Peter L. Lagus, Brian S. Flanagan
Lagus Applied Technology Inc.
San Diego
Michael E. Peterson
Tennessee Gas Pipeline Co.
Middleton, Tenn.
Sam L. Clowney
Tenneco Gas
Houston
A technique for measuring compressor flowrate through an operating natural-gas centrifugal compressor has been tested and found to have a precision approaching 1.5%. The technique employs constant-flow tracer dilution.
Testing demonstrated that use of a critical-flow nozzle to inject a constant, known flow of tracer into a flowing natural-gas stream is feasible. Effects of potential pulsation on a tracer flow measurement appear to be eliminated by this technique.
With experimental and operational streamlining, the constant-flow tracer dilution technique is capable of being used to measure the flowrate through operating centrifugal compressors with sufficient precision and accuracy to allow actual compressor operating characteristics to be determined.
This technique is especially useful in situations in which an orifice-flow measurement cannot be performed because of physical space limits or economic considerations.
PROGRAM OBJECTIVES
The research program was conducted to demonstrate the viability and utility of a constant-flow tracer dilution technique for measuring the flowrate through a high-pressure natural-gas centrifugal compressor.
Testing was performed by injection of a known flowrate of tracer gas, sulfur hexafluoride (SF6), into a natural-gas stream immediately upstream of an operating centrifugal compressor. Measurements of the diluted tracer gas concentration downstream of the compressor were made by means of electron-capture gas chromatography.
Knowledge of the injection concentration and injection flowrate coupled with the downstream diluted concentration allowed inference of the actual flow through the compressor. Tracer-inferred flowrates were then compared to flowrates measured by a custody-transfer station downstream of the compressor station.
This technique appears to be ideally suited for proving the flow through a compressor or calibrating a flow-measurement device at its installed location. The technique is not suited for use on a continuous measurement device.
TRACER INJECTION
Inference of flowrate by the tracer-dilution technique entails introduction of a constant, known flowrate of tracer into a flowing gas stream and measurement of the equilibrium concentration of tracer downstream of the injection point. Fig. 1 shows the basic idea.
Illustrating the underlying concept involves assuming ideal gas behavior. In this case it can be shown that if one injects a tracer at a constant rate (P), and one measure of an equilibrium tracer concentration (C) downstream of the injection point, then the flowrate (0) is calculated as Q = P/C.
Thus, if one injects, for example, 100 cfm of tracer and measures a downstream concentration of 1 unit, then the inferred flowrate would be 100 cfm; i.e., the initial flowrate of tracer was diluted by a factor of 100.
The actual analysis of the tracer data is only slightly more complicated than this.
For compressor flow testing, SF6 was used as a tracer gas because it has proven to be an ideal tracer for a wide variety of flow-measurement applications. It is inert, nontoxic, nonreactive, and easily detectable in quantities approaching 1 part per trillion (1012 ) by means of electroncapture gas chromatography.
All tracer-gas measurements during testing were performed with S-Cubed tracer-gas monitors. These are measurement-specific, electron-capture chromatographs that have been specifically manufactured and optimized for the detection of SF6.
In order to increase the likelihood of a successful result, tracer-injection flowrate was controlled by a critical flow nozzle which had been specially calibrated for the pressure and flowrates anticipated in the actual compressor testing.
For the measurements performed during testing, different flowrates were obtained by changing the upstream pressure.
FIELD TESTS
During spring 1988, compressor flowrate testing was performed at Pacific Gas Transmission Co.'s Bonanza station near Klamath Falls, Ore.
The station consists of two centrifugal compressors, 14A and 14B, whose specifications are provided in the accompanying box.
Fig. 2 shows the station layout of compressor locations as well as injection and sampling locations.
Testing of Compressor 14B was effected by tracer injection at Point B, with sampling at Point C. For those experiments in which both compressors were operating, tracer injection occurred at Point A, while sampling took place at Point C.
