Answers are grounded in this paper and cite their sources.
Automated gas lift optimization improves oil production in the Permian Basin, indicating AI's significant role.
_ This article, written by JPT Technology Editor Chris Carpenter, contains highlights of paper SPE 219553, “Gas Lift Optimization in the Permian Using Machine Learning and Artificial Intelligence,” by Pooya Movahed, Derek Burmaster, SPE, and Emmanouil Karantinos, ExxonMobil, et al. The paper has not been peer reviewed. _ Gas lift optimization has historically been a time-consuming process. Automated gas lift optimization can add incremental volumes without the need for major expenditure. This paper presents a closed-loop iterative well-by-well gas lift optimization workflow deployed to more than 1,300 operator wells in the Permian Basin. The workflow conducts multirate tests through remote control of gas lift injection-rate set points in combination with automated well‑data acquisition. The optimization workflow has resulted in an average oil production uplift of approximately 2.0%. Methodology The workflow described in this paper is based on a core iterative process composed of the following steps: 1. Multirate-test scheduling 2. Multirate‑test‑data analysis 3. Downhole-pressure-model construction and storage 4. Optimal set-point calculation and delivery This process is repeated for each well on a roughly monthly basis, which allows the process to iterate continuously toward the optimal value while well conditions change. Between cycles, set points are updated based on new production-well-test data. Data are collected and transmitted between sites and a cloud-based computational hub through a communication infrastructure, as are the resulting injection-rate set-point changes. Finally, the downhole-pressure models (detailed in the complete paper) and their associated data are stored in a model inventory, used to train a machine-learning (ML) model to enable optimization of wells that are missing data. Data Collection and Analysis. Meaningful gas lift injection-rate variations are required to create a useful downhole-pressure model. The system creates this data by conducting three-step multirate tests. Three rates are checked to provide the data required to define a quadratic approximation of the gas lift performance curve (GLPC). The default duration for a single step is 15 hours, but the duration is increased to 24 hours for slugging wells. The gas lift rate is changed by 100,000 scf/D from the initial gas lift injection rates to ensure the variations in pressure are significant enough. Smaller changes were attempted but often resulted in downhole pressure changes that were insufficient to produce meaningful models. The complete paper includes subsections devoted to components of this task, including paired multirate testing, multirate test‑data handling and fitting, productivity‑index determination, and workflow quality control and automation. Communication Infrastructure. General Architecture. The system’s communication infrastructure serves twin purposes—receiving telemetry data from the multitude of distributed devices in the field for logic and modeling and sending out the resulting calculated values as set points to the related devices. To achieve this, the system makes use of several existing components—the local Permian controls network, oblivious transfer (OT) protocols, the message-queuing telemetry transport (MQTT) protocol, the supervisory control and data‑acquisition (SCADA) architecture, and near‑real‑time data flow to and from the cloud environment.
No takes yet. Share an insight, caveat, or question.
Chris Carpenter (2025) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: