Industrial Internet of Things Application Deployed for Liquid Unloading in Gas Wells
Chris Carpenter_
This article, written by JPT Technology Editor Chris Carpenter, contains highlights of paper SPE 229390, “Smart Liquid-Unloading IIoT Application for Gas Wells in the Haynesville Basin,” by Agustin Gambaretto, SPE, and Carl J. Kemp, SLB, and Rogelio M. Nunez, SPE, Consultant, et al. The paper has not been peer-reviewed.
_
Intermittent wells, particularly in gas-producing basins such as Haynesville, traditionally are managed with manual or partially automated controls, requiring frequent operator intervention and relying on static supervisory-control and data-acquisition (SCADA) infrastructure. This approach limits production efficiency and responsiveness to changing well conditions, especially in the presence of liquid loading. To address these limitations, this paper presents an autonomous, data-driven solution deployed at the edge, designed specifically for intermittent well optimization.
Introduction
This paper introduces a field-proven Industrial Internet of Things application that manages the liquid-unloading cycle autonomously using a data-driven, edge-based control system. Running directly on a local gateway device at the well site, the application ingests real-time pressure, flow rate, and temperature data continuously. It applies embedded physics-based models and lightweight machine-learning (ML) algorithms to calculate key flow parameters and dynamically adjusts choke settings in a closed-loop manner without requiring cloud connectivity or operator intervention.
Building on the architecture successfully deployed in artificial lift applications such as autonomous sucker-rod-pump optimization frameworks that leverage local intelligence to adapt rod-pump control logic without cloud reliance, this solution demonstrates how real-time control logic can be extended to intermittent gas wells to maximize production, minimize downtime, and enable scalable automation. Similar data-driven approaches have shown strong performance in gas lift optimization scenarios, where edge-deployed models adapt to variable flow regimes with minimal human tuning.
Methodology
Architecture Overview.
The smart unloading solution is built around an edge-deployed application running on a rugged field gateway. This device integrates directly with existing wellhead instrumentation (casing pressure, tubing pressure, temperature, flowrate, and choke actuator), requiring no modifications to upstream SCADA or communications infrastructure.
The edge application is composed of four tightly integrated layers:
- Data-Acquisition Layer: Continuously collects and validates telemetry from the wellhead, including pressure, temperature, and flow-rate sensors.
- Calculation Engine: This module computes flow parameters essential to understanding unloading behavior. It applies embedded physics-based models to derive gas velocity, critical velocity, and liquid-column height. These calculations are updated continuously to reflect changing wellbore conditions.
- Autonomous Decision Logic: The core control algorithm replaces static calendar-based cycling with real-time, condition-driven logic. It uses calculated parameters to decide autonomously when to shut in the well. Specifically, it monitors gas velocity vs. critical velocity thresholds and estimated liquid level and liquid-accumulation rate.
- ML-Based Shut-In Duration Estimation: Instead of applying a fixed shut-in time, the duration is computed dynamically by an ML model that could be deployed at the edge; however, to enhance scalability and agility, it was moved to the cloud. The model predicts the optimal shut-in duration to allow the required pressure recharge to resume effective flow.