Evaluating Natural Gas Emissions from Pneumatic Controllers from Upstream Oil and Gas Facilities in West Virginia

Abstract

In April of 2018, an optical gas imaging (OGI) and full flow sampler (FFS) emissions measurement study of pneumatic controllers (PCs) was conducted at 15 oil and natural gas production sites in West Virginia. The objective of the study was to identify and characterize PC systems with excessive emissions caused by maintenance issues or nonoptimized process conditions. A total of 391 PC systems were found on the sites and all were classified by the operator as snap-acting (on/off) intermittent venting PCs (IPCs) that should exhibit little gas release while the PC is closed between actuation events. The population was comprised of two groups, 259 infrequently actuating, lower emitting (LE) IPCs and 132 gas processing unit (GPU) liquid level IPCs and associated dump valve actuators that vent more frequently and have larger emission volumes. Using a PC-specific OGI inspection protocol with an assumed whole gas OGI detection threshold of 2.0 scfh, only 2 out of 259 LE-IPCs exhibited OGI detectable emissions indicating good inspection and maintenance practices for this category. Due to combined (ganged) GPU exhaust vents, the OGI inspection of the GPU liquid level IPCs was comparatively less informative and determination of single component IPC emissions by the FFS was more difficult. The time resolved FFS measurements of GPU IPCs defined three categories of operation: one that indicated proper function and two associated with higher emissions that may result from an IPC maintenance or process issues. The overall GPU IPC emission distribution was heavy tailed, with a median value of 12.8 scfh, similar to the 13.5 scfh whole gas IPC emission factor (EF). Total emissions were dominated by non-optimal temporal profile high-emitter IPC cases with the top 20% of IPC systems accounting for between 51.3% and 70.7% of GPU liquid level IPC emissions by volume. The uncertainty in the estimate was due to the ganged nature of the GPU exhaust vents. The highest GPU IPC emission came from a single malfunctioning unit with a measured whole gas value of 157 scfh. Up to six IPCs exceeded 100 scfh. An analysis of FFS emission measurements compared to liquids production per IPC unit employed indicated that production sites operating at a high level of liquids production test the limits of the site engineering, likely resulting in higher IPC emissions. Overall, this study found that the LE-IPCs with OGI-verified low closed bleed rates may emit well below the IPC EF while GPU liquid level IPC systems are likely well represented by the current IPC EF. IPCs that are experiencing a maintenance or process issue or that are operating at sites with a very high product throughput per IPC employed can emit at rates exceeding ten times IPC EF.

1. INTRODUCTION

In the United States (U.S.), petroleum and natural gas systems combined represent the largest source of CH₄ emissions, a potent greenhouse gas, by inventory estimates (EPA, 2020; EPA, 2021a; EPA, 2021b). Measurement studies indicate that U.S. oil and natural gas (ONG) production emits greater than 7.0 teragrams of CH₄ per year, significantly exceeding emissions inventories, which may not properly account for abnormal operating conditions (Alvarez et al., 2018). ONG production also emits volatile organic compounds (VOCs) that can impact local and regional airsheds if not properly controlled (Bari and Kindzierski, 2018; Lyon et al., 2016; McMullin et al., 2018). As a component class, PCs represent a significant emission category for ONG production sites. For oil production, emission inventory estimates indicate that PC emissions constituted 53% of the total petroleum system CH₄ emissions for the year 2018, which is more than all other petroleum system activities combined (EPA, 2020). PCs in natural gas production are the third largest source of CH₄ emissions, responsible for 18% of the total estimated national CH₄ emissions (EPA, 2020).

Natural gas-driven PCs are used to automate production site process variables, such as pressure, temperature, and liquid level (Simpson, 2014). The PC pilot is the sensing body that is used to initiate or terminate a mechanical action. The dump valve actuator is the device that converts the gas signal from the pilot into the mechanical action for the purposes of this article (also called the actuator). As a normal part of operation, PCs emit a fraction of the production field natural gas produced on the site to the atmosphere. Field natural gas is primarily comprised of CH₄, however VOCs and hazardous air pollutants, such as benzene, are also emitted, with the specific composition varying by basin (Brantley et al., 2015; Eisele et al., 2016). While PCs designed to prevent field natural gas emissions are increasingly employed, most U.S. ONG production sites continue to use natural gas-venting PCs (EPA, 2020).

Potential emissions from PCs are determined by site engineering, process operational condition, and production rate. The service type, physical design, and maintenance condition of the specific PC pilot and dump valve actuator also affects emission levels (Simpson, 2014; Allen et al., 2015; D’Antoni, 2018; Luck et al., 2019; Stovern et al., 2020; Thoma et al., 2017). When the need for a process adjustment is sensed, the natural gas-driven PC actuates a valve or other system to cause the process change, using field natural gas (supply gas) as motive force. Intermittent venting PCs (IPCs) are designed to vent natural gas only during actuation and can be configured in snap-acting (on/off) or throttling mode, based on service need (Simpson, 2014). The current U.S. EPA whole gas emission factor (EF) for IPCs is 13.5 scfh (EPA, 2021a). Some research suggests that a significant fraction of IPCs, called here low-emitting IPCs (LE-IPCs), are infrequently actuating and release relatively small volumes of natural gas to the atmosphere (e.g., <2 scfh), if they are well-maintained (Stovern et al., 2020; Thoma et al., 2017; Tupper, 2019). Other IPCs, such as those used in liquid/gas separation service for gas processing units (GPUs), have larger depressurization volumes and actuate more frequently, or may be in throttling service; their emissions are likely more in line with the current IPC EF under proper operating conditions (Thoma et al., 2017; Allen et al., 2015).