Actual injection and sampling locations coincided with pre-existing valve and tap locations located on the compressor piping. Test Points B and C were, respectively, 20 ft upstream and downstream of Unit 14B.
Compressor flow for the Bonanza site was measured at a custody-transfer station which was located 11 miles downstream of the station. Because only a minimal amount (less than 0.1%) of gas is taken out of the line between the Bonanza station and the custody-transfer station, a very precise measurement of the actual flow through the compressor can be obtained.
All pressure transducers used at the custody-transfer station are calibrated monthly. The test period was chosen to occur immediately after this calibration was performed in order to increase the likelihood that accurate flow data could be obtained.
Compressor flowrate testing was also performed at the Tennessee Gas Pipeline Co. Station 17 near East Bernard, Tex., during the last 2 weeks of January 1990. A single compressor, whose specifications are presented in the accompanying box was tested over several days.
Compressor flowrates were measured by a 30-in. AGA Class III orifice located immediately upstream of the compressor. All pressure transducers were calibrated daily just prior to the actual testing.
Tracer gas was injected approximately 6 ft upstream of the compressor, while the tracer was sampled approximately 6 ft downstream of the compressor.
For both test locations, tracer gas was injected with a manifold similar to that shown in Fig. 2. The tracer gas injection system consisted of a high-pressure (4,000 psig) cylinder of SF6 diluted in nitrogen which was connected to a critical flow nozzle, whose output was directed to the upstream side of an operating centrifugal compressor.
INJECTION, SAMPLING
For testing at Bonanza, tracer gas-injection pressure was measured by a 0-3,000 psi dial gauge with a 0.25% (full-scale) accuracy. The tracer-gas injection temperature through the nozzle was measured with a Type J thermocouple.
During testing at East Bernard, tracer-gas injection pressure was measured with a strain-gauge transducer with an accuracy of 0.1% of full-scale. Temperature was measured with a platinum RTD (resistance-temperature device).
Cylinders of SF6 diluted in nitrogen at high pressure were used as tracer-gas sources. These cylinders were analyzed by an independent specialty-gas house to a precision of 1% in concentration.
Each test was performed according to a defined written procedure. A new procedure booklet was filled out for each test.
Testing was performed by injection of tracer gas at a known flowrate upstream of the compressor. This was done by adjustment of the tracer-injection source cylinder regulator to the desired injection pressure. For most tests, the injection tracer-gas pressure was approximately 1,800 psig.
Sampling was performed by stainless-steel probes inserted into the natural-gas stream through a packing gland. The probe was connected to a shut-off valve, which was in turn connected to a vent line. The vent line itself possessed a septum tee.
Sampling was performed by allowing gas from the downstream side of the compressor to vent to atmosphere for 1 min. No sampling was attempted, however, until 3-5 min had elapsed after initiating tracer flow into the compressor stream.
Syringe samples were then drawn through the septum tee during which time the natural gas continued to vent. Gas samples were withdrawn with polypropylene syringes downstream of the compressor using the sampling line (Fig. 3).
Immediately after the syringe needle was withdrawn from the sampling septum, the needles were plugged with small pieces of rubber so that no leakage or dilution of the sample occurred. The syringes were then taken to an adjacent building in which the chromatographs had been set up.
Prior to the beginning of each day of testing, the calibration of the instruments was checked.
After an individual experiment was completed, a calibration check was performed to ensure that no instrument drift had occurred during the analysis period. Instrument responses were reduced to actual tracer gas concentration by means of calibration equations.
After performance and analysis of an individual test, tracer-inferred compressor flowrates were calculated with Equation 1 and then compared to the custody-transfer station inferred flowrate (Table 1).
DATA AGREEMENT
Tables 1 and 2 present a comparison of tracer inferred and orifice-measured flowrates. The agreement for the East Bernard data is better than for the Bonanza data.
