On-line monitoring of H 2 O using tunable diode laser absorption spectroscopy (TDLAS) to optimize recovery of natural gas liquids

Transcription

1 On-line monitoring of H 2 O using tunable diode laser absorption spectroscopy (TDLAS) to optimize recovery of natural gas liquids by W. Gary Engelhart, Product Line Marketing Manager, SpectraSensors, Inc. Brian Spencer, Process Engineer, Enterprise Products Company David Beitel, Sales Engineer, Davis & Davis Company Introduction Raw wellhead natural gas is a complex mixture of methane, hydrocarbon condensates (natural gas liquids NGLs), water and contaminants; hydrogen sulfide (H 2 S), carbon dioxide (CO 2 ), nitrogen, mercury, and other compounds. Natural gas processing involves separating methane (CH 4 ) from NGLs and removing entrained contaminants. The design and complexity of natural gas processing plants varies based upon the end products being recovered and entrained contaminants being removed. Plants designed to recover C 2 + NGLs employ cryogenic processing. 1 Operation and control of molecular sieve dehydration dryer vessels has a direct impact on cryogenic processing equipment in NGL recovery plants. One study noted; The most common disturbances in downstream cryogenic units due to molecular sieve systems occur as a result of having a freshly regenerated bed being brought on-line. 2 Molecular sieve dehydration Sweet natural gas exiting an Amine Treatment Unit is saturated with water vapor. Some water can be removed from the wet gas by passing it thorough a knock-out drum, compression and cooling. Molecular sieve dehydration must be used to obtain the very low H 2 O concentration (< 0.1 ppm v ) required in low temperature and cryogenic processes for NGL extraction and liquefied natural gas (LNG) production. 1,3 The Gas Processors Suppliers Association (GPSA) Engineering Data Book states; Molecular sieve dehydrators are commonly used ahead of NGL recovery plants designed to recover ethane. These plants operate at very cold temperatures and require very dry feed gas to prevent formation of hydrates. Dehydration to a -150 F dew point is possible with molecular sieves. 4

2 Three or four dryer vessels containing molecular sieves are typically operated in parallel with a piping system that allows a saturated adsorbent bed to be taken off line for regeneration with heated gas, as shown in Figure 1. Wet feed gas flows down through the adsorbent bed where mass transfer and adsorption occurs. An adsorbent bed has three operating zones; The top zone is called the saturation or equilibrium zone. The desiccant in this zone is in equilibrium with the wet inlet gas. The middle or mass transfer zone (MTZ) is where the water content is reduced from its inlet concentration to < 1 ppm. The bottom zone is unused desiccant and is often called the active zone. If the bed operates too long in adsorption, the mass transfer zone begins to move out the bottom of the bed causing a breakthrough. At breakthrough the water content of the outlet gas begins to increase 4 Monitoring the H 2 O concentration in the outlet gas from each dryer vessel enables the operator to rapidly detect H 2 O breakthrough in an adsorbent bed and switch gas flow to a dryer vessel with a freshly regenerated adsorbent bed. A general practice is to design the bed to be on-line in adsorption mode for a period of 8 to 24 hours. 4 Rapid detection of H 2 O breakthrough helps prevent gas with elevated levels of H 2 O from entering downstream cryogenic equipment used to separate and recover natural gas liquids. Spent Regeneration Gas Wet Feed Gas Dryers (Regeneration) Dryers (Adsorption) Condensed Liquid Separator Regeneration Heater AX AX AX Regeneration Gas AX Product Gas Figure 1. Molecular Sieve Dehydration System with 3 Dryer Vessels In theory molecular sieve dryers can be run until breakthrough is detected to maximize cycle time and extend the life of the adsorbent bed by reducing the frequency of bed regeneration, which is a major factor in bed degradation and regeneration costs. Most plants operate their molecular sieve dryers on a timed cycle rather than detection of bed breakthrough. One reason for operating on a timed cycle rather than detection of bed breakthrough is the slow response time of certain types of H 2 O analyzers to changes in H 2 O concentration. The lag time associated with these analyzers precludes use of measurements to directly control bed switching and regeneration. Using a timed cycle for bed switching reduces the risk of breakthrough occurring and elevated levels of H 2 O reaching cryogenic equipment before an analyzer detects the condition. The water capacity and adsorption rate of molecular sieves decline as the material ages. The loss of capacity over time varies considerably. A common estimate for this decline in capacity is a loss of 35% capacity over a 3 to 5 year period, or a loss of 50% capacity after approximately 1,600 regeneration cycles. 1 This decay in water capacity is primarily attributable to deposition of carbon and sulfur compound residues causing adsorbent fouling, and bed caking resulting from breakdown of the clay binder. These negative effects are associated with regeneration of the adsorbent bed. 1 2

3 TDLAS analyzer evaluation at Enterprise Products Meeker plant Endress+Hauser approached Enterprise Products and proposed conducting an evaluation of a SpectraSensors TDLAS H 2 O analyzer at their Meeker plant in western Colorado. The Meeker plant processes approximately 0.9 Bcfd of natural gas supplied primarily by Encana, Williams and XTO. The plant has two processing trains, each equipped with three molecular sieve dehydration vessels. Raw natural gas is processed to separate and recover NGLs from the natural gas using a demethanizer. The residue natural gas product is returned to the suppliers of the gas. The NGL output, approximately 50,000 bpd, now belonging to Enterprise is transported via their Mid-America Pipeline (MAPL) to New Mexico where additional NGL product is introduced. The NGL stream is then sent to the Seminole Pipeline in west Texas for transport to Enterprise s storage and fractionation facilities in Mont Belvieu, Texas. The process engineering and management team agreed to evaluate a TDLAS analyzer installed in the common outlet of the molecular sieve dryer vessels in the Meeker II process train. Figure 2. Process train with three molecular sieve dehydration vessels and demethanizer at the Enterprise Products Meeker plant The Meeker plant has used Aluminum Oxide (Al 2 O 3 ) electrochemical sensors to measure H 2 O on the outlets of their molecular sieve dryer vessels. The molecular sieve adsorbent beds are regenerated on a time basis, with the sensors being used for trending measurements rather than for process control. Problems with traditional H 2 O measurement technologies While Aluminum Oxide sensors are relatively simple instruments they have some limitations for trace-level H 2 O measurements in natural gas streams. These electrochemical sensors operate on the capacitance principle with H 2 O 3