Ideally, IPCs minimize natural gas emissions by limiting pilot supply gas venting between actuation events to a low seepage rate (e.g., <1 scfh) by design, also called the closed bleed rate. IPC closed bleed rates vary by manufacturer, model design, and are affected by the maintenance state of the PC pilot (Simpson, 2014; Stovern et al., 2020; Thoma et al., 2017; Tupper, 2019). For example, dirt, debris, wax, or seat failures can prevent proper seal closure, causing much greater closed bleed rates than design values. This condition represents an IPC maintenance issue, with cases ranging from slightly elevated closed bleed rates to significant, sustained emissions that are multiple times the EF. The latter category represents one type of “high emitter” source observed in numerous studies (Alvarez et al., 2018; Allen et al., 2015; Luck et al., 2019; Tupper, 2019). Elevated emissions from IPCs may also indicate issues with the production process, for example inefficient control loop design, suboptimal operational settings, process and equipment malfunctions, or extreme production rates.

Recent long-term measurement studies indicate the temporal emission profiles of some IPCs can be complex and exhibit a variety of transient effects (Luck et al., 2019). Whereas longer-term installed measurements provide understanding of IPC emissions over time, less invasive emission survey approaches can assist in identification of IPC maintenance issues and suboptimal processes that require attention (Stovern et al., 2020; Thoma et al., 2017; Tupper, 2019; Methane Guiding Principles Partnership, 2019). Leak detection and repair (LDAR) surveys using optical gas imaging (OGI) are routinely performed by operators to detect CH₄ and VOC emissions and affect equipment repairs as required by applicable regulations. Traditionally, LDAR surveys focus on fugitive emissions, which are leaks from component fittings, flanges, valves, or other equipment that is not designed to intentionally emit gas. ONG operators and regulators are now implementing or exploring the use of OGI to identify potentially excessive emissions associated with IPC maintenance as part of routine LDAR surveys (Stovern et al., 2020; Thoma et al., 2017; EPA, 2021c; Colorado, 2022; New Mexico, 2022).

In April 2018, OGI surveys and short-duration IPC emission measurements were conducted at 15 ONG production sites managed by one operator in Ohio County, West Virginia. This work was part of a larger study that investigated other sources of emissions at these sites using onsite and offsite measurements (Johnson et al., 2022). These sites serviced a total of 66 operating wells in total that produced on average 9.3 million standard cubic feet (MSCF, or 263,000 m³) of gas, 325 barrels (bbl) of condensate and 291 bbl of water per day over the month of April 2018, with site selection detailed in Supplemental Information (SI) Section 1. Following a survey of PC systems found on site, OGI inspections identified IPCs and associated actuators with continuous emissions potentially indicative of maintenance or process issues. The CH₄ emission rates from the subset of OGI-detected IPCs were then quantified using a full flow sampling (FFS) device developed by West Virginia University (Johnson et al., 2015; Johnson et al., 2016; Johnson and Heltzel, 2016). This paper describes the population of encountered PCs, results of the OGI survey and FFS measurements, and recommendations for best practices to improve observations of suboptimal IPC function as part of routine LDAR surveys.

2. METHODS

2.1. PC Type and OGI Survey

The type and intended operation of each PC associated with working wells at each site were documented and operationally classified through discussions with operator personnel (SI Section 2). This approach differed from a previous Appalachian basin study that classified PC types based on the characteristics of the time-resolved measurement of PC supply gas (Allen et al., 2015). A trained and experienced study team executed the OGI survey of PCs and associated actuators using a FLIR GF320 (FLIR systems, Billerica, MA) or Opgal EyeCGas^® (Opgal Optronic Industries ltd, Tirat Karmel, Hefa, Israel) OGI camera. OGI cameras developed for leak detection of invisible gases operate according to principles of optical remote sensing; optical elements (such as filters and lenses) narrow the bandwidth of incoming electromagnetic radiation and focus it onto a detector array to make a visual infrared signal. At the time of this study, handheld OGI cameras used for detecting hydrocarbon emissions from ONG operations use a cooled detector for increased sensitivity and methodical survey practices common with LDAR surveys. Although emerging versions of the OGI technology are able to estimate gas emission rates under certain conditions (Zeng et al., 2017), this technology was not available for use in this study. Additional detail on OGI technology and applications is available elsewhere (EPA, 2021c).

Upon site arrival, the study team conducted a whole-site OGI scan to identify any potential emission-related safety issues (e.g., large leaks or process malfunctions). Once the site was cleared for study, the study team OGI operator identified and recorded component-level observations of emissions from typical LDAR components as well as PCs and tanks for subsequent measurement by FFS. The PC portion of the OGI survey focused on identifying and assessing IPCs with potential maintenance or process issues, as evidenced by continuous OGI-observable emissions. Since IPCs are designed to close between actuations, no discernable emissions between actuations should occur. If continuous OGI emissions from an IPC were observed, the FFS was used to measure emissions. If no OGI-observable emissions from the IPC pilot vent, body, or associated dump valve actuator were observed, the IPC system was determined to have no gross maintenance or process issues and was not measured by FFS. FFS measurements were also made in some cases where OGI observations could not be performed due to viewing constraints.

