Operators managing large inventories of artificial lift systems and intermittent wells face a persistent challenge: how to detect equipment anomalies, optimize production, and reduce manual intervention across geographically dispersed assets in real time.
Traditional approaches, such as periodic well checks, centralized SCADA polling, and reactive maintenance, leave operators blind to rapidly developing downhole conditions, resulting in avoidable production losses, premature equipment failures, and excessive personnel exposure.
This case study describes how edge computing and industrial internet of things (IIoT) platforms were deployed to automate and optimize production operations across four distinct basins: the Oriente Basin in Ecuador’s Amazon region (IPTC 25145), the Permian Basin in Texas (SPE 216829) the Williston Basin in the Bakken (SPE 222618), and the Haynesville Basin in Louisiana (SPE 229390). Each deployment targeted a different artificial lift method or well type, demonstrating the breadth of the edge‑based approach.
The Challenge
In Ecuador’s Amazon region, a mature brownfield with electric submersible pump (ESP) wells faced compounding constraints. These included no permanently available rig for interventions, a remote jungle location more than 100 km from the nearest city, and a reduced workforce where a single operator was responsible for over 60 wells.
Manual operations exposed personnel to high-pressure, high-temperature, and electrical hazards, while delayed anomaly detection led to frequent ESP failures and deferred production.
In the Permian, an operator managing unconventional horizontal gas-lift wells struggled with gas-lift injection rate optimization. Traditional simulation-based approaches depended on well-calibrated models that could not keep pace with the severe slugging and rapidly changing conditions characteristic of unconventional completions.
In the Bakken, sucker-rod-pump (SRP) wells experienced excessive cycling, with some wells averaging six shutdowns per day, and operators lacked real-time diagnostics to distinguish between gas interference, fluid pound, and tagging events.
In the Haynesville Basin, intermittent gas wells managed with manual or calendar-based shut-in cycling suffered from suboptimal liquid unloading, prolonged downtime, and frequent operator intervention.
The Solution
All four deployments share a common architectural foundation: ruggedized edge computing devices installed at wellsites that ingest high-frequency sensor data and execute analytics locally, enabling closed-loop control with sub-second response times (Fig. 1). Only preprocessed summaries and alerts are transmitted to the cloud, reducing data transmission volumes by 85 to 95% compared with streaming raw sensor data (SPE 202252, SPE 201411).
In the Amazon, the automated well operator (AWO) edge application integrated four workflows:
- Smart production surveillance
- Smart chemical injection
- Smart well test
- Smart surface equipment
These workflows fed into a single digital twin interface enabling remote and autonomous ESP operation, chemical dosing, and automated well testing with artificial intelligence and machine learning (ML)-based stabilization algorithms.
In the Permian, a data-driven gas-lift optimization application ran directly on an IIoT gateway device, iteratively testing injection-rate setpoints and implementing optimized rates via closed-loop actuation without requiring well models or field personnel.
In the Bakken, an edge-based workflow combined ML-dynamometer-card classification with fast-loop mitigation algorithms and a production-optimization algorithm, all functioning autonomously in unison on the edge gateway.
In the Haynesville, an autonomous liquid-unloading application used physics-based critical velocity calculations combined with ML-driven shut-in duration prediction to dynamically control choke actuation without manual intervention.
Field Results
Amazon Basin, Ecuador: Over 17 months of continuous AWO operation, the deployment achieved a 6% production increase equivalent to 22,300 additional cumulative bbl of oil. The ESP failure index decreased from 0.5 to 0.26, avoiding at least one major workover. Well test duration was reduced by 60%, or from 10 hours to 4 hours, with 95% measurement accuracy. Operator efficiency improved by 80%, and 26 tons of CO2 emissions were avoided through reduced field mobilizations.
Permian Basin, Texas: The data-driven gas-lift optimization application was deployed on eight unconventional horizontal gas-lift wells. In single-well optimization mode, the candidate well outperformed manually managed wells by 5%. In multiwell optimization across three-well groups, production improvements ranged from 5 to 25%, with one previously underperforming well achieving a roughly 20% step-change in production within a single optimization cycle. The entire workflow (i.e., data gathering, optimization, and setpoint implementation) was executed fully autonomously.