4 molecules diffusing through a microscopically porous layer and contacting the surface of the sensor element. H 2 O (and other polar compounds) in contact with the sensor change the capacitance in proportion to the concentration of H 2 O. Because electrochemical sensors depend upon physical diffusion and contact of water vapor with the sensor element, the speed of response can be rather slow. Lag time is most pronounced when the gas stream experiences changes (increases or decreases) in H 2 O concentration. Recommended practice calls for electrochemical sensors to be removed from service for factory calibration on an annual basis. New or recalibrated sensors require hours or days to dry down and stabilize for trace-level H 2 O measurement. There is a further complication to using Al 2 O 3 electrochemical sensors for low level H 2 O measurements in natural gas. When polar compounds (methanol) injected upstream to inhibit hydrate formation, (or ethylene glycol) carried over in gas exiting a field glycol dehydration unit come in contact with the sensor element they cause a change in capacitance that is indistinguishable from the sensor response to water molecules. This measurement bias will falsely indicate a higher concentration of H 2 O than is actually present. An electrochemical sensor s non-specific response to both H 2 O and polar compounds makes their use for trace-level H 2 O measurement highly questionable. Gas processing plants typically use a 3A molecular sieve adsorbent. Methanol will pass through a 3A bed because it does not fit into the pores of the molecular sieves. 5 Consequently, gas exiting a molecular sieve dryer vessel and reaching an Al 2 O 3 electrochemical sensor will contain methanol. These technical issues have been known and documented by molecular sieve manufacturers for over 20 years. A 1996 guidance booklet by one manufacturer states; Instrumentation, Probe type moisture analyzers, such as Panemetrics, are often satisfactory for detecting water breakthrough. However, readings tend to drift with time and temperature, so that absolute dew point numbers may not be correct. The probes should be replaced and/or recalibrated on a regular basis. Methanol, being a polar compound, will also register on these moisture analyzers. As the molecular sieve dryers are sized to remove water, large concentrations of methanol in the feed gas will cause methanol breakthrough in the dryer product gas. Methanol will register on a probe-type moisture analyzer as a lower concentration of water. 6 Quartz Crystal Microbalance (QCM) moisture analyzers have also been used for low level H 2 O measurements in natural gas processing plants. In a QCM analyzer a pair of electrodes apply voltage to a quartz crystal element, which induces an oscillation frequency. The surface of the oscillating quartz crystal is coated with a hygroscopic polymer. When H 2 O molecules come in contact and adhere to the sensor surface, the resulting change in mass changes the oscillation frequency which is measured and used to calculate H 2 O concentration. Unfortunately, entrained polar compounds (methanol, ethylene glycol, and H 2 S) also condense on the QCM sensor along with water. QCM analyzers measure a change in oscillation frequency resulting from a change in mass (H 2 O and other molecules) on the sensor surface. This is an indirect, physical measurement that does not measure the molar concentration of H 2 O and is biased by the presence of contaminant molecules. QCM sensor life and service interval is shortened by the presence of entrained contaminants. To mitigate this problem the support documentation of one supplier advises users to operate in a sensor saver mode where the process gas containing liquid contaminants such as glycol, compressor oils or high boiling point hydrocarbons bypasses the analyzer and reference gas flows through the analyzer. This is done to minimize the time the sensor is in contact with the process gas. Tunable diode laser absorption spectroscopy Many natural gas processing plants have transitioned from older, direct-contact H 2 O measurement technologies to TDLAS analyzers. TDLAS analyzers are designed to selectively and specifically measure H 2 O and other analytes (H 2 S, CO 2, C 2 H 2, and NH 3 ) in hydrocarbon process streams. 7,8 The near infrared (NIR) laser wavelength emitted from a tunable diode laser and used to detect molar absorptivity is identified using special spectral analysis software to optimize analyte measurement in a particular process gas stream such as natural gas. The basic design of a measurement cell in a TDLAS analyzer is depicted in Figure 3. The principal components of the cell are; an optical head housing the laser and thermo-electric cooler (TEC) and a solid state detector, the cell body with a mirror positioned at the end opposite the laser, gas inlet and outlet connections, and temperature and pressure sensors. 4

5 In operation, process gas from a sampling probe is introduced to the sample cell of the TDLAS analyzer. A tunable diode laser emits a wavelength of near-infrared (NIR) light that is selective and specific for the target analyte into the sample cell where it passes through the gas and is reflected back by the mirror to a solid state detector. A window isolates the laser source and solid state detector components from the process gas. This design allows measurements to be performed with absolutely no contact between the process gas (and entrained contaminants) and critical analyzer components. Analyte molecules present in the gas sample absorb and reduce the intensity of light energy in direct proportion to their concentration according to the Lambert-Beer law. The difference in light intensity is measured by the solid state detector and this signal is processed using advanced algorithms to calculate analyte concentration in the process gas. Figure 3. The measurement cell in a TDLAS analyzer The laser wavelength used in TDLAS analyzers to measure H 2 O in natural gas does not have an absorbance interference from methanol. Therefore, methanol present in gas exiting a molecular sieve dryer vessel does not affect the H 2 O concentration measured and reported by a TDLAS analyzer. Figure 4 shows the spectrum of 164 ppm v H 2 O in natural gas overlaid with the spectrum of 164 ppm v H 2 O and 4,000 ppm v methanol in natural gas. The 164 ppm v H 2 O measurement is unaffected by the presence of 4,000 ppm v of methanol in the natural gas sample stream. Figure 4. Spectrum of 164 ppm v H 2 O in natural gas overlaid with spectrum of 164 ppm v H 2 O and 4,000 ppm v methanol in natural gas demonstrating TDLAS analyzer response to H 2 O is unaffected by the presence of methanol 5