OGI detection limit performance for identification of unknown leak locations during LDAR site surveys varies based on site and weather conditions, observation protocol, and operator training (Zimmerle et al., 2020). This study used an experienced OGI operator and a special observation protocol focused on the known locations of the PC systems. In any survey scenario, the OGI emission detection limit is not a single value but represents a performance band. The detection threshold band maximum (DTBM) represents the leak emission rate at which confident (~90%) OGI detection can be obtained using the specific equipment, observational protocol, training, and conditions of the survey (Ravikumar et al., 2018; Stovern et al., 2020; Zimmerle et al., 2020). Under ideal conditions, OGI can detect emissions well below the DTBM, but with decreasing probability as emission rate decreases. This study used a custom PC observation protocol to produce the best possible emission detection performance. As in studies conducted by Stovern et al. (2020) and Thoma et al. (2017), the OGI team member carefully examined and documented each IPC, actuator, and exhaust vent using extended observation dwell times (typically >1 minute), optimized delta temperature (ΔT) background conditions, and short observation distances of about 2 m for IPC pilots and dump valve actuators and about 5 m for extended IPC exhaust vent pipes (with partial sky background). For this study, the OGI DTBM was assumed to be 2.0 scfh (0.057 m³ h⁻¹) whole gas, in line with other similarly executed component-focused OGI detection studies (Stovern et al., 2020; Thoma et al., 2017; Ravikumar et al., 2018; Zeng and Morris, 2019. IPCs, dump valve actuators, and vents with continuous emissions over the DTBM were assumed to be detectable greater than 90% of the time using this OGI protocol.

As part of the study, the site operator’s LDAR inspector separately performed OGI inspection of the sites using a FLIR GFx320 with company-standard survey procedures that focused on traditional LDAR components rather than PC emissions. The operator’s LDAR inspector preceded the study team and applied LDAR flags to found leaking components. As part of normal LDAR survey procedures for these sites, the LDAR inspector manually actuated many (potentially all) of the indoor GPU liquid level IPC pilots to clear and reset the pilot. For this reason, the “as-found” state of the GPU liquid level IPCs could not be determined in this study. It is generally assumed that an IPC reset reduces continuous emissions by clearing accumulated seal debris and reducing closed bleed rate emissions from the as-found state, but this has not been systematically studied.

2.2. FFS Emission Measurements

Time resolved CH₄ mass emission rates of IPC systems (SI Sections 3-5) were measured by the FFS. With the inlet of the FFS system positioned over an emission source, an explosion-proof blower creates higher sample flow rates through the sampling hose, past sensing components, to exhaust out the system outlet (Figure 1). The sampling hose captures the entire CH₄ plume plus dilution air. The air passes through a fixed mixing length of 10 diameters after the blower exit. The total diluted sample then passes through a calibrated mass air flow meter to quantify the total flow rate. A sampling port delivers a portion of the total diluted sample at a constant volume to a laser-based CH₄ off-axis integrated-cavity output spectrometer (the Los Gatos Research Inc. Model GGA-24P, San Jose, CA) that measures concentration. Using the total flow rates and CH₄ concentrations, the FFS provides real-time continuous CH₄ mass and volume emissions measurements. The system records a separate time-stamped data file for each measured component at a time resolution of 1 Hz (1 second). Data analysis, quality assurance testing, uncertainty, and background effects are discussed in SI Sections 3 and 4 and elsewhere (Johnson et al., 2015; Johnson et al., 2016; Johnson and Heltzel, 2016; Johnson et al., 2022). Evacuated canisters were collected from the FFS exhaust outlet during emission measurements to provide supporting speciation data. The canisters were analyzed with Method TO-14A and EPA Method 18 ALT-100.

Figure 1.

FFS operation schematic where the sample is drawn into the inlet on the left, passes through the system, and exits at the outlet to the right.

Measurements collected using the FFS occurred over a typical time duration of 5 minutes. The FFS measurements included the exhaust vents of the PC and dump valve actuators as well as discrete emission measurements from OGI-detected emission points. For infrequently actuating IPCs, exhaust vent FFS measurements established the continuous emission rate of the supply gas, or closed bleed rate, that may indicate a maintenance or process issue. For more frequently actuating IPCs (or combined groups of these), one or more actuation events could be observed, allowing the study team to verify IPC seal closure in real time. The measurement profiles generated for each IPC were slightly delayed and diffused compared to actual IPC operation due to air entrained by FFS emission capture and IPC exhaust gas flow impedance from extensive exhaust vent pipe lengths. Longer duration FFS measurements (up to 20 minutes) included zero checks of the sampler where the FFS inlet was removed from the emission source to sample ambient air and then returned to the emission measurement position to check measurement reproducibility. Observations were also prolonged in cases where temporally complex emission behavior was noted (SI Section 5).

3. RESULTS

3.1. IPC Classification

A total of 391 PC systems associated with 66 operating wells were found on the 15 sites surveyed. Each PC system included the pilot and the dump valve actuator, either in an integrated housing or spatially separated form connected by supply gas tubing. Through discussions with the operator, it was determined that the intended operating mode for all PCs on the sites was snap-acting (on/off). Two primary categories of IPCs were formed, (1) assumed LE-IPCs, and (2) harder working GPU liquid level IPCs, which carry the default 13.5 scfh EF (Table 1). One distinction between these two groups is that, due to infrequent venting, it is easier to assess the LE-IPC group for gross malfunction using OGI since no continuous emission flow should be observed between actuations (indicating a properly seated IPC pilot). The average number of IPCs per well was 5.9, (4.9 excluding emergency shutdown devices).

Table 1.

Summary of PC survey data.

IPC Description	Study Group	Number Found (N)	Percent of IPC Population (%)	Estimated Actuation Frequency (Time)
Emergency Shutdown	LE-IPC	66	16.9	Monthly to Yearly
Temperature	LE-IPC	132	33.8	Hours to Days
Pressure	LE-IPC	53	13.6	Hours to Days
Level (other)	LE-IPC	8	2.0	Minutes to Days
GPU Liquid Level	IPC	132	33.8	Minutes to Hours

The 259 LE-IPCs included emergency shutdown, temperature control, pressure control, and auxiliary level IPCs. These systems were considered LE-IPCs primarily due to infrequent actuations but this assignment relied on the IPC to be well maintained (e.g., not leaking) and the underlying process to be designed and operating properly. For example, emergency shutdown devices could have extremely low integrated annual emissions and may in fact not be considered as IPCs by some operators. A back pressure regulator may vent significantly under certain process conditions, so assignment as a LE-IPC will depend on site design and operation. Temperature controllers and associated actuators are typically low volume, low frequency PC systems which should exhibit emission rates below 1.0 scfh if not leaking (Thoma et al., 2017). Depending on process design, non-GPU level IPCs may not fit in the LE-IPC category but have been categorized as such for these sites based on findings and for simplicity in discussion.

