In artificial lift, the new name of the game may be adaptability. From hardware to automation, the goal is to match lift performance to changing well conditions for more productivity from maturing assets.
Operators are turning to different technologies to meet the challenges posed in these aging wells, and they highlighted some of those solutions during SPE’s Artificial Lift Conference and Exhibition in Houston in August.
While mature horizontal wells typically hold remaining recoverable oil and gas, declining reservoir pressure and liquid accumulation make producing those hydrocarbons increasingly difficult, especially from laterals with geometric low points. Current commonly used artificial lift methods haven’t been suitable for addressing waning production from these late-life laterals, Ryan Hieronymus, artificial lift lead at Occidental, said while presenting SPE 232971.
“Thousands of horizontal wells across US onshore are starting to exhibit liquid loading below the depth that we typically reach with conventional artificial lift methods,” he said.
He cited a 2012 Utica study of more than 1,000 wells that indicated more than 80% had at least one low point, or sump area, in the lateral where water can choke production from the wellbore.
“If you have a very tortuous wellbore with low points and choke points and sumps, that can have a big impact on your production, especially your late-life production,” he said.
The operator ran a field trial on a well in Larimer County, Colorado, using a mechanically simple form of intermittent gas lift solution called Gas-Assisted Liquid Lift Using Oscillating Pressure (GALLOP). The system uses two flow strings, a check valve, and intermittent lift-gas injection to produce low to moderate liquid volumes from a selected depth within the lateral, which can be located anywhere along the horizontal section. According to the authors, once installed, the system is similar in complexity to intermittent single-point gas lift.
In June 2022, Oxy installed GALLOP for the first time in a pilot well producing from the Codell formation in the Denver-Julesburg (DJ) Basin that had experienced severe loading below the production tubing for the previous 2 years. It started production in July 2022, and in the 4 years the system has been online, the well has “produced consistently at the capacity of the reservoir, even weathering and recovering from offset frac communication,” the authors wrote.
GALLOP’s lift depth can be placed in a sump much deeper than those used with traditional artificial lift methods (Fig. 1).
Hieronymus said GALLOP has a fill phase and an injection phase. In the fill phase, liquid enters the tubing system through the downhole check valve, and in the injection phase, injection gas is introduced. That check valve closes, and the liquid is lifted to the surface.
“This is where we’re starting to get slugs into the separator. And then, finally, when we remove all the liquid, we go into pressure-down. The injection stops, and we pressure down as quickly as possible to the production separator so that that downhole check valve opens again and we can bring more fluid into the tubing system,” he said.
Testing of the system at a flow loop at Colorado School of Mines indicated 75% of liquid removal was possible, he said, referencing a separate evaluation documented in SPE 205962 and a 2020 doctoral thesis.
Because the system was expected to run for 2 years, components needed to function at least that long. But no commercially available check valves achieved the target 17,000 cycles that represented 2 years of expected GALLOP cycling, Hieronymus said, so a custom check valve was required.
“The advantage of GALLOP is really just its mechanical simplicity,” he said. “There’s only one moving part downhole, and so far, we’ve seen great performance. We’ve seen over 4 years of production with the pilot in the DJ Basin.”
The pilot well, which was near but not connected to an existing gas lift compression system, had experienced chronic liquid loading after being shut in for about a year and a half. The well was effectively nonproducing, and Oxy was considering plugging and abandoning it but instead opted to use it for the GALLOP trial.
“Overall, it’s been incredibly resilient. It initially came on right at about the 20 barrels of liquid per day target, but it quickly declined to between 2 and 5 barrels per day. So, we believe that that is inflow related and not outflow related,” Hieronymus said.
The pilot validated the GALLOP concept and showed it could produce at very low liquid rates, according to the paper.
“The ability to produce a horizontal well down to only a few barrels per day may unlock many years of additional production and reserves, bringing hyperbolic forecast technical limits that were previously difficult to achieve within practical reach,” the authors wrote.
Hieronymus said the best horizontal candidate wells for GALLOP are those that are close to existing gas compression systems.
Because the pilot is still in operation after 4 years, many of GALLOP’s opportunities for future development revolve around operability and quality of measurements rather than equipment robustness, the authors wrote.
Economic Retrofits
In an industry where a majority of wells are considered marginal, operators often cannot justify the cost of retrofitting conventional automation systems. One operator sought to determine if AI-driven pumpoff automation was economically feasible on wells that produce only a few barrels of oil equivalent per day.
Charles-Henri Clerget, chief operating officer at Amplified Industries, said while presenting SPE 233005 that a well producing under 15 BOE/D may take 18 to 24 months to pay out after an operator invested the $10,000–$15,000 cost to cover sensors, controller, and telemetry in typical pumpoff controller retrofits, “which is not going to be very appealing to most operators.”
The budget needed for retrofitting marginal wells in this way wouldn’t really move the needle much in terms of production, he added.
“You end up with a lot of those wells that have no monitoring, no automation. And the problem is, while those wells do not move the needle for you on the production side, they’re still a big source of cost because, at the end of the day, they’ll fail like a big well,” Clerget said.
A lower-cost approach could help operators with a fleet of marginal wells improve their production numbers, he said. What it would take, he said, is a hardware cost that could be justified at output rates as low as 3 BOE/D, with simple installation and the ability to detect failure and send alerts from day one. One such solution using the Internet of Things (IoT) combines a wireless, clamp-on load cell with edge-side pumpoff control and cloud-based AI diagnostics (Fig. 2).
Siena Natural Resources deployed the automated solution across 150 wells for 2 years and found it had a payback period of less than a month and reduced lifting costs by $11.80/BOE, the authors wrote. During the pilot, the mean time between failures rose from 8 months to 2.5 years, production increased by 7%, 22 surface spills were prevented, and the team noted net annual benefits of $12,950/well. Each installation cost about $1,000 and carried an annual operating cost of $1,200.
Clerget said, “The relative benefit in terms of failure reduction, power saving, all these kinds of things, are outsized compared to what I think would be more conventionally expected on wells that are receiving more attention and more care.”
Wil McCabe, vice president of production at Siena Natural Resources, said while presenting the paper that the wells in the pilot had initially received daily visits.
“Essentially, if a problem happened within that 24 hours, you found it the next day. After implementation of the system, we developed an alarm-driven approach that was based on criticality of the events that happened within that 24 hours,” he said. “Most of the events that happened during that time were able to be closed out the same day versus waiting for 24 hours for the next operator to show up. Based on this alone, we were able to add 10% to 20% the number of wells managed per person per lease operator, just on this technology alone.”
He attributed the 7% uplift in production, which equated to about $3,800/well, to runtime optimization, faster response, and less downtime due to more quickly catching the wells being down.
“It changed not only the way that we operated, viewed wells, (but) it changed the way that we approach problems. We stopped visiting normal wells for the sake of being normal and approached problems on a daily basis as they came,” McCabe said.
Scaling Solutions
While engineers review data and troubleshoot underperforming electrical submersible pumps (ESPs), operators face the challenge of scaling those activities across large well populations. A two-phase pilot project validated physics-based models to establish a trustworthy engineering model before layering in artificial intelligence (AI) in a digital workflow that could help identify critical pump conditions and promote optimal production, Omar Iskandar, senior director of digital operations for Leucipa at Baker Hughes, said while presenting SPE 232996.
The integrated digital workflow was used on over 100 ESP wells in the Middle East and led to production uplift from already high-rate onshore wells, more efficient ESP surveillance, and improved early warning for potential ESP failure, according to the paper’s authors.
Iskandar said six production optimization recommendations involving adjusting frequencies within operational constraints delivered an average of 16% uplift for a total of 2,116 BOPD from the six wells (Table 1).
Well | Oil Production Gain | |
BOPD | % | |
Well 1 | 150 | 5 |
Well 2 | 600 | 20 |
Well 3 | 500 | 9 |
Well 4 | 200 | 25 |
Well 5 | 216 | 48 |
Well 6 | 450 | 56 |
Total | 2,116 | ~16 |
Table 1—Results from implementing production optimization recommendations in six of the wells. Source: SPE 232996.
The authors said the workflow recommended other production optimization possibilities that were not executed because of operational limitations and field conditions.
Iskandar said the combined physics-based and AI-informed workflow flagged critical conditions on ESPs in the pilot. The system is able to identify nearly two dozen critical conditions, including down/upthrust, low efficiency, motor loading, motor overheating, gas-lock probability, and minimum production allowed.
“During the period of the evaluation, there were like 15 alerts per day,” he said, noting improvements in the model and AI improved the accuracy of the alerts.
The critical-conditions model reached 72% accuracy, just over the target accuracy level of 70% or greater. The workflow was able to estimate ESP failure within the 45–90 days about 57% of the time.
The results were “very positive,” Iskandar said. Ongoing exposure to new data will enable the AI and machine learning models to continuously refine and improve their performance, he said.
And filling in gaps in data will help with improving the remaining useful life-estimation performance, he added.
“What we are now working on is trying to move this to another phase, where we can keep increasing and enriching that (remaining useful life) model to be able to predict more accurately the remaining useful life,” he said.
For Further Reading
SPE 205962 Mechanistic Model for the Design and Operation of an Intermittent Gas Lift System for Liquid Loaded Horizontal Gas Wells by D. Croce and L. Zerpa, Colorado School of Mines.
SPE 232971 Development and First Field Pilot of Gas-Assisted Liquid Lift Oscillating Pressure (GALLOP): A Novel Artificial Lift Method for Unloading Mature Horizontal Wells From the Lateral by R. Hieronymus, Occidental; S. Wilson, Ryder Scott Company, LP; and D. Green, Well Master Corporation.
SPE 232996 From Data to Action: Field Deployment of Live Physics-Based Models for ESP Optimization and Operational Efficiency Gains by F. Bosia, R. Cafarelli, D. Chessa, C.P. Gomez, C. Guglielmo, A. Kurd, G. Rizza, S. Pilone, and L. Vigano, Eni; M. Aljanabi, Boc; and H. Alrashid, S. Barbarino, M.E. Gonzalez, M. Hamoud, O.A. Iskandar, M. Khan, D. Marquez, Y. Montana, M. Orooq, and A. Valente, Baker Hughes.
SPE 233005 Unlocking Value Through AI-Driven Pump-Off Automation: A Two-Year IoT Deployment Across 150 Marginal Wells by C. Clerget, Amplified Industries Inc.; N. Chugunov, Siena Natural Resources; S. Mannai, Amplified Industries; and W. McCabe, Siena Natural Resources.
