which is not optimal for measuring the lower DPs encountered during the test. In consequence, the Scanner 2000 dP reading fell below 25 inwc on 27 tests (which is about 1/3 of them). In field use the dP range would need to be be better optimized. The current multiphase correction is based on coefficients for a meter having a significantly smaller diameter. The worst case uncorrected errors were slightly above 20% with larger errors tending to coincide with lower DPs. possibly due to transmitter hysteresis. Preliminary Regression The Dec 2010 data was fitted to the same equation as the 2009 data set. As before, low velocity data was inconsistent and was excluded. 2009 Equation

Ln (PLR ) Xlm = Exp p + q DR + r PLR 2


where

Density Ratio

DR =

gas liq
dP5d dPvc

Pressure Loss Ratio

PLR =

This resulted in a similar response surface - Color Topograph below Figure 11

Figure 11 ; Plot of XLM from Pressure Loss Ratio PLR and Gas-Liquid Density Ratio DR

Conclusions The fitted coefficients are close to those obtained from the 2009 data. The use of the system as a wet gas device will be tested in the field in February 2011 Initial in house testing shows a viable technology. A gas calibration is most important to determine the downstream recovery pressure The cone meter behaves according to previous larger D.P wet gas meter tests (Murdoch) based on both over reading versus XLM tests. Next Steps a) Further work on the PLR-DR regression using the 2009 data, the 2010 data, separately and in combination. Investigate other correlation equations. b) Test the correlations against the 2009 and 2010 data. (c) Improve the liquid density correction. The XLM equation is based on the gas-liquid density ratio, both densities being at flowing conditions. The gas density can be accurately predicted using AGA-8; however the liquid density is more complex because it is affected by temperature and by the absorption of gas, which is largely a function of the pressure. (d) A correlation for the dry gas DP. recovery may be useful in establishing limits of the DP recovery correction.

References
Lawrence PA & Steven R Research Developments In Wet Gas Metering with V-Cone Meters Kinney J ISHM Class # 1320 Wet Gas Measurement ISHM O.K. USA NSFMW 2003 2006 2006 2007 2007 2007 2009 2010

ISO TC 193 WG 1.0 Allocation Metering in the Upstream Area (white paper) Steven R A Discussion on Horizontal Wet Gas D.P. Flow Meters St Andrews Scotland UK..NSFMW Lawrence PA Wet Gas Measurement ISHM Class 2007 #1320 Wet Gas Data Release Estes Park Colorado (CEESI - Wet Gas Laboratory Nunn CO). Wet Gas test Data on a 2 inch cone meter Courtesy Cameron Houston Inc - Lawrence -Braid -Madden Wet Gas Data on a 2 inch cone meter CEESI December 2010 Courtesy Cameron Canada Braid /Madden

Você também pode gostar