6 Differential spectroscopy for trace-level H 2 O measurements Molecular sieve dehydration is a highly efficient means of removing H 2 O from natural gas. Under normal operating conditions gas exiting a molecular sieve dryer vessel has only trace-level (sub-ppm) concentrations of H 2 O. Natural gas streams being processed contain different amounts of NGLs (ethane, propane, butane) in addition to methane. These compounds absorb some NIR energy at or near the wavelength used to measure H 2 O. In some cases the light energy measured by the TDLAS analyzer is attenuated by this background absorption effect. SpectraSensors developed and patented 9,10 a spectral subtraction technique that enables trace-level (sub-ppm) measurements of H 2 O (and H 2 S or NH 3 ) to be made when a process gas sample contains very low levels of an analyte and background gas interferences. In operation the TDLAS analyzer performs a sequence of steps to obtain a zero spectrum and process spectrum that are used to calculate analyte concentration by spectral subtraction as depicted in Figure 5. The zero spectrum is obtained by passing the process gas sample through a high-efficiency dryer which selectively removes the trace analyte, without altering the process gas composition and background absorbance. The analyzer records the resulting dry spectrum of the process gas and automatically switches the sample gas flow path to bypass the dryer and collect the process gas spectrum. Subtraction of the recorded zero spectrum from the process spectrum generates a differential spectrum of the trace analyte which is free of background interferences. The concentration of H 2 O is calculated from the differential spectrum. Gas with H₂O (wet) H₂O - Free Gas (dry) a. Process Gas Spectrum with H2O b. H2O - Free Zero Spectrum Differential Measurement a - b = H2O Spectrum Figure 5. Differential Spectroscopy using Spectral Subtraction A particular advantage of the Differential Spectroscopy technique for trace-level H 2 O measurements in natural gas streams is that it dynamically corrects for changes in gas stream composition, temperature, or pressure, when spectral distortion from such changes exceed preset values in the analyzer firmware. Factory calibration and field validation Every SpectraSensors TDLAS analyzer is factory tested and calibrated using a test mixture blended to simulate the customer s process gas stream. The dilution ratio of the standard and mixing ratio of the background stream gases are controlled by digital mass flow controllers with NIST certifications. 6

7 The resulting calibration report is included in the documentation package shipped with the analyzer. Customers can elect to have a Factory Acceptance Test (FAT) to witness analyzer calibration at SpectraSensors manufacturing site in Rancho Cucamonga, California. The solid state laser and detector components used in TDLAS analyzers are intrinsically stable, so no field calibration is required over the lifetime of the analyzer. Users perform periodic validation checks to verify the analyzer is operating properly within its factory-certified calibration range and ensure measurements are accurate for process control. Validation of analyzers for trace-level measurements (0-10 ppm v ) is particularly important because the analyte (H 2 O in this case) being measured will not be present for extended periods of time under normal process operating conditions. The analyzer will read zero ppm. The purpose of validation is stated in section 3.3 of API RP 555; Validation is observing and noting the difference (if any) between the analyzer reading and the agreed analysis of a standard introduced into the analyzer, but with no adjustment made to the analyzer. 11 The term accuracy is defined by the American Society of Testing and Materials (ASTM) as the closeness of agreement between a measurement result and an accepted reference value. 12 NIST-traceable standards provide an accepted reference value for establishing the accuracy of a measurement. Unfortunately, cylinders of NIST-traceable certified reference gases are not commercially for trace-level (low ppm to sub-ppm) concentrations of certain analytes, e.g; H 2 O, NH 3, C 2 H 2, etc. To address this situation SpectraSensors TDLAS analyzers for trace-level measurements are equipped with a permeation system to perform automated validation checks at user-designated time intervals. This enables users to verify analyzer performance and accuracy in the field when certified reference gas standards are unobtainable. The interior of a TDLAS analyzer configured for trace-level H 2 O measurement in natural gas exiting a molecular sieve dryer vessel is shown in Figure 6. This analyzer is equipped with a permeation tube to perform automated validation checks. TDLAS analyzers are built inside enclosures certified to comply with hazardous area classifications of natural gas processing plants around the world. SpectraSensors Application Note details performance specifications for TDLAS analyzers in this application. 13 Figure 6. Interior View of a TDLAS Analyzer for Trace-level H 2 O Measurement 7

