A Data-Driven Journey into Liquid Loading Detection and Prediction

Abstract Liquid loading is a persistent challenge encountered in onshore and offshore gas wells, particularly at low gas rates, where the accumulation of wellbore liquids leads to flow instabilities, operational disruptions, prolonged shut-ins leading to early well abandonment and reduced overall recovery. Typical liquid loading mitigation strategies such as velocity strings, well cycling, intermittent or permanent artificial lift, surfactants/soaps etc. involve additional cost and therefore, their operational success relies on selecting the right wells with proper understanding of expected critical rates at the right time. Models based on droplet, film or multiphase transient flow can be used to model liquid loading phenomenon. However, empirical correlations are used commonly to detect liquid loading, which often lack precision in field applications due to oversimplified assumptions regarding liquid behavior and flow regime consistency. In this work, we introduce an innovative data-driven approach for liquid loading detection and prediction (LLDP) that harnesses high-frequency gas rate and tubing head pressure measurements to identify the onset of liquid loading and use it to correct critical rates computed by empirical methods. The LLDP methodology involves the computation of diagnostic statistical proxy features, enabling the characterization of flow instability arising from liquid loading. Upon detection, subsequent adjustment to gas rates via feedback control facilitate the determination of corrected critical rate by calibration of empirical correlations such as Turner and Coleman. This also allows us to predict when the reservoir energy will no longer be sufficient to lift the liquids combined with the corrected critical gas curve and estimate time to liquid loading. Through extensive application across various unconventional and conventional gas fields, the LLDP method has consistently matched detected liquid loading events with manual field observations, eliminating biases and subjective interpretations. In a specific case study, the utilization of adjusted critical rates to optimize gas lift operations resulted in substantial cost savings by minimizing the need for buy-back gas. In another example, the detection of liquid loading and prediction helped optimize onshore gas well operations and field management practices. This presentation highlights a rapid and systematic approach for the detection, quantification, and prediction of liquid loading, utilizing readily available field data without relying on assumptions. By accurately identifying critical gas rates and optimizing artificial lift configurations, the LLDP method offers a practical solution to the challenges posed by liquid loading in gas wells, ultimately enhancing operational efficiency and maximizing production.