A. Gambaretto, C. Kemp, M. Perezhogina, V. Er, D. Davalos, G. Martinez Loya, M. Nethi, and V. Salvi, SLB
R. Marin Nunez, Independent;
E. Gies, Expand Energy
Intermittent gas wells frequently suffer production losses due to liquid loading and the limitations of manual or SCADA-driven cycling. This work introduces an edge-native autonomous control system that optimizes liquid unloading and flowback behavior using a hybrid physics-based and machine-learning (ML) framework. Deployed directly at the wellsite on rugged IIoT gateways, the system continuously ingests surface pressure, temperature, and flow data to compute real-time gas velocity, critical velocity, and inferred liquid-column dynamics. These insights are used to automatically determine optimal shut-in timing and choke-actuation decisions without requiring cloud connectivity or operator oversight.
To enhance unloading efficiency, shut-in duration is predicted by a cloud-hosted ML workflow trained on pressure-buildup trends, cycle outcomes, and historical production behavior, producing tailored per-well recommendations that are executed autonomously at the edge. The combination of deterministic modeling, adaptive ML forecasts, and closed-loop decision logic eliminates reactive, calendar-based operation and reduces unnecessary downtime.
Field deployment across nine Haynesville wells demonstrated significant production uplift, with cumulative gas increases of 70–139% and average daily gains reaching 350 MCFD. The approach delivered over 80 MMCF of incremental gas per well annually while requiring minimal infrastructure changes. Results confirm that hybrid edge-cloud intelligence provides a scalable, low-cost pathway to modernizing intermittent well management, enabling production optimization, reduced emissions, and improved operational consistency across diverse asset conditions.