8 Observations and findings from evaluation at Enterprise Products Meeker plant Performance of the SpectraSensors TDLAS analyzer was evaluated in side-by-side testing with an existing GE Panametrics Al 2 O 3 sensor at the common outlet of a 3-bed molecular sieve drying system. Figure 7 shows the H 2 O concentration measured over a 6-month period by the two analyzers. The data plotted in green is from the GE Panametrics Al 2 O 3 sensor, and the data plotted in blue is from the SpectraSensors TDLAS analyzer. Some noteworthy differences can be observed between the two data plots. In the center of the graph there is a major spike (in green) of the H 2 O concentration as measured by the GE Panametrics Al 2 O 3 sensor. This spike (circled in red) followed preventative maintenance to replace the Al 2 O 3 sensor. The response of this new sensor incorrectly indicated a high level of H 2 O was present and remained present for several days until the sensor element dried out. During the same time period the SpectraSensors TDLAS analyzer did not show any excursion in the H 2 O concentration, providing a more accurate indication of the actual H 2 O concentration. The graph also shows two time periods (circled in yellow), where no H 2 O concentration is registered by the GE Panametrics Al 2 O 3 sensor. It is not immediately clear why no H 2 O is being detected / reported. This may be attributable to the use of dew point measurement and conversion calculations at extremely low H 2 O levels by the Al 2 O 3 sensor, versus direct measurement of molar absorptivity to detect and report the H 2 O concentration used in the TDLAS analyzer. Figure 7. H 2 O concentration in gas at common outlet of Meeker II molecular sieve dryers measured over a 6 month period by GE Panametrics Al 2 O 3 sensor (in green) and SpectraSensors TDLAS analyzer (in blue) Conducting the evaluation enabled process engineering personnel to calculate projected cost savings attainable by installing TDLAS H 2 O analyzers at the Meeker plant. The projected cost savings are derived from gains in operational efficiency in several aspects of molecular sieve dryer operation. Increased adsorbent life As noted previously the water capacity and adsorption rate of molecular sieves decline over time. This decline in capacity and efficiency is attributable to secondary effects of high temperature regeneration. The TDLAS analyzer evaluation demonstrated the fast response time of the analyzer to H 2 O concentration changes, and that the measurement was unaffected by methanol or glycol in the gas stream. These performance characteristics provide more reliable data for process control and optimization. 8