As the higher emitting PC system category for these sites, the 132 GPU liquid level IPC pilots and associated dump valve actuators managed the transfer of the produced water and condensate from the GPU separation vessels to onsite storage tanks (two PC systems required per well). The GPU liquid level pilots were typically variations of the 1001 model series (Norriseal-Wellmark, Houston, TX), matched to SMA/SMT PO series dump valve actuators (Kimray Inc. Oklahoma City, OK). The liquid level IPC pilots were located in enclosure boxes inside of the GPU shelter while the dump valve actuators were located outside of the shelter and had separate exhaust lines with check valves (detail in SI Section 2). In this configuration, the depressurization of the dump valve actuator bonnets, which inflate with supply gas when actuation is triggered, did not vent through the IPC pilots (as typically occurs), so were not part of the IPC exhaust vent emissions. As explained by the operator, separating the exhaust streams between the IPC pilot and dump valve actuators both enhanced safety inside of the GPU shelter and reduced pilot seal closure impedance to minimize IPC pilot emission transients and lower closed bleed rates. The GPU shelters on these sites typically combined multiple IPC exhaust streams to a common roof vent and were fitted with lower explosive limit monitors to protect against potential safety issues caused by indoor leaks. The comingled IPC pilot exhaust vents coupled with atypical independent dump valve actuator vents precluded summed assessment of the PC system emissions under this study design since the PC system elements could not be matched or simultaneously measured.

3.2. OGI IPC Survey Results

Both the LE-IPC and GPU liquid level IPC groups were surveyed using the PC-specific OGI protocol, with the former group more effectively observed. For LE-IPCs, if the OGI observation failed to detect a continuous emission in its as-found state, we assumed that IPC was well maintained and that the underlying process was operating nominally (no evidence of gross malfunction). Since the OGI DTBM was assumed to be 2.0 scfh whole gas for this study, an LE-IPC closed vent rate at or below this level is inferred. Of the 259 LE-IPCs surveyed only two IPCs exhibited OGI-detectable continuous emissions, a backpressure controller with FFS-measured emissions at 0.2 scfh CH₄ and a stabilizer liquid level controller measured at 2.9 scfh CH₄. The backpressure controller measurement was performed sometime after the OGI observation, therefore the FFS measurement may not be representative of the OGI-observed venting state. No other LE-IPCs were measured by FFS in this study.

The independent and easily observable exhaust ports of LE-IPCs, coupled with low actuation frequency makes detection of potential maintenance issues by OGI relatively straight forward. The observation of LE-IPCs during OGI surveys is like that of standard LDAR fugitive component inspection and can be executed with little additional burden. On these sites it was believed that the operator’s LDAR inspector was including LE-IPCs as part of routine OGI inspection, given the low OGI detection rate found during this study.

In contrast to the LE-IPC group, OGI inspection of the GPU liquid level IPC pilots and dump valve actuators on these sites provided less definitive information. The higher actuation frequency and larger depressurization volumes of the GPU liquid level IPCs, coupled with the site engineering configurations used made OGI detection of potential maintenance issues more complicated than the LE-IPC case. Because the IPC pilots inside of the GPU shelters were typically combined (ganged) together into a common roof exhaust vent, identification of individual malfunctioning IPC pilots was difficult to determine. In most cases, the IPC liquid level exhaust vent pipes were routed over the top edge of the GPU shelter roof making direct OGI observation of the vent pipe difficult from ground positions. Further compromising the assessment was the standard procedure used by the operator’s LDAR inspector that preceded our OGI assessment whereby most (possibly all) GPU liquid level IPC pilots were manually actuated to clear and reset the pilot as part of the operator’s standard LDAR survey. As discussed, this action invalidated the assumption of “as-found condition” for GPU IPC pilots.

Even with these challenges, some useful OGI survey information was obtained from this group. Outside of the GPU shelter, OGI inspection for leaks in dump valve actuator bodies or supply gas connections mimicked fugitive-type inspection and was performed with clear interpretation. Similar to the LE-IPC case only three examples of fugitives out of the 132 dump valve actuators were found indicating generally good routine LDAR survey performance. The elevated vents of the dump value actuators were observed and continuous emissions indicative of a potential diaphragm leak was found in one case. The FFS team measured these OGI-detected issues at the emission source and performed additional vent assessments producing 25 dump valve actuator measurements in total (3 leaks and 22 vents).

OGI observations of GPU liquid level IPC pilots were performed both inside the GPU shelter and at the roof vent pipe exit. In total there were 60 GPU roof vents pipes supporting the 132 GPU liquid level IPC pilots. Five GPU roof vent pipes representing 10 total IPCs were confirmed by OGI to have no visible continuous emissions between actuations and therefore possessed acceptable closed bleed rates. These five roof vents were not measured by FFS. The other 55 roof vent pipes representing 122 IPCs were either unable to be assessed by OGI or had clear OGI observable emissions, which included an unknown fraction of actuation venting (expected). In all 55 cases these emissions were measured by FFS.

In three cases, OGI surveys discovered emissions from a closed IPC pilot box (Figure S2) inside the GPU, likely indicative of leaking IPC pilots. In these cases, FFS measurements were directed to these emission points. It was realized that the closed IPC pilot box routed the majority of these emissions to the roof vent as part of safety measures. Early in the study prior to systematic FFS measurement of the IPC roof vents, the IPC pilot box was opened and directly measured by FFS in three cases. As these specific IPCs were independently vented, these three measurements were believed to be equivalent to a roof vent measurement of the same system, so are counted as part of the 55 GPU roof vents monitored for discussion simplicity.

3.3. FFS Measurements of GPU Liquid Level IPCs

The FFS team measured time-resolved CH₄ emissions from 84 separate emission points associated with liquid level IPC pilots and dump valve actuators (Figure 2). The white (and blue) filled data points represent the CH₄ mean emission rates (MER) in scfh, with error bars representing ± one standard deviation (1 σ). The black filled data points represent the normalized MER where the measured FFS data has been divided by the number of ganged IPC pilots or dump valve actuators (2 or 4) that could contribute to the emission. The dotted and dashed lines show the IPC whole gas EF of 13.5 scfh and the assumed OGI DTBM of 2.0 scfh, respectively, for reference. FFS timeseries data are detailed in SI Section 5 as report numbers (RNs) listed in order of Figure 2 abscissa values. All measurements correspond to the 132 GPU-related PC systems except RN33, which was a stabilizer liquid level IPC measurement discussed in Section 3.2.

Figure 2.

FFS-determined CH₄ MER from liquid level IPC pilots (white/blue circles) and dump valve actuators (white triangles), seven values below 0.2 scfh CH₄ are not shown. Closed black circles represent the normalized MER for CH₄. Emission categories A, B, and C presented in Figure 3 panels A1, A2, B1, B2, and C1, are shown, with C2* the combined result of a single pilot into a split header (RN82A, RN82B). The EF and OGI reference lines are scfh whole gas.

Of the twenty-five FFS measurements of dump valve actuators (RN1-RN25), 22 were executed at the actuator exhaust vent and three were low level fugitive leaks found by OGI at the valve stem or body (RN1, RN4, and RN5). Dump valve actuator vent data includes expected depressurization behavior (e.g., RN8 and RN12) as well constant emissions up to 6.3 scfh (RN 25), indicative of a diaphragm leak. Fifty-two (52) FFS measurements were made on GPU liquid level IPC exhaust vent pipes that connected directly to one or more closed IPC pilot boxes. In three instances (RN37, RN45, and RN66), open pilot box measurements were made in lieu of vent pipe measurements and were assumed to be equivalent because of their direct connection (yielding 55 total IPC vent measurements). In two cases, a closed pilot box was observed by OGI to be additionally leaking. One pilot box had two leaking points producing RN27 and RN31 (3.6 scfh CH₄ total) and is related to exhaust vent measurement RN43. Another pilot box leak, RN46 at 10 scfh, was related to vent measurement RN79 at 69 scfh CH₄, clearly representing a maintenance or process issue. The observation of continuous emissions from a closed IPC pilot box or vent could be due to an excessive closed bleed rate (IPC pilot maintenance issue) and/or a connector-related fugitive leak inside the IPC enclosure. In either case, most of the emissions were directed to the roof vent, complicating OGI maintenance issue detection by OGI.

Approximately 25% of the FFS CH₄ MER measurements were below the assumed OGI DTBM of 2.0 scfh. There are several reasons for this apparent discrepancy. OGI detects CH₄ and other compounds that form the whole gas emission whereas FFS measures only CH₄. In some cases, the GPU IPC pilot and dump valve actuator vents were measured by the FFS team without OGI detection. In other cases, the emissions were temporally variable or the OGI observation conditions were optimal resulting in lower detection limits. Due to the physically separated nature of the IPC pilot and dump valve actuator, the exhaust vents were measured by the FFS at different times. This, coupled with the ganged nature of the vents, made it impossible to meaningfully combine FFS readings into a single result for each of the 132 GPU liquid level IPC systems on a component-by-component basis. Of equal importance, the snapshot-in-time nature of the FFS assessment may not sufficiently represent longer term PC activity factors (i.e., actuation frequency). Even with these limitations, information on the performance of these PC systems can be found in analysis of the FFS time series measurements.

3.4. Three IPC Pilot Emission Categories

From the FFS timeseries data and the histograms of SI Section 5, liquid level IPC pilot emissions were classified into one of three categories, called A, B, and C. Measurements ranged from the expected behavior of a snap-acting IPC pilot [Figure 3(A1), (A2)] to continuous emissions indicative of an IPC maintenance or process issue [Figure 3(C1), (C2)]. Category A cases were characterized by a cessation of emissions between actuations (or no actuations) to a low closed bleed rate with an emission rate histogram typically dominated by the lowest bin. For linkage to OGI inspection protocols, we propose the measured closed bleed threshold should be below 2.0 scfh for a single IPC pilot. Typically, the CH₄ MER for this class was at or below the IPC EF of 13.5 scfh. For multiple IPCs ganged into a common exhaust vent, the acceptable category A closed bleed rate was multiplied by component count (e.g., < 8.0 scfh CH₄, for four ganged IPCs). Figure 3(A2) provides an example of four ganged IPC pilots. The high actuation duty cycle of the group may be due to multiple IPCs actuating or one pilot working at an unusually high frequency. The apparent baseline at ~ 4.5 scfh may be caused by one IPC pilot with an elevated closed bleed rate or the additive effect of multiple pilots. Due to the entrained air flow of the FFS, the actuation peaks are broadened and possibly overlap neighboring actuation peaks, so the actual baseline of Figure 3(A2) may be lower. In terms of OGI inspection, low closed bleed rates of IPCs could be confirmed if the observed emission periodically falls below the OGI detection threshold between actuations. This assessment is complicated by ganged pilot exhausts, suboptimal observation geometries (folded exhaust ports over GPU roofs), and the long sections of exhaust vent pipes that temporally attenuate the emission event under low exhaust pressure.