Much of this results from the use of a precise pressure transducer as opposed to a dial gauge to measure tracer injection pressure. In addition, between the Bonanza and the East Bernard tests, all calibration gases were precisely reanalyzed to a precision of 1%.
Initially, the analyses used to generate the calibration curve for the Bonanza testing were precise to approximately 2%. This reanalysis resulted in a more accurate calibration curve for the East Bernard tests.
Test 17 in Table 1 was performed with two compressors (14A and 14B) in series. The error is not significantly different from, say, Test 12. Therefore, the results of Test 17 suggest that significant absorption/solution of the tracer gas in the compressor oil did not occur.
Note, also, that the Bonanza station data appear to possess a slight positive bias, i.e., tracer-determined flow higher than orifice-determined flow. This bias (if any) is significantly less for the East Bernard data.
To investigate if bypass valve leakage could be adversely affecting these results, a test was performed in which tracer was injected at Point C and sampled for at Point A. No tracer was measured at Point A, implying that bypass leakage was negligible.
Possible sources of uncertainty in measurement technique due to sampling, mixing, and tracer-gas adsorption in compressors' lubricity oil were investigated and found to be negligible.
BYPASS FLOW
Another possible source of error is possibly bypass flow in the pipeline due to partial failure of the main line valve.
To investigate this possibility at Bonanza, an additional test was performed in which tracer was injected downstream of Compressor 14B at Point C and sampled at Point A upstream of Compressor 14A.
The existence of any measurable tracer upstream of Compressor 14A would imply that the main line valve was allowing flow past and thus allowing recirculation of gas through the compressor station. No tracer was found, indicating that bypass flow was negligible.
Accurate knowledge of the tracer-injection flowrate is essential. Any inaccuracy in the knowledge of this flowrate would manifest itself as a discrepancy between tracer inferred and custody-transfer flowrate.
The injection flow nozzle was calibrated in a system which is directly traceable to the National Institute for Standardized Technology (NIST). The calibration itself possessed an intrinsic precision of better than 0.5% of reading.
The major uncertainty in the tracer injection flow would be in measurement of pressure and temperature of the flowing gas upstream of the nozzle.
For testing at the Bonanza station, pressure directly upstream of the nozzle was measured by a 3,000-psi dial gauge, while the temperature was measured with a Type J thermocouple.
The pressure gauge was initially calibrated at an independent laboratory and was found to be within manufacturer's specifications ( 10 psi). The thermocouple system was calibrated and capable of providing temperature measurements to within 0.1 C.
For testing at East Bernard, the pressure transducer was calibrated with a deadweight tester and found accurate to within 0.1% of fullscale. The platinum RTD system evinced precision on the same order as the thermocouple system.
The concentration of the SF6 tracer gas injected into the flowing natural gas was independently analyzed at a specialty-gas house by techniques traceable to NIST and possessed a measurement precision of approximately 1 %.
A final possible source of discrepancy is in the actual sample measurement using the gas chromatograph.
In order to produce very precise data, an extensive calibration protocol was undertaken, as described previously. The calibration equations generated for the monitors each possessed a correlation coefficient in excess of 0.99. This indicated that at least 99% of the variation of the data was explained by the functional form of the calibration equation.
Additional testing is planned for the Didsbury test facility operated by NOVA Corp. near Calgary and again at the Bonanza station.
Thus, the major sources of error imply that a precision approaching 1.5% is possible with a single measurement.
While the precision is slightly worse than this for the Bonanza data, it is better than 1.5% on the average for the East Bernard data.
ACKNOWLEDGMENTS
This article is based primarily upon research funded and directed under contracts with the AGA on behalf of its Pipeline Research Committee (PRC).
The work was overseen by the Compressor Research Supervisory Committee of the PRC, chaired by Ralph Hessje, whose support is appreciated.
Thanks are also due to Colton Meyer and Frank Timmons, Pacific Gas Transmission, and to Charlie Gilbert, S-Cubed.
Copyright 1991 Oil & Gas Journal. All Rights Reserved.