9 The Meeker plant replaced the adsorbent in their molecular sieve dryers shortly before the TDLAS analyzer evaluation was conducted. The adsorbent supplier has an on-line software tool to help plants assess and optimize adsorbent performance and life by monitoring process parameters; gas pressure, temperature, flow rate, and H 2 O concentration. By utilizing this process optimization software and data from a TDLAS analyzer the Meeker plant expects to be able to extend the operating (drying) time of each molecular sieve dryer bed from 34 to 48 hours before regeneration. Operating on this basis will extend the life of the adsorbent by 40%. This translates into an annual savings of $70,000 for a 3-bed dryer system based upon an estimated adsorbent replacement cost of $500,000 per dryer and the current 6 year replacement cycle. Decreased regeneration costs Extending the operating time of each molecular sieve dryer from 34 hours to 48 hours decreases the regeneration time to dry the adsorbent by 40%. This translates into an estimated annual fuel consumption reduction of 9,332 MMBTU/ year. Using an estimated fuel cost of $4/MSCF discounted 30% (because Enterprise is using its own fuel) the estimated annual savings from reduced fuel usage for bed regeneration is $11,200. Avoiding freeze-up events in cryogenic equipment An undetected H 2 O breakthrough from a molecular sieve dryer will introduce elevated levels of H 2 O to cryogenic equipment and can cause freeze-up and interrupt process operation. Based upon historical operating data for this plant a breakthrough event could be expected to occur and cause the plant to operate in a dew point mode an average of 19 hours per year to remove hydrates from cryogenic equipment. Operating a plant with a capacity of 700 MMSCFD in dew point mode rather than normal cryogenic recovery mode for a period of two days incurs extra costs of approximately $150,000. Projected annual operational cost savings The evaluation conducted at the Meeker plant demonstrated that the more accurate and reliable data obtained from a TDLAS analyzer used in conjunction with an on-line molecular sieve optimization software program can deliver significant operational cost savings. In this case the combined annual cost savings is estimated to be over $230,000. Next steps Both of the gas processing trains at the Meeker plant have three bed molecular sieve dryer systems. The recommendation is to install two SpectraSensors TDLAS analyzers in each train. One analyzer will be installed on the common outlet of the molecular sieve dryer system. A second analyzer equipped with a three stream sample switching system will be installed and programmed to switch between the outlets of the three dryer vessels. Both analyzers will be installed inside heated cabinets for protection from winter weather conditions that can reach -40 F at the plant site in Colorado. Based on the results of their evaluation, process engineering and management personnel at the Meeker plant have also recommended SpectraSensors TDLAS H 2 O analyzers to the Enterprise Products Pioneer plant in Opal, Wyoming. In February 2018, Enterprise Products placed an order for two TDLAS analyzers, heated cabinets, a stream switching system, sample probes, and heated sample transfer tubing lines. Summary On-line monitoring of the H 2 O concentration in gas exiting molecular sieve dryer vessels helps detect moisture breakthrough and prevent gas with elevated levels of H 2 O from entering cryogenic equipment used for separation and recovery of NGLs. 9

10 The fast response time of TDLAS analyzers to changes in H 2 O concentration is an important performance characteristic for detecting breakthrough in molecular sieve dryer beds. TDLAS analyzers selectively and specifically measure the molar absorptivity of H 2 O in natural gas, and are unaffected by the presence of methanol and glycol. Consequently trace-level H 2 O measurements from TDLAS analyzers are more accurate than those obtained from analyzers with sensing elements in direct contact with natural gas and entrained contaminants. An in-field evaluation at the Enterprise Products Meeker plant demonstrated that data from a TDLAS analyzer used in conjunction with an on-line molecular sieve process optimization software program can improve process efficiency resulting in significantly lower operating costs. References 1. Fundamentals of Natural Gas Processing, CRC Press, Taylor & Francis Group, Designing Molecular Sieve Dehydration Units to Prevent Upsets in Downstream NGL/LPG Recovery Plants, 62nd Laurance Reid Gas Conditioning Conference, Univ. of Oklahoma, Natural Gas Processing Technology and Engineering Design, Elsevier, Engineering Data Book, Twelfth Edition, Gas Processors Suppliers Association, Optimal Design and Operation of Molecular Sieve Dehydration Units Part 1, R.H.M. Herold, S. Mokhatab, Gas Processing, July/August, Operating a Cryogenic Molecular Sieve Dryer, Booklet 3, UOP, U.S. Patent 6,657,198 B1 8. U.S. Patent 7,679,059 B2 9. U.S. Patent 7,704,301 B2 10. U.S. Patent 7,819,946 B2 11. API Recommended Practice 555 Process Analyzers, Downstream Segment, 2nd Edition 12. ASTM E Standard Terminology Relating to Quality and Statistics 13. SpectraSensors Application Note Water Measurement in Molecular Sieve Dryer Vessel Outlets 10

11 Contact 4333 W. Sam Houston Pkwy N. Suite 100 Houston, TX Tel Fax WP On-line Monitoring of H2O Using Tunable Diode Laser Absorption Spectroscopy (TDLAS) to Optimize Recovery of Natural Gas Liquids