Liquid Loading Prediction in Deviated Wells: An Explorative Analysis of the Possibility of Complex Interplay Between the Entrained Droplet Model and Continuous Film Model

Liquid loading in gas wells remains a critical challenge, particularly in deviated and horizontal wells where multiphase flow dynamics become increasingly complex. Traditional models such as the Entrained Droplet Model (EDM) and the Continuous Film Model (CFM) have independently provided frameworks for understanding flow regimes and predicting critical velocities. However, recent investigations suggest that the interaction between these two fundamental models may not be entirely isolated, especially under varying well inclinations and flow conditions. This study aims to explore the potential complex interplay between the two models in deviated wells, proposing that a hybrid understanding could bridge the prediction accuracy gap for critical velocities. By leveraging empirical data and machine learning regression techniques, the analysis evaluates the transitions between droplet-dominated flow and film-dominated flow, offering insights into optimized flow predictions for gas wells unloading. Results show that by combining EDM and CFM, a more accurate prediction of the loading onset was achieved. Although the two models align in near-vertical wells, their predictions diverge at 30 - 60 degrees medium inclinations, where increased film thickness and reduced gravity raised the required critical velocity. In these cases, CFM tends to overpredict loading while EDM underpredicts it, indicating that neither model alone fully captures the onset of instability. By combining both criteria or applying an ML-assisted hybrid, prediction accuracy improves markedly, achieving R² values of ~0.93 for load classification and ~0.96 for critical-velocity regression. This confirms that an integrated EDM–CFM approach more reliably represents liquid-loading behaviour in deviated wells and enhances deliquification decision-making.