Bakken, North Dakota: A pilot on eight SRP wells demonstrated that the combined ML classification, fast-loop mitigation, and production optimization workflows could operate in unison with minimal human intervention. Inferred production increased by an average of 15%, runtime improved by 3%, and pump cycling decreased by 29% through maintenance of optimal pump fillage. On one well, daily shutdowns were reduced from an average of six to one through systematic VFD (variable frequency drive) speed optimization.
Haynesville Basin, Louisiana: Deployed across nine intermittent gas wells on eight pads, the autonomous liquid-unloading application delivered cumulative gas production increases of 70 to 139% over optimization periods of 63 to 83 days (Fig. 2 and Table 1). Daily production gains reached up to 350 Mscf/D. Analysis estimated over 80 MMscf of additional gas production per well per year.
Conclusion
The consistent results across four distinct operating environments with varying artificial lift methods (ESP, gas-lift, SRP, intermittent gas), geography, connectivity, and organizational maturity validate the edge IIoT architecture as a broadly applicable platform rather than a single-application solution. In each case, the edge device’s ability to execute closed-loop control locally proved critical: whether responding to dynamometer card anomalies within a minute in the Bakken or maintaining autonomous well cycling for weeks without cloud connectivity in the Haynesville.
These implementations demonstrate that the combination of physics-based models with data-driven analytics at the edge enables autonomous optimization workflows that were previously impossible with cloud-dependent or manual approaches. The modular architecture supports horizontal scaling; the AWO framework in the Amazon is designed for replication across additional pads with minimal hardware expansion; and the Haynesville solution’s containerized deployment requires no SCADA modifications.
For Further Reading
SPE 216829 A Robust Method for Data-Driven Gas-Lift Optimization by A. Gambaretto and K. Rashid, SLB.
IPTC 25145 Automated Well Operator—AWO: The Future of Production Operations by S. Guaigua, H. Quevedo, and L. Bustamante, SLB, et al.
SPE 202252 Edge Computing: A Powerful and Agile Platform for Digital Transformation in Oilfield Management by A. Sharma, P. Samuel, and D. Gupta, SLB, et al.
SPE 201411 Edge Computing: Continuous Surveillance and Management of Production Operations in a Cost-Effective Manner by A. Sharma, P. Samuel, and G.M. Gey, SLB, et al.
SPE 222618 Enhancing Edge-Based SRP Production Optimization Algorithm With Fast-Loop Mitigation by Z. Hyder, M. Yermekova, and C. Kemp, SLB, et al.
SPE 229390 Smart Liquid-Unloading IIoT Application for Gas Wells in the Haynesville Basin by A. Gambaretto, C. Kemp, and R. Marin Nunez, SLB, et al.
Akshay Dhavale, SPE, is a product champion for Agora Edge AI at SLB, based in Houston. He leads the development and global deployment of edge-enabled solutions across well and facility operations including artificial lift, flow assurance, and safety systems for energy assets. Under his leadership, Agora Edge AI has achieved global deployment spanning Southeast Asia, West Africa, and the Americas. With over 16 years in the software industry, Dhavale has progressed from lead developer through solution architect and project manager to his current role, bringing rare depth across the full product stack, from system architecture to go-to-market strategy. He is an active contributor to SPE, with peer-reviewed conference papers on autonomous well optimization and edge-enabled production technologies. He holds 4 US patents, one granted and three pending, and a MSc in computer engineering from Pune University, India.
Zeshan Hyder is a product champion in the Agora Edge AI group at SLB, based in Houston, where he leads the development of edge-enabled solutions for energy assets, including artificial lift, flow assurance, and safety across well and facility operations. With more than 25 years of experience in the oil and gas industry, his career spans production engineering, operations, and digital solution development across both operators and service companies in domestic and international environments. His expertise includes a broad range of production-optimization technologies, particularly in artificial lift systems, with a focus on integrating advanced analytics, machine learning, and edge computing into field operations. Hyder has authored multiple SPE papers on autonomous optimization and edge-enabled production technologies. He holds a BSc in chemical engineering from Texas A&M University and an MBA from the University of Calgary.