Figure 3.

FFS CH₄ timeseries for three categories GPU IPC pilot emissions: (A1), (A2) category A, (B1), (B2) category B, and (C1), (C2) category C, with (—) indicating valid FFS data, (—) indicating quality assurance-removed FFS data from zero check or other, and (—) showing CH₄ for valid FFS data; histograms adjacent right. Shaded box inserts contain SI RNs, number of IPCs potentially contributing to the measurement, and CH₄ MER.

Category B measurements show complex temporal behavior where the IPC pilots(s) may demonstrate the ability to fully seat to a low closed bleed rate between actuations, but the production process, pilot setting and/or maintenance state were not optimized and resulted in higher than expected emissions (Allen et al., 2015; D’Antoni, 2018; Luck et al., 2019; Methane Guiding Principles Partnership, 2019). When observing ganged IPC exhaust vents, complex temporal patterns may result from the combined action of multiple IPCs. Figure 3(B1) represents two ganged IPC pilots. Initial actuations indicate that the pair can achieve a low closed bleed between depressurization events, however, a complex emissions pattern emerges that could be caused by a suboptimal interaction between the IPC setup and process settings from one or both IPCs. Figure 3(B2) provides another example of a category B measurement from a dual IPC exhaust while RN58 shows an example of a single controller with the ability to approach zero emissions. These examples exhibit a highly modulated temporal profile indicating conditions analogous to a throttling mode, however the pilot was in snap acting (on/off) operation. Category B timeseries measurements possess broad FFS emission histogram distributions. Whereas category A classifications typically show normalized MER below the IPC EF, category B measurements may exceed the EF (Figures 2 and 4).

Figure 4.

(a) Estimate of whole gas emissions from the 132 GPU liquid level IPC systems by category (A, B, or C) found using the equal apportionment (low limit) case assumptions with inset showing emission breakdown by category. (b) same using the high limit case assumptions.

Further departing from ideal snap-acting IPC operation, category C temporal profiles were dominated by elevated continuous emissions indicating significant IPC maintenance and/or underlying process issues. This category exhibited elevated normalized MER typically exceeding the IPC EF, frequently with narrow emission histograms that were well-separated from the lower bins. These continuous emissions never returned to the baseline level indicative of an IPC closed bleed rate. In ganged category C cases, category A or B emissions may also be present but the category C behavior of one or more IPCs dominated the profile.

3.5. Canister Comparisons and Translation to Whole Gas Emission Rates

Eighteen (18) evacuated canister grab samples associated with GPU IPC operation were collected in the FFS exhaust stream for speciated laboratory analysis by EPA Method TO-14A and EPA Method 18 ALT-100 (SI Section 4). The average relative percent difference for CH₄ concentration determined by the FFS and canisters was 0.4% with a largest difference of −38.3% (Figure S4). Estimated whole gas emission rates (CH₄ + VOC), were determined from the canister analysis. The whole gas emission rates were found to be 37% higher by volume than the CH₄-only measurements for these sites, which was in good agreement with the reported value for produced gas in Ohio County, WV. The canister analysis was consistent across all canister samples collected from IPC sources (Figure S5). The same canister analysis was conducted on samples collected from tank battery sources and showed to be quite variable (see companion paper by Johnson et al., 2022).

4. DISCUSSION

Estimates of whole gas emissions from the 132 GPU liquid level PC systems (Table 2, Figure 4) were derived by combining the dump valve actuator and IPC pilot FFS data. Due to the independent dump valve actuator and IPC exhausts, ganged exhaust vents, and the short time duration and single point nature of the FFS measurement, the actuator-related FFS data (Figure 2, RN1-RN25) could not be directly combined with the IPC pilot FFS data on a component-by-component basis. To explore the uncertainty in assumptions two limit cases were developed, called the equally apportioned (low limit) case and the high limit case. As a first step for both cases, the three actuator-related leak measurements were combined with the appropriate actuator vent measurement (22 total) to yield an average whole gas emission rate of 1.99 scfh. This mean emission rate was then applied to the entire population of 132 dump valve actuators as a default. The veracity of this estimate is in part confirmed by the low number of OGI continuous emission detections for this group.

Table 2.

Summary of GPU liquid level IPC system whole gas emissions.

		Equally Apportioned Case (Low Limit) For Ganged GPU IPC Vents			High Limit Case For Ganged GPU IPC Vents
	Overall Whole Gas	Category A Whole Gas	Category B Whole Gas	Category C Whole Gas	Category A Whole Gas	Category B Whole Gas	Category C Whole Gas
Mean (scfh)	19.1	4.19	15.8	36.5	4.27	27.3	79.4
Median (scfh)	12.8	4.35	15.6	29.2	4.4	27.7	65.0
σ (scfh)	22.3	1.28	8.9	30.9	0.84	13.9	52.4
Minimum (scfh)	2.05	2.05	3.91	8.41	2.05	3.91	20.6
Maximum (scfh)	169	6.88	44.9	168.9	6.88	51.9	176
Total (scfh)	2520	159	822	1530	375	710	1430
Count (N)	132	38	52	42	88	26	18
Percent of Total (%)	100.0	6.3	32.7	61.0	14.9	28.2	56.8

Similarly, the three IPC closed pilot box leak measurements were combined with the appropriate IPC vent measurements. The 55 total IPC FFS vent-type measurements represented 122 of the 132 IPCs. These measurements were classified as category A, B, or C. Eleven (11) of the 55 measurements were classified as category A, representing 28 IPCs. The average whole gas emission rate for category A was 2.36 scfh. There were five unmeasured GPU IPC vents representing 10 IPCs that exhibited no continuous emissions by OGI inspection and were therefore classified as category A cases and assigned the mean whole gas emission rate value of 2.36 scfh. In total, 38 liquid level IPCs, or 28.8% of the population of 132 GPU IPCs, were ascribed to category A. Of the 55 IPC vent FFS measurements, 26 were category B and represented 52 (39.4%) of the 132 IPCs. The remaining 18 FFS IPC vent measurements were category C determinations that represented 42 (31.8%) of the population.

The first estimate of emissions combined the dump valve actuator whole gas value of 1.99 scfh with each of the FFS vent measurements (or category A mean values). The ganged vent FFS measurements plus the appropriate number of actuator emission units were equally apportioned among the number of IPCs that contributed to the measurement (two or four). As an example, RN69 [Figure 3(B1)] had a CH₄ MER of 36.1 scfh, which equated to a whole gas estimate 49.5 scfh, with two IPC pilots contributing to the FFS measurement. This became two separate entries in the final tally, each with a whole gas emission rate of 1.99 scfh plus 24.75 scfh, or 26.7 scfh total. This procedure was repeated to form a set of 132 IPC systems yielding the overall and an equally apportioned (low limit) sections of Table 2, with the latter visually represented by Figure 4(a).

For the equally apportioned case, the FFS timeseries and OGI survey results provided a minimum number for the category A IPCs and the simplest assumption for distribution of the ganged category B and C cases was employed. However, it was likely that a significant number of the IPCs represented by category B and C FFS measurements exhibited category A behavior that was masked by a subset malfunctioning IPCs in the ganged measurement. A high limit case was derived where the fewest number of category B and C cases were assumed [Table 2, Figure 4(b)]. For this case, each of the category B and C FFS vent measurements (26 and 18, respectively) was assumed to be caused by a single nonoptimal IPC. This approach therefore assumed a total of 88 category A cases, which were formed by either the whole gas FFS measurement or the group average value of 2.36 scfh for the non-measured cases, plus the dump valve actuator value of 1.99 scfh. The same procedure was used for the category B and C values with the singles vent case. For the ganged vents, one or three units of the default category A value was subtracted from the FFS measured value for the two and four ganged cases, respectively [Table 2, Figure 4(b)].

The overall emission distribution was heavy tailed, as is commonly observed in oil and gas studies, with a median value of 12.8 scfh (Table 1) that was similar to the IPC EF of 13.5 scfh. Total emissions are dominated by high-emitter IPCs. The top 10% of IPC systems (N=13) were all classified as category C and were responsible for 34.7% and 49.2% of total emissions under the equally apportioned and high limit cases, respectively. The top 20% of IPC systems (N=26) were all classified as category C in the equally apportioned case and represent a mixture of categories B (31%) and C (69%) in the high limit case. The top 20% of GPU IPC systems account for 51.3% and 70.7% of GPU liquid level IPC emissions under the equally apportioned (low limit) and high limit cases, respectively.

With the uniform addition of the 1.99 scfh whole gas dump valve actuator value, 32 of the 55 FFS measurements representing 65 GPU IPC systems exhibited normalized whole gas emission rates in excess of the 13.5 scfh EF, an equal number to that represented in the equally apportioned case. This group represents 83.8% of emissions by volume. For the high limit case, 36 IPCs, all category B and Cs, exceeded 13.5 scfh EF and represented 82.7% of emissions. The highest GPU IPC pilot vent emission came from a single malfunctioning unit with a continuous normalized CH₄ MER of 122 scfh (RN 82), or whole gas value of 157 scfh. This is also the highest value in the equally apportioned case since it involved a single vent measurement. The highest emitter in the high limit case was 176 scfh with six IPCs exceeding 100 scfh. These high emitter vents were easily observable by OGI. The dominant factor for high emitting IPC vents was continuous category C type behavior. Atypical temporal profiles have been linked to higher emissions; one study indicated PCs with abnormal actuation patterns result in mean emissions five times higher than PCs with normal actuation patterns (Luck et al., 2019).

Investigating site-specific factors, Figure 5 combines the total GPU IPC emissions by site (1-15), with April 2018 site production data found in a companion paper (Johnson et al., 2022). Figure 5 includes only the FFS-measured values of the GPU liquid level IPCs and dump valve actuators (Figure 2) by site with the CH₄ readings converted to whole gas emission rates. Figure 5 plots the sum of measured emissions on each site, normalized by the expected emissions (%), against the approximate combined condensate and water daily liquids production levels (bbl/IPC). The expected emissions were determined by multiplying the number of working liquid level IPCs by the 13.5 scfh EF. Occurrences of FFS measurements above the EF (Figure 2), with emission categories (Figure 3) and their SI RN numbers are indicated in Figure 5. One third of the sites exhibited emissions greater than 200% of expected values. The highest emitting sites were dominated by Category C emission profiles, indicating possible maintenance or process issues. Three sites (8, 9, and 10) possessed higher than expected throughput per IPC in service. The high production rates per IPC observed at these sites may be the cause of IPC maintenance issues, and/or the category B and C temporal profiles found. Site 8 exhibited the highest IPC emissions and was experiencing a malfunction associated with liquids handling and tank storage (Johnson et al., 2022). Site 4 possessed reasonable liquids production per IPC in service but had a collection of IPC maintenance issues that could, in principle, be diagnosed by OGI survey.

Figure 5.

Comparison of approximate daily liquids production per liquid level IPC installed versus liquid level IPC site whole gas emission ratio. The label indicates the production site number with the RNs for all high emitter IPC devices shown. The dashed line indicates 200% of expected emissions level (whole gas basis).

5. SUMMARY and CONCLUSIONS

An onsite study of 15 West Virginia ONG production sites servicing 66 working wells was conducted over a two-week period in April of 2018. One part of the study aimed to identify and characterize PC systems with excessive emissions caused by maintenance issues or nonoptimized process conditions, using OGI survey and short duration FFS CH₄ emission rate measurements. The initial site survey included 391 PC systems, all of which were classified as snap acting (on/off) IPCs through consultation with the site operator. The average number of IPCs per well was 5.9, (4.9 excluding emergency shutdown controllers).

The population of IPCs was divided into 259 infrequently actuating LE-IPCs and 132 GPU liquid level IPC systems. For the LE-IPC group, it was assumed that properly maintained IPCs servicing normally operating processes should exhibit very low continuous emission rates between actuations (low closed bleed rate), that should be below OGI detection limits. A PC-specific OGI inspection protocol with an assumed 2.0 scfh DTBM found continuous OGI emissions in 2 of 259 LE-IPC observations. This low occurrence of leaking IPC systems was believed to be in part due to the relative ease of inspection of this category as part of regular OGI LDAR leak surveys. OGI inspection of the GPU liquid level IPC systems was comparatively less informative. The expectation of vented emissions from these frequently actuating and larger volume liquid level IPC systems, coupled with the ganged nature and physical configuration of the GPU pilot exhaust vents on these sites made determination of potential operational issues by OGI difficult in comparison to the LE-IPC case. Several site engineering best practices, such as independent and easily visible exhaust vents could facilitate OGI diagnostics of GPU IPCs as part of regular LDAR surveys. The use of ganged header vents also made determination of single component emissions and two limit cases were presented.

Time-resolved FFS measurements of CH₄ emissions from GPU IPC liquid level pilot exhaust vents found three classes of temporal profiles that could be used to infer operational state. Category A profiles indicated normal IPC operation, as evidenced by low closed bleed rates between actuations (< 2.0 scfh/IPC CH₄ by FFS). Category B and C profiles exhibited complex temporal behavior or more continuous high bleed rates, respectively, with both cases indicative of an IPC maintenance issue or non-optimized process condition exhibiting higher emissions. The overall GPU IPC emission distribution was heavy tailed, with a median value of 12.8 scfh whole gas that was similar to the 13.5 scfh IPC EF. Depending on the limit case assumptions, between 27% and 50% GPU IPC systems exceeded the EF and represented approximately 83% of emissions. Total emissions were dominated by category B and C high-emitter IPCs with the top 20% of systems (N=26) accounting for between 51.3% and 70.7% of GPU liquid level IPC emissions. The highest GPU IPC pilot vent emission came from a single malfunctioning unit with a measured whole gas value of 157 scfh. Up to six IPCs exceed 100 scfh. An analysis of FFS emission measurements compared to liquids production per IPC unit employed indicated that production sites operating at high liquids production test the limits of the site engineering, likely resulting in higher IPC emissions.

Considering the OGI LE-IPC survey results, the distribution of FFS-measured liquid level emissions around the EF, and the observations of abnormal temporal profiles and higher emission rates likely indicative of maintenance or process issues (and/or abnormally heavy liquid production throughput per IPC utilized), we conclude that three levels of IPC emissions were observed on these sites:

• LE-IPCs with OGI-verified low closed bleed rates with EFs likely < 2.0 scfh whole gas.

• Active GPU liquid level IPC systems with emissions likely well represented by the current whole gas IPC EF of 13.5 scfh.

• Higher emitting IPCs that are either experiencing a maintenance or process issue or are operating at very high throughput rates and are not well represented by the current IPC EF of 13.5 scfh. These IPCs can emit at over ten times the EF.

With significant variations in site engineering and process design by ONG operators in different basins, coupled with mixes of IPC makes and models, additional research is required to understand the broader applicability of these findings. Production operations using natural gas-driven IPCs could benefit from OGI observations of these systems as part of regular LDAR inspections. FFS-type assessments of harder working IPCs and processes could provide valuable performance benchmarks. Significant maintenance or process issues can be diagnosed and mitigated with these tools. OGI and FFS survey findings can be utilized to refine emission estimates in the context of key metadata such as production throughput per IPC in service.

SYNOPSIS.

• As part of routine OGI LDAR inspections, LE-IPCs can be screened for maintenance issues.

• Short duration FFS time-resolved emission measurements of higher emitting GPU liquid level IPCs can be a useful diagnostic to detect issues with the IPC or underlying process.

• The top 20% of GPU liquid level IPC systems accounted for between 51.3% and 70.7% of emissions in this study.

• Site engineering can be optimized to support OGI and FFS testing leading to emissions reductions.

ACRONYMS AND ABBREVIATIONS

ΔT

temperature differential

bbl

barrels

CH₄

methane

DTBM

detection threshold band maximum

EF

emission factor

EPA

Environmental Protection Agency

FFS

full flow sampler

g/h

grams per hour

GHG

greenhouse gas

GPU

gas processing unit

Hz

hertz

IPC

intermittent venting pneumatic controller

LDAR

leak detection and repair

LE-IPC

low-emitting intermittent venting pneumatic controller

m

meters

m³ h⁻¹

meters cubed per hour

MER

mean emission rate

MI

maintenance issue

MSCF

thousand standard cubic feet

OGI

optical gas imaging

ONG

oil and natural gas

ORD

Office of Research and Development

PC

pneumatic controller

RN

report number

Scfh

standard cubic feet per hour

SI

supplemental information

U.S.

United States

VOC

volatile